ES2436845B1 - PROCEDURE TO MEMORIZE AND TO PLAY A CAREER CARRIED OUT BY A VEHICLE, AND SYSTEM THAT IMPLEMENTS IT. - Google Patents

Abstract

A procedure is presented to memorize a trajectory followed by a vehicle in order to be subsequently reproduced, by means of the procedure to reproduce a trajectory. A system for memorization and for reproduction carried out by the steps of the respective method is also shown. It is of main interest in the automatic assistance to parking vehicles in parking lots to free the driver from repetitive or complex tasks. Safety has been studied in the possible variation of the conditions with respect to those in which the trajectory and the feasibility of rectifications were memorized.

Description

image 1

PROCEDURE TO MEMORIZE AND TO PLAY A CAREER CARRIED OUT BY A VEHICLE, AND SYSTEM THAT IMPLEMENTS IT

TECHNICAL SECTOR

The invention falls within the driving assistance systems for automobiles. More specifically, it relates to parking assistance systems.

BACKGROUND OF THE INVENTION

The following documents related to the present invention are known in the state of the art: ES 2318725T discloses an auxiliary parking device and a procedure for parking assistance. A travel path is recorded during a journey of the motor vehicle from an initial point to an end point and then the motor vehicle is automatically driven back along the recorded travel path from the end point to the starting point. Orders are issued by the driver from a remote control system. It is able to reproduce the path from the starting point to the end point again. However, it raises a solution with a significant limitation for an expected total path length of approximately 5 meters. DE 10331948 discloses a maneuver assistance method, storing registered maneuvers and supporting the performance of stored maneuvers. He says it facilitates the performance of repetitive maneuvers, such as parking in a specific parking space, however it lacks details and explanations of the technological aspects implemented to achieve this result. In addition, it does not conceive the possibility of trajectory rectifications, e.g. before an obstacle, to later recover the memorized trajectory.

DESCRIPTION OF THE INVENTION

The present invention aims to free the driver from the need to perform tedious and repetitive tasks at the controls of a car while performing any type of repetitive maneuver. The vehicle can reproduce routes

image2

previously memorized completely autonomously, including long journeys of more than 100 meters, without the presence of the driver inside the vehicle. To achieve this, the present invention proposes a method and a system for memorizing a trajectory made by a vehicle according to claims 1 and 18 respectively. The object of the present invention is also a method and a system, both linked to the corresponding one mentioned above, to reproduce a previously memorized path according to independent claims 11 and 28 respectively. Preferred or advantageous embodiments are defined in the dependent claims.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1: Shows the sensors used. FIG. 2A-2C: Shows the field covered by different sensors in greater detail. FIG. 3: Shows a flow chart corresponding to the memorization of the trajectory. FIG. 4: Shows a flow chart corresponding to the reproduction of the memorized path. FIG. 5: Shows a flow chart corresponding to the reproduction of the memorized path in a second phase. FIG. 6: Shows a block diagram for an embodiment. FIG. 7: Shows a flow chart corresponding to the rectification of the path in the presence of an obstacle. FIG. 8A, 8B: Shows a drawing of two situations in which the scope of the measurements of the laser sensor 11 is observed during the reproduction of an obstacle course. FIG. 9: Shows a flow chart corresponding to the rectification of the path in the presence of an obstacle. FIG. 10: Shows an example of a parameterizable curve of the center of mass of the vehicle by avoiding an obstacle. FIG. 11: The nominal or memorized trajectory is displayed. FIG. 12: Corrected trajectory is displayed.

image3

FIG. 13: Shows the main addresses of the gradients of an area of interest for the calculation of descriptors.

DESCRIPTION OF A PREFERRED EMBODIMENT

The following is an example of an embodiment with reference to the figures. This example is for illustrative purposes and should not be considered as limiting the scope of the invention. The example of a path assistance system can be activated and deactivated manually via an interface, for example a touch screen. Deactivation can also be executed automatically in certain circumstances. For example: detecting a fault, if the driver presses the brake pedal, maneuvers the steering wheel or activates the electromechanical parking brake. The range of speeds and accelerations is expected to be limited to avoid both safety and comfort problems.

Memorization of the initial position and positions in the trajectory With the vehicle stopped, the characteristics of the environment are analyzed in order to determine distinctive characteristics of said position as the initial position of the trajectory to constitute position descriptors. Preferably, it will be based on obtaining references of the environment through a laser scanner or lidar 11 and an image sensor or camera 12. In Figure 3 a possible sequence of steps for memorizing positions can be observed. Identifying (to later place) the vehicle in the initial position is very important since the information provided by the inertial navigation unit (INS) allows only knowing the relative position of a body from linear acceleration and angular velocity values measured by accelerometers and gyros respectively. With the INS, the measurements for navigation are obtained in the form of relative coordinates, that is, with respect to the previous position. It is noted that a cumulative error occurs when obtaining data that lacks absolute references. In general, both in the initial position and in positions during the trajectory, to obtain the descriptors of a position, the image obtained by the camera 12 must be processed and vertical edges (corners, edges, sides, etc.) greater than one threshold (eg 100 pixels) through a processor 10. It has been proven that these elements have more options to be characteristic and invariant references to changes in brightness, translation, scale ... These edges can be identified by edge detection techniques between which are worth mentioning Sobel, Roberts, Prewitt, Frei-Chen or Canny. On them measurements are made with the laser sensor to collect the relative distance. In the case of the initial position, a detection of invariant points that hardly change in the environment must be carried out additionally. By choosing invariant points that belong to edges, greater reliability is achieved. There are several techniques that allow to obtain the desired results. These include SIFT, Shi-Tomasi corner detector or Harris corner detector. Thus, the transformation of the existing information into an image is achieved and invariant coordinates are obtained against changes in scale, orientation, lighting changes, translations and rotations. This meets the requirement to identify exactly the initial position. Also in the case of establishing the initial position it is advisable to use at least 2 different references to locate the vehicle later for the reproduction of the memorized trajectory. As mentioned, the initial position is where the memorized maneuvers are to be repeated when the trajectory has to be reproduced in the future. For security, it may be advisable that the user additionally decides whether such references are correct or not taking into account the objects in the environment they represent. For example, if some features do refer to mobile or changing elements (cars, people, vegetables ...) the user will decide that such references are not satisfactory. An interface 18 can serve this validation function. The degree of requirement during the memorization of non-initial points of the trajectory is lower. There is no user validation when the vehicle is moving memorizing the trajectory. It is not necessary to identify invariant points during trajectory capture. The identification of these invariant points would have to be done in real time which would excessively complicate the process capacity requirements necessary for their implementation. In sum, it is enough that during the duration of the trajectory, edges are captured and relative distances are measured with them at time intervals.

image4

image5

As a common element, both the initial positioning and the correction of the successive positions of the trajectory will require at least two characteristic elements of the image to make an exact positioning of the vehicle with respect to the environment. Consequently, at least two descriptors of each position will be required. Each path descriptor contains visual information of the detected vertical edge and its relative position. A linear correspondence between the horizontal field of view of the laser sensor 11 and the horizontal field of view of the chamber 12 can optionally be included by a parallel between the longitudinal axes of both. This point is advantageous to omit the inclination of the road in the calculation of separation distance to vertical axes and distance to be traveled by the vehicle. The distances in the plane through which the vehicle moves are calculated, thus obtaining simpler movement estimates. The fusion of the 2D laser map and visual images begins with the definition of the region of interest of the images, where the objects detected by the laser 11 are represented on the 2D map. Data fusion based on the linear correspondence between the horizontal field of view of the laser sensor 11 and the horizontal field of view of the vision sensor 12 or camera is implemented. Consequently, the 2D plane representing the angle-distance of the environment captured by the laser sensor 11, will correspond linearly to the X pixels wide of the visual image. If the results of the fusion are satisfactory, the descriptors with the characteristics (position and orientation) of the axes chosen as representative of this image are stored in a memory 30. This information will be consulted by the processor 10 to perform the positioning in the reproduction of the posterior path. Together, it can be collected by means of three gyroscopes 14 and three accelerometers 15, spatial information relative to the successive points of the path for better characterization. Information can also be incorporated with a linear speed sensor 16 and angular 17. During the path memorization, additionally, the lateral distance of the vehicle to possible obstacles can be measured by an ultrasonic sensor 13 similar to how it is done with laser sensors 11. The scope of an ultrasonic sensor 13 is of the order of several meters, about 2 meters to maintain good measurement accuracy. Generally, vehicles that have an inertial navigation unit (INS) can perform these measurements. The main advantages of an INS compared to other navigation systems are the accuracy, frequency of data collection and coverage. Obtaining measurements in relative coordinates does not present a problem because the need is to reproduce a trajectory from the initial position, regardless of its absolute location. In the same way, an incorrect installation of the measuring devices in the vehicle, even providing incorrect measurements, does not affect the correct operation since both the measurement and the reproduction will be carried out with the same error. In figure 1 you can see an example where the location of some of the sensors used in the vehicle itself is illustrated. As the main disadvantage to take into account the INS for proper operation, is the cumulative error already mentioned. In order that the system is not significantly affected, the process described above is used, which is based on obtaining references during the journey to be able to parameterize and correct said error. In the absence of such reference points, reproduction of short or medium distance paths will be imposed as an essential condition. In this way, the cumulative error will not be a significant value that prevents the satisfactory reproduction of the memorized trajectory.

image6

Reproduction of the memorized trajectory Once the trajectory has been memorized, the conditions so that it can be repeated without the driver's assistance are set in the first instance by recognizing the memorized initial position and validating that the vehicle is located on said position with little deviation. Then, the maneuvers performed by the vehicle can be reproduced when said trajectory was memorized. The comparison of the current position against the memorized position is therefore very important. To do this, the descriptor stored in memory 30 corresponding to the starting point of the path must be retrieved. In turn, a current image of the environment of the vehicle must be collected by the camera 12. The image must be processed with the processor 10 to identify the vertical edges that meet the required conditions (eg be greater than a threshold) and then establish these vertical borders as current references. Once established, with the laser sensor 11 the relative position of each vertical edge with respect to the vehicle is measured and the current descriptors can be generated with visual information of each detected vertical edge and its measured relative position. The comparison is made between said current descriptors with the descriptors of the starting point of the trajectory. This can confirm whether the current position is correct to start the maneuver's reproduction. If this is not the case, the necessary corrections can be made to adjust the current initial position to the one memorized and with it that the successive maneuvers that entail tracing the memorized trajectory can be carried out with precision, comparing the current and memorized relative distance. Figures 4 and 5 show a possible sequence of steps for the process of reproducing a memorized path. When it is estimated that the current initial position is valid to begin the reproduction of the path, the processor 10 through a controller 20 will order the different actuators to carry out the necessary maneuvers. It is therefore necessary that there is communication with the actuators responsible for the admission of the fuel 21, on the brake 22, on the parking brake 23 and, of course, on the direction 24. Simultaneously, the ultrasonic sensor 13 can remain activated for warn of possible obstacles at a predefined lower distance. In the presence of an obstacle, the movement associated with the movement to the next point of the memorized path can be interrupted for a predefined period of time. This avoids aborting the maneuver in case the obstacle is temporary (e.g. a pedestrian crossing). It is estimated that an appropriate waiting interval is less than 1 minute, preferably 20 s. Otherwise, if the presence of the obstacle is maintained after the predefined time interval, the laser sensor 11 measures the relative position of the obstacle with respect to the vehicle and its width and depth are estimated to calculate the feasibility of circumventing it based on successive relative displacements. of the trajectory, of the minimum distance of the vehicle to obstacles arranged in the nominal trajectory and of the minimum available lateral space obtained by the lateral ultrasound sensors 13. Subsequently, this specific case will be treated.

image7

Mathematical description The definition of the position of the vehicle in each data update is necessary to plot the trajectory and to make corrections in it. The position of the vehicle is defined by means of a status vector. Two parts can be distinguished, the estimated state vector of the vehicle and the definition of the environment marks. This description includes:

image8

- Position of the vehicle with respect to the previous position, Xveh - Orientation of the vehicle with respect to the previous position, Ωveh - Linear speed of the vehicle in the center of the rear axle, Vveh - Angular speed of the vehicle, ωveh

5 Of the previous parameters that define the instantaneous kinematics of the vehicle, the variation of the angular speeds are obtained directly from an inertial measurement unit, for example of gyroscopes 13. The obtaining of the linear speed (V) of the vehicle is also immediate to through wheel rotation sensors or angular velocity sensors 17.

10 The kinematic parameters are obtained from:

image9

It follows that with the proposed integrations the variation of

15 position and angle to the immediately previous position. For reproduction under normal conditions of the nominal path these values are sufficient. On the other hand, to make rectifications it will be essential to know these values globally, that is, to be able to characterize each position independently of the

20 previous positions. This positioning in each data update "i" will be carried out depending on the initial position of the vehicle. In this way:

image10

25 Secondly, those parameters that will allow the search of correspondences between characteristic elements of an image that must also be defined are defined

image11

be memorized These elements will be used as references of the nominal trajectory. As a summary, the descriptor of a characteristic point of the nominal initial position comprises at least the following information:

•

Location in the image of the point of interest

•

Main orientations of the region studied

•

Distance to vertical axis

•

Vertical axis orientation

In addition, the descriptor of a characteristic axis of any position of the nominal path comprises at least the following information:

•

Average horizontal position of a vertical axis in the image

•

Distance to vertical axis

• Vertical axis orientation Distance and vertical axis orientation are obtained directly from the laser scanner (responsible for knowing the relative position of the marks). Due to the negligible vertical field of view, the appearance of the vertical axes and the identical vertical orientation between the vehicle and the laser, the orientation angle of the marks in this direction is disregarded (θz → 0). Therefore, in this definition it is assumed valid to work in a 2D plane. For reasons of comfort of the occupants and also to guarantee the operation, a maximum speed must be established that allows both the correct acquisition of data and its complete processing, as well as the comfort of the occupants. Likewise, a maximum acceleration must be established. Preferably, the maximum acceleration value is 0.2 g; the path braking value is -0.2 g; The emergency braking value is -0.5 g, where g is the acceleration of gravity. Regarding the speed, it is recommended that it be the minimum of the following speeds:

During the memorization of the trajectory, the instantaneous vehicle speed for each position is also recorded. It can be established as the first premise that the speed of the vehicle in the reproduction of a trajectory is lower than the speed of the vehicle driven during the recording stage.

image12

In order to make a correct perception of the environment, some sensors specify an optimal working range depending on the absolute speed of the vehicle. At speeds greater than those specified, the manufacturers do not guarantee the correct perception or the correct functioning of the measuring devices, so a maximum operating speed must be set. Another parameter to consider is the distance traveled by the vehicle during braking. It must be ensured that, before the perception and detection of an obstacle in the path to be memorized, the vehicle must be stopped before a collision, at the most, performing an emergency braking. Therefore, it should be evaluated that the braking distance is significantly less than the range of the sensors integrated in the vehicle. There may be other situations where the vehicle speed must be reduced so that the quality of the measurements obtained by the devices or sensors implemented is optimal. The inclination of the plot with respect to the horizontal plane is going to be one of the parameters to take into account. A displacement with a remarkable slope (such as parking ramps) can imply sudden changes in grade. In addition, in descents, it can mean an increase in braking distance. Another situation of uncertainty may come from the sinuosity of the trajectory. In sharp curves the absence of obstacles on the road cannot be guaranteed since the arrangement of the environment can limit the field of vision of the perception devices. In summary, vehicle braking must be guaranteed in situations of uncertainty such as those presented above.

Generation of the alternative path An embodiment for the calculation and execution of the alternative path is described. 1) Upon detection of an obstacle that prevents progress along the nominal path, the vehicle is ordered to stop. Depending on the distance to the obstacle, the vehicle is decelerated with a comfort braking (-0.2g) or an emergency braking (-0.5g). 2) Once the vehicle is immobilized and with the parking brake applied, it is advisable not to take any action for a while. This is due to the fact that, in certain situations, the obstacle may deviate from the trajectory (another car driving, a person walking, an object that someone has removed, etc.). The waiting time is preferably set at 20 seconds. Once this time has elapsed, if the obstacle remains, it is parameterized using the values obtained by the laser scanners. The characteristic points of the obstacle are the ones with the greatest angle, since they will be the ones that will mark a first approximation of the space occupied by it. A maximum of 2 values (distance, angle) are obtained from this simplification. Thus, lidar 11 measures the left extreme obstacle: [dobsi; βobsi]; and right end of the obstacle: [dobs; βobsd].

image13

3) Once the position of the ends is known (in the horizontal field of view of the scanner), it is assessed whether rectification is feasible based on:

•

The evolution of the nominal trajectory, that is, the direction in which the trajectory evolves once the obstacle has been overcome.

•

The width and depth estimation of the obstacle (it is proposed to consider a depth of 2.6 times the total width of the obstacle).

•

Sufficient maneuverability available space (obtained by lateral ultrasound sensors).

•

Direction of the most feasible maneuver (left or right maneuver).

4) If rectification is feasible, choose the direction on which to begin rectification and calculate the relative position of the obstacle:

image14

5) Once the parameters that define the location of the end of the obstacle are known, the rectification to be executed is determined. It can be determined by a kinematic control, which is based on converting the movement specification into an analytical trajectory in space. Specifically, a control based on parametric curves is used. These curves must meet the vehicle's movement restrictions. As a main drawback, the small errors made during the computation and reproduction of the curve accumulate. In order to avoid or minimize the possible deviation produced, the parametric curve to be followed is recalculated from time to time.

image15

It is important to keep in mind that this process can result in multiple possible trajectories, as well as the possibility of the absence of a feasible solution. 6) Due to the chassis characteristics and steering angle, not all the paths calculated through the equation y = f (x) will be able to be executed by any vehicle. You should know the maximum curves that a vehicle in question can execute and compare if the execution of a curve is feasible. 7) If the main cause that makes it impossible to rectify a path is analyzed (assuming that the surface is flat, there are no changes in grade and the lateral distance is sufficient), this is that the distance-to-cross distance of the obstacle relationship is greater than acceptable. A previous displacement of reverse gear is proposed, where the following will be taken into account: - collision surveillance in the rear area - calculation of the available lateral space (lateral ultrasound sensors) - deviation from the nominal path - subsequent movement in phases of 0.5 meters After the subsequent movement, the rectification capacity is assessed by analyzing the available lateral space and obtaining the parameters (X0; Y0) of the obstacle. Four possible solutions are proposed: a) The rectification is feasible and the movement begins. b) The rectification is not feasible due to lack of space to the obstacle and the feasibility of a new subsequent displacement is raised c) It is not possible to carry out the subsequent displacement and the entire process is aborted. 8) In the situation where rectification is feasible, the calculated trajectory is executed. Once finished, the vehicle is parallel to the obstacle or has passed it. In the case of exceeding it, the process of recovering the nominal trajectory (described in the following points) would begin. Otherwise, the vehicle approximately reproduces the outer lateral surface of the obstacle thanks to the distance values obtained by the lateral and frontal ultrasound sensors 13. As in the previous case, once the obstacle has been overcome, recovery of the memorized trajectory will begin. Figures 7 and 9 show the sequence of processes and decision-making executed in a rectification.

image16

An example of a parametric curve that could be used to generate the trajectory on a flat surface that the vehicle will follow during the rectification process is the following:

image17

where X1 is the space traveled by the vehicle during rectification in the direction of the longitudinal axis [m] and Y1 is the space traveled by the vehicle during rectification in the direction perpendicular to the longitudinal axis [m]. The graph in Figure 10 shows an example of the maneuver in question (X1 = 3; Y1 = 2.3). The main features of this fifth grade polynomial are:

•

Describe the trajectory of the center of mass of the vehicle.

•

Start and end of the trajectory with the vehicle oriented towards its longitudinal axis.

•

Progressive variation of the steering wheel turning angle allowing continuous and uniform speed control. This will prevent vehicle stops.

Recovery of the nominal trajectory Once the vehicle has passed the obstacle, the system intends to resume the nominal trajectory. This will be a complex and delicate operation since the system does not have references to follow, but a trajectory with which to connect. It is proposed to solve the problem, at a conceptual level, by means of a kinematic control similar to that proposed in the “generation of the nominal trajectory”, which is based on converting the movement specification required by the “Virtual driving module”. Specifically, a control based on parametric curves is used.

The main difficulty of this process lies in creating a closed model that allows the recovery of the nominal trajectory before a wide range of situations and possible destination points. It is recalled that the execution of the rectifications was imposed on almost flat surfaces and without changes of grade or notable potholes. There are several ways to solve the aforementioned problem. The first conflict arises when defining the target point of the nominal trajectory. It should be taken into account that its location with respect to the current position of the vehicle is unpredictable, as is the orientation of the vehicle in that position. It should also be taken into account that the kinematic characteristics of the vehicle do not allow the realization of any type of curve.

image18

A possible embodiment is presented below. The initial challenge appears in the development of a pattern that makes it possible to narrow the longitudinal distance between this point and the vehicle, which is sufficient to allow maneuverability but, at the same time, not excessive. Figures 11 and 12 show some parameters to take into account. Once the longitudinal distance is defined, the memory 30 searches for those points of the memorized path that are in a line perpendicular to the current orientation φr of the vehicle at a distance X1. Figure 11 shows the line t, at a certain longitudinal distance, where the point of the nominal trajectory with which it is desired to link (Xfnom Yfnom) is found. The point obtained will be the destination point where the nominal trajectory will resume. Its coordinates (Xfnom Yfnom) and vehicle orientation (φfnom) will be recovered. At this point it is intuited that the problem lies in developing a parametric curve suitable for most cases. This conflict could be resolved by complex dynamic control models, by Béizer curves, Hemite curves, splines, combination of parametric curves. A possible simple embodiment would be to connect the current point with the target point tangently to its orientation with an arc of circumference and a fifth degree polynomial. In this way, it is intended to cover a large multitude of points along the line (t). By means of the defined radius circumference arc, the vehicle is intended to be oriented so that its longitudinal axis is parallel to the orientation of the known destination point (φfnom). On the other hand, the use of the fifth degree polynomial will allow the vehicle to reach the target position with continuous and uniform control of speed and direction. The same fifth grade polynomial presented above is proposed. Figure 12 shows the raised trajectory composed of two parametric curves, differentiating the initial position (Xr0 Yr0), the intermediate position resulting from the defined angle rotation (Xr1 Yr1) and the final position (Xfnom Yfnom) which belongs to The nominal trajectory.

Image processing Once the invariant points of an image are identified, which are in a vertical axis of length greater than a threshold value, we proceed to obtain those characteristics that will allow us to identify it in future occasions.

image19

For this, a study area is defined around each characteristic point (as an example an 8x8 pixel matrix). Once this region is defined, the gradient of each pixel (vector) is calculated, which comprises a module and an orientation. The next step is to make a weighting of the orientations of the gradients according to their module in order to obtain one or several main orientations of the region. Figure 13 shows a 4x4 matrix that encompasses the main directions included. The visual descriptor can be obtained by calculating the gradient of a characteristic point. The information stored in this visual descriptor will be: -Location in the pixel image (point of interest). -Main orientations of the region studied. The reason for memorizing the location of the pixel in the image is that the computational cost of the comparisons could become very high if restrictions are not set. Therefore, if we know that certain areas of an image A must appear in a similar area in image B, the search is centered on a certain area. To calculate the degree of similarity between two visual descriptors (set of elements that indicate the most important orientations around a point of interest) there are several alternatives. Performing a series of operations between the visual descriptor of the image A with each of the image B will know which are most likely corresponding. At this point the necessary information has been memorized in order to identify this point in an image in future phases of initial positioning. However, for this phase to be satisfactory it will be necessary to consult the relative position data provided by the laser scanner 11. The stored positioning data will be: -Distance to the vertical axis. - Relative orientation of the axis with respect to the longitudinal axis of the device. As a summary, the descriptor of a characteristic point of the nominal initial position comprises at least the following information: -Location in the image of the point of interest. -Main orientations of the region studied. -Distance to vertical axis. - Orientation to the vertical axis.

image20

Obtaining the characteristics that define a vertical axis and allow its subsequent identification is not as robust as that of the invariant points. The reason lies in its own constitution, since neither gradients nor orientations make sense. One embodiment is the memorization of the gradient defined at the extreme points of a vertical edge. However, it is ruled out due to the possible defragmentation of the axes and the associated computational cost. Therefore, it is preferred to use a less robust definition but of lower computational cost, which consists in memorizing the average horizontal position of the vertical axis in the image. By itself, this definition would be ineffective for two similar images. However, given the uniqueness of dealing with a succession of sequentially arranged images and knowing the approximate area where a vertical axis should be located in the image, acceptable behavior could be obtained. As is known, the process of associating points on vertical lines accepts as valid almost vertical lines. In this fact lies the need to calculate the average horizontal value of a vertical axis resulting from a horizontal calibration of the vision device. In a simplified way:

∑ x

x = i ∀ xi ∈ l

n

Where n is the number of pixels evaluated and l any border. As in the case of descriptors of invariant points, for a correct positioning it will be necessary to consult the relative position data provided by the scanner. As a summary, the descriptor of a characteristic axis of any position of the nominal path comprises at least the following information: average horizontal position of a vertical axis in the image, distance to the vertical axis and orientation to the vertical axis.

Claims (30)

image 1 1.-Procedure to memorize a trajectory made by a vehicle characterized in that it comprises the following steps: -to collect at least one image of the vehicle environment at a point in the path, by means of at least one camera (12), -process the image to detect the presence of at least two essentially vertical edges in the image greater than a predefined threshold, -measure, by means of a sensor (11), the relative position of each essentially vertical edge with respect to the vehicle at said point of the path, -generate a descriptor of the path point with visual information of each essentially vertical edge detected and its measured relative position, -memorize each descriptor of the path point. 2-Procedure for memorizing a path according to claim 1, characterized in that the step of processing the images of the vehicle environment, when stopped to start the path, comprises establishing at least two different descriptors of the starting point of the path. 3-Procedure for memorizing a path according to claim 2, characterized in that the step of establishing a descriptor of the starting point of the path comprises detecting at least one invariant point on an essentially vertical edge of the collected image. 4. Procedure for memorizing a path according to claim 3, characterized in that the detection of invariant points is carried out by at least one of the following techniques: -Scale Invariant Features Transform (SIFT), -Shi-Tomasi corner detector, -detector from Harris corners. 5. Procedure for memorizing a trajectory according to claim 1, characterized in that it comprises at least one additional step of confirmation by the user image2 6. Method for memorizing a trajectory according to claim 1, characterized in that a plurality of descriptors of successive points of the trajectory are memorized at intervals of time during the duration of the trajectory. 7. Method for memorizing a trajectory according to any one of the preceding claims, characterized in that it establishes a descriptor of the trajectory point with visual information of the essentially vertical edge detected and its relative position comprises performing a linear correspondence between the field of vision essentially horizontal of the sensor (11) and the essentially horizontal field of view of the camera (12) by a parallel between the longitudinal axes of both. 8. Method for memorizing a path according to claim 6 or 7, characterized in that it comprises collecting, at time intervals by at least one gyroscope (14) and at least one accelerometer (15), spatial information relating to a plurality of points successive of the trajectory. 9.-Procedure to memorize a trajectory according to any one of the previous claims, characterized in that the images are processed to identify a vertical edge by at least one of the following techniques: -Sobel, -Roberts, -Prewitt, -Frei-Chen, -Canny. 10. Method for memorizing a trajectory according to any one of the preceding claims, characterized in that it comprises measuring the lateral distance of the vehicle to obstacles during the trajectory by means of at least one distance sensor (11,13). 11.-Procedure for reproducing a memorized trajectory of a vehicle characterized in that it comprises the following steps: -recovering a descriptor of the starting point of the memorized trajectory, -collecting at least one image of the vehicle environment by means of at least one camera (12 ), image3 -process the image to identify at least two essentially different vertical edges greater than a predefined threshold and establish said vertical edges as current references, -measure, by means of a sensor (11), the relative position of each essentially vertical edge with respect to the vehicle, - generate at least two current descriptors with information on each vertical edge detected and its relative position measured with respect to the vehicle itself, - compare said current descriptors with the descriptors of the starting point of the memorized path, - move the vehicle based on the result of the previous comparison. 12-Procedure for reproducing a path according to claim 11, characterized in that the step of generating a descriptor of the starting point of the path comprises detecting at least one invariant point comprised on a vertical edge of the collected image. 13. Method for reproducing a memorized path according to claim 11, characterized in that the step of comparing the current descriptor with the descriptor of the start point of the memorized path comprises in turn comparing the current and memorized relative distance between said vertical edges and the vehicle itself. 14. Method for reproducing a memorized path according to any one of claims 11 to 13, characterized in that, if the result of the comparison of at least two descriptors coincides with a predefined threshold, the vehicle moves to a next point of the memorized trajectory 15. Method for reproducing a memorized path according to any one of claims 11 to 14, characterized in that it comprises measuring at least one sensor (13) (11) the vehicle environment at a point on the current path, to detect the presence of obstacles at a distance less than a predefined threshold to define the braking conditions to be performed. image4 16. Procedure for reproducing a memorized path according to any one of claims 11 to 15, characterized in that in the presence of an obstacle, the movement is interrupted for a predefined period of time. 17. Procedure for reproducing a memorized trajectory of claim 15, characterized in that if the presence of the obstacle is detected after the predefined time interval, the sensor (11) measures the relative position of the obstacle with respect to the vehicle and its width and depth to calculate the feasibility of circumventing it based on - successive relative displacements of the memorized trajectory, -minimum distance from the vehicle to obstacles arranged in the memorized path, - Minimum available lateral space obtained by one of the lateral distance sensors (11,13). 18.-System to memorize a trajectory made by a vehicle characterized in that it comprises: -mediates to collect at least one image of the vehicle's surroundings at a point in the trajectory, by means of at least one camera (12), -mediates to process the image to detect the presence of at least two vertical edges in the image greater than a predefined threshold, - means for measuring, by means of a sensor (11), the relative position of each vertical edge with respect to the vehicle at said point of the memorized path, -a processor (10) configured to generate a descriptor of the path point with visual information of each essentially vertical edge detected and its relative position measured with respect to the vehicle itself, -a memory (30) configured to store the descriptor of the point of the trajectory. 19. System for memorizing a trajectory according to claim 18, characterized in that the processor (10) is configured to establish at least two different descriptors of the starting point of the trajectory in the images of the environment of the car, when said car is stopped for Start the trajectory. image5 20. System for memorizing a path according to claim 19, characterized in that the processor is configured to detect at least one invariant point on an essentially vertical edge of the collected image. 21.-System for memorizing a path according to claim 20, characterized in that the processor is configured to detect invariant points by at least one of the following techniques: -Scale Invariant Features Transform (SIFT), -Shi-Tomasi corner detector, - Harris corner detector. 22. System for memorizing a path according to claim 20 or 21, characterized in that it comprises an interface (18) configured to establish confirmation by the user. 23. System for memorizing a trajectory according to claim 18, characterized in that the memory (30) stores, at time intervals during the duration of the trajectory, a plurality of descriptors of successive points of the trajectory. 24. System for memorizing a path according to any one of the preceding claims 18 to 23, characterized in that the processor is configured to make a linear correspondence between the essentially horizontal field of view of the sensor (11) and the essentially horizontal field of vision of the chamber (12) by means of a parallel between the longitudinal axes of both. 25. System for memorizing a path according to claim 23 or 24, characterized in that it comprises at least one gyroscope (14) and at least one accelerometer (15), configured to collect relative spatial information a plurality of successive points of the path to timeslots. 26. System for memorizing a path according to any one of the preceding claims 18 to 25, characterized in that the processor (10) is configured to identify an essentially vertical edge by at least one of the following techniques: image6 -Sobel, -Roberts, -Prewitt, -Frei-Chen, -Canny. 27. System for memorizing a trajectory according to any one of the preceding claims 18 to 26, characterized in that it comprises at least one ultrasonic sensor on each side of the vehicle (13) configured to measure the lateral distance of the vehicle to obstacles during the trajectory. . 28.-System to reproduce a memorized trajectory of a vehicle characterized in that it comprises the following steps: -mediates to retrieve the descriptor of a memorized trajectory starting point, -mediates to collect at least one image of the vehicle environment by means of the at least one camera (12), -at least one processor (10) configured to process the image to identify at least two different essentially vertical edges greater than a predefined threshold and to establish said essentially vertical edges as current references, -means for measuring, by a sensor (11), the relative position of each vertical edge with respect to the vehicle, where the processor is further configured to generate at least two current descriptors with visual information of each essentially vertical edge detected and its relative position measured with respect to the vehicle itself, and to compare said current descriptors with the descriptors of the starting point of the trajectory, -a actuators (21, 22, 23, 24) configured to move the vehicle based on the result of the previous comparison. 29-System for reproducing a path according to claim 28, characterized in that the processor (10) is configured to calculate at least one invariant point comprised on an essentially vertical edge of the collected image. image7 30. System for reproducing a memorized path according to claim 28, characterized in that the processor (10) is configured to compare the current and memorized relative distance between said vertical edges and the vehicle itself. 31. System for reproducing a memorized path according to any one of claims 28 to 30, characterized in that the processor (10) is configured, if the result of the comparison of at least two descriptors is positive, to send instructions to the actuators ( 21, 22, 23, 24) to move the vehicle to a next point on the memorized path. 32. System for reproducing a memorized path according to any one of claims 28 to 31, characterized in that it comprises at least one sensor (13) (11) configured to detect the presence of obstacles at a distance less than a predefined threshold to define the braking conditions to be performed. 33. System for reproducing a memorized path according to claim 32, characterized in that in the presence of an obstacle, the movement associated with the displacement to the next point of the memorized path is interrupted for a predefined period of time. 34. System for reproducing a memorized path according to any one of claims 28 to 33, characterized in that the sensor (11) is configured to measure the relative position of the obstacle with respect to the vehicle and estimate its width and depth and transmit the information to the processor (10) to calculate the feasibility of circumventing it based on - successive relative displacements of the memorized trajectory, -minimum distance from the vehicle to obstacles arranged in the memorized path, - minimum available lateral space obtained by the lateral distance sensors (11) (13)