An automatic parking path planning system based on image recognition

Abstract

The present invention relates to the technical field of parking path planning, and specifically to an automatic parking path planning system based on image recognition, which includes: a parking demand condition extraction module combines the current vehicle position area corresponding to the target vehicle in the constructed three-dimensional environment model to obtain the parking calibration target area of the target vehicle, and generates parking demand conditions based on the relative position information of the current vehicle position area. The present invention not only takes into account the environmental information around the target vehicle to achieve accurate locking of the parking area, but also uses a multi-process parallel analysis method, and at the same time adopts a variety of planning methods to implement the planning of the parking path based on different data sources, realizes real-time feedback on the automatically planned parking path, and improves the response speed and response result accuracy of the automatically planned parking path.

Description

Technical Field

[0001] The present invention relates to the technical field of parking path planning, and specifically to an automatic parking path planning system based on image recognition. Background Art

[0002] With the rapid development of autonomous driving technology, the automatic parking system, as an important part of autonomous driving technology, has received extensive attention and research. The automatic parking system realizes the automatic parking function of the vehicle without human intervention by integrating a variety of sensors (such as cameras, radars, ultrasonic sensors, etc.) and advanced algorithms. Among them, the automatic parking path planning system based on image recognition has become a research hotspot due to its intuitive and efficient characteristics.

[0003] Existing automatic parking path planning systems based on image recognition often use image recognition technology to collect the position information of parking spaces and obstacles around the vehicle, and directly plan the corresponding automatic parking path through a path planning algorithm. In this method, the influence of incomplete parking markings collected on the locking of the parking position is not considered; and the dynamic update of obstacle information is not realized according to the environmental information around the locked parking position, thus affecting the subsequent planning result of the automatic parking path; at the same time, there is a single situation in the existing automatic parking path planning method, which cannot effectively balance the response speed and the accuracy of the response result of the automatic parking planning path; therefore, there are relatively large defects in the existing technology. Summary of the Invention

[0004] The purpose of the present invention is to provide an automatic parking path planning system based on image recognition to solve the problems raised in the above background art.

[0005] To solve the above technical problems, the present invention provides the following technical solution: An automatic parking path planning system based on image recognition, the system includes: a parking environment model building module, a parking area dynamic locking module, a parking demand condition extraction module, and a parking path intelligent planning module.

[0006] The parking environment model building module constructs a three-dimensional environment model around the target vehicle by collecting the environmental information around the vehicle in real time through an in-vehicle radar and a high-definition camera built in the target vehicle.

[0007] The parking area dynamic locking module identifies the position information of obstacles and the position information of parking markings in the three-dimensional environment model around the target vehicle, and locks the parking calibration target area of the target vehicle.

[0008] The parking demand condition extraction module combines the current vehicle position area corresponding to the target vehicle in the constructed three-dimensional environment model, obtains the relative position information of the parking calibration target area of the target vehicle based on the current vehicle position area, and generates parking demand conditions;

[0009] The parking path intelligent planning module uses a multi-process method to parallelly analyze the parking path schemes corresponding to the first parking path planning method and the second parking path planning method, generates parking planning path feedback information and feeds it back to the target vehicle display terminal. After the driver confirms, the target vehicle is automatically parked according to the confirmed parking planning path;

[0010] The first parking path planning method obtains the historical parking records of the target vehicle, calculates the adaptation values of the parking demand conditions and each historical parking record, and obtains the historical parking path planning scheme according to the adaptation values;

[0011] The second parking path planning method obtains the parking demand conditions and the obstacle position information in the three-dimensional environment model around the target vehicle, and generates the optimal reverse path planning solution of the target vehicle position based on the obstacle position information through an iterative method to obtain the optimal parking planning path of the target vehicle.

[0012] In the present invention, when planning the automatic parking path of the target vehicle, not only the environmental information around the target vehicle is considered to accurately lock the parking area, but also through a multi-process parallel analysis method, multiple planning methods are simultaneously used to plan the parking path based on different data sources (the first parking path planning method and the second parking path planning method), realizing real-time feedback on the automatic parking planning path, improving the response speed and response result accuracy of the automatic parking planning path; in the present invention, as the response time increases, the response result accuracy of the automatic parking planning path will also continuously improve, and during the response process, the driver of the target vehicle can independently select the automatic parking planning path according to his own needs (response time requirement and parking path planning method requirement), realizing effective management of the automatic parking path planning data.

[0013] Furthermore, the environmental information includes images and videos collected by the built-in vehicle-mounted radar and high-definition camera on the target vehicle within a preset radius area around the target vehicle;

[0014] The three-dimensional environment model around the target vehicle is constructed based on the collected environmental information within a preset radius area around the target vehicle and through image recognition technology and 3D generation technology of 2D images.

[0015] Further, the obstacle position information includes vectors formed by each position point in the three-dimensional environment model around the target vehicle, where the difference in height corresponding to the reference plane is greater than a preset height, based on the center point of the target vehicle; the preset height is the minimum height difference between the target vehicle and the ground; the reference plane is a plane parallel to the plane where the bottom crossbeam of the target vehicle is located and with a corresponding height lower than the preset height of the plane where the bottom crossbeam of the target vehicle is located;

[0016] The position information of the parking markings includes each parking marking area identified by image recognition technology in the three-dimensional environment model around the target vehicle and vectors formed by each position point in each parking marking area based on the center point of the target vehicle;

[0017] When locking the parking calibration target area of the target vehicle, obtain the minimum parking space area corresponding to the target vehicle preset in the database, denoted as the reference contour area; calibrate the parking marking position information in the three-dimensional environment model around the target vehicle according to each parking marking contour model preset in the database to obtain the parking marking position calibration information;

[0018] Randomly enclose reference contour areas at different positions in the three-dimensional environment model around the target vehicle, and the intersection of each random enclosure scheme for the reference contour area and the corresponding obstacle position points in the obstacle position information is empty; obtain the calibration matching value corresponding to each random enclosure scheme for the reference contour area, denoted as BD; the locking result of the parking calibration target area of the target vehicle is the reference contour area in the random enclosure scheme with the largest calibration matching value;

[0019] The BD = MU / MZ + r·L,

[0020] where MU represents the intersection area between the reference contour area in the corresponding random enclosure scheme and the area corresponding to the parking marking position calibration information; MZ represents the area corresponding to the parking marking position calibration information; L represents the shortest distance between the reference contour area in the corresponding random enclosure scheme and the obstacle position points in the three-dimensional environment model around the target vehicle; r represents the preset calibration matching conversion factor;

[0021] During the process of calibrating the parking marking position information in the three-dimensional environment model around the target vehicle according to each parking marking contour model preset in the database, calculate the contour model matching value, denoted as P, of the parking marking position information in the three-dimensional environment model around the target vehicle based on the parking marking contour model, and use the area covered by each preset parking marking contour model in the three-dimensional environment model around the target vehicle when the value of P is the largest among the corresponding Ps as the calibrated parking marking position information;

[0022] The P is the ratio of the number of position points in the parking line position information in the three-dimensional environment model around the target vehicle that belong to the corresponding preset parking line contour model to the total number of position points in the corresponding parking line position information when the parking line contour model corresponding to the parking line position information in the three-dimensional environment model around the target vehicle is the corresponding preset parking line contour model.

[0023] In the present invention, when querying the minimum parking space area preset for the target vehicle in the database, it is considered that different vehicle models of the target vehicle will have different body parameters, which will in turn lead to differences in the corresponding minimum parking space areas when the corresponding vehicles are parked.

[0024] Furthermore, the obtained parking calibration target area of the target vehicle is based on the relative position information of the current vehicle position area as a set composed of the corresponding current vehicle position area in the three-dimensional environment model and the parking calibration target area of the target vehicle.

[0025] The parking demand condition is the relative obstacle area within the parking demand reference area corresponding to when the parking environment risk of the target vehicle for the locked parking calibration target area of the target vehicle is 0.

[0026] When obtaining the parking environment risk of the target vehicle based on the locked parking calibration target area of the target vehicle, extract the union area of the area in the three-dimensional environment model around the target vehicle whose distance from the locked parking calibration area of the target vehicle is less than or equal to the first associated distance and the locked parking calibration area of the target vehicle, and denote it as the parking demand reference area.

[0027] The first associated distance is the maximum value of the distances between the front and rear ends of the vehicle and the wheels of the vehicle closest in distance.

[0028] Denote the position point with the lowest horizontal height in the parking demand reference area as the obstacle reference point, and extract the area composed of all position points in the parking demand reference area whose difference in corresponding horizontal height from the horizontal height of the obstacle reference point is greater than the minimum height difference between the target vehicle and the ground, and denote it as the relative obstacle area.

[0029] If the relative obstacle area in the parking demand reference area is an empty set, then determine that the parking environment risk of the target vehicle for the locked parking calibration target area of the target vehicle is 0; otherwise, determine that the parking environment risk of the target vehicle for the locked parking calibration target area of the target vehicle is 1.

[0030] When the parking environment risk of the target vehicle for the locked parking calibration target area of the target vehicle is 1, mark the random enclosure scheme corresponding to the parking calibration target area of the target vehicle, jump back to the parking area dynamic locking module, and re-lock the parking calibration target area of the target vehicle after removing the marked random scheme.

[0031] In the process of locking the parking calibration target area of the target vehicle in the present invention, considering the vehicle body parameters of the target vehicle and the obstacle information in the surrounding environment of the target vehicle compared with the target vehicle, the effective planning of the parking area of the target vehicle is realized, providing data support for planning the automatic parking path of the target vehicle in the subsequent steps.

[0032] Further, the parking planning path feedback information includes the feedback time, the parking path plan corresponding to the first parking path planning method, and the parking path plan corresponding to the second parking path planning method;

[0033] The parking planning path feedback information adopts a real-time feedback method, and the content of the parking planning path feedback information corresponding to different feedback times is different;

[0034] There is a case where the parking path plan corresponding to the first parking path planning method and the parking path plan corresponding to the second parking path planning method of the parking planning path feedback information are empty sets.

[0035] Further, when calculating the adaptation value between the parking demand condition and each historical parking record,

[0036] Obtain the historical parking records of the target vehicle, extract the vehicle parking routes corresponding to each parking record in the historical data, and denote the adaptation value between the parking demand condition and the k-th parking record as SPk.

[0037] SPk = PCk·YCk·YZk,

[0038] where PCk represents the parking area deviation value between the parking area after the target vehicle executes according to the vehicle parking route corresponding to the k-th parking record and the locked parking calibration target area of the target vehicle; the value of PCk is equal to the quotient of the intersection area between the parking area after the target vehicle executes according to the vehicle parking route corresponding to the k-th parking record and the locked parking calibration target area of the target vehicle divided by the union area between the parking area after the target vehicle executes according to the vehicle parking route corresponding to the k-th parking record and the locked parking calibration target area of the target vehicle;

[0039] YCk represents the influence value of the relative obstacle area in the area passed by the target vehicle after executing according to the vehicle parking route corresponding to the k-th parking record based on the parking demand reference area;

[0040] If the intersection of the area passed by the target vehicle after executing according to the vehicle parking route corresponding to the k-th parking record and the relative obstacle area in the parking demand reference area is an empty set, then YCk = 0; otherwise, YCk = 1;

[0041] YZk represents the influence value of the obstacle position points in the obstacle position information in the three-dimensional environment model around the target vehicle on the area passed by the target vehicle after executing the vehicle parking route corresponding to the k-th parking record;

[0042] When the intersection of the set of obstacle position points in the obstacle position information in the three-dimensional environment model around the target vehicle and the area passed by the target vehicle after executing the vehicle parking route corresponding to the k-th parking record is an empty set, then YZk = 0; otherwise, YZk = 1;

[0043] Each vehicle parking route corresponding to a parking record whose adaptation value to the parking demand condition is greater than the preset adaptation value is recorded as a historical parking path planning alternative;

[0044] The historical parking path planning scheme is the historical parking path planning alternative with the largest adaptation value to the parking demand condition among all the obtained historical parking path planning alternatives;

[0045] If there is no historical parking path planning alternative, then the historical parking path planning scheme is empty.

[0046] Further, when obtaining the best parking planning path of the target vehicle,

[0047] Obtain the parking demand condition and the obstacle position information in the three-dimensional environment model around the target vehicle, simulate the parking planning path of the target vehicle from the initial state based on the locked target vehicle parking calibration target area to the current vehicle position state, and use the obtained simulated parking planning path as the best parking planning path of the target vehicle;

[0048] The initial state means that the center point of the target vehicle is aligned with the center point of the locked target vehicle parking calibration target area, and the central axis from the front of the target vehicle to the parking space is parallel to the long marking line in the target vehicle parking calibration target area;

[0049] During the process of simulating the parking planning path of the target vehicle from the initial state based on the locked target vehicle parking calibration target area to the current vehicle position state, when the intersection of the area of each forward driving route simulated by the target vehicle and the parking demand condition or the obstacle position information in the three-dimensional environment model around the target vehicle is not an empty set, trigger a reverse operation in advance when the distance between the target vehicle and the obstacle is the preset obstacle distance; when the intersection of the area of each reverse driving route simulated by the target vehicle and the parking demand condition or the obstacle position information in the three-dimensional environment model around the target vehicle is not an empty set, trigger a forward operation in advance when the distance between the target vehicle and the obstacle is the preset obstacle distance; generate the best reverse path planning solution of the target vehicle position based on the obstacle position information through an iterative method;

[0050] When the target vehicle performs a forward operation or a reverse operation during the simulation process, the deflection angle of the target vehicle is obtained by matching the forward deflection angle and the reverse deflection angle corresponding to the relative position between the simulated vehicle position in the database and the actual position of the current vehicle.

[0051] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: The present invention not only takes into account the environmental information around the target vehicle to accurately lock the parking area, but also uses a multi-process parallel analysis method, and at the same time adopts multiple planning methods to plan the parking path based on different data sources, realizes real-time feedback on the automatically planned parking path, and improves the response speed and response result accuracy of the automatically planned parking path;

[0052] In the present invention, as the response time increases, the response result accuracy of the automatically planned parking path will also continuously improve. And during the response process, the driver of the target vehicle can independently select the automatically planned parking path according to his own needs, realizing effective management of the automatically planned parking path data. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0054] Figure 1 is a schematic structural diagram of an automatic parking path planning system based on image recognition according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] Please refer to Figure 1 , the present invention provides a technical solution: an automatic parking path planning system based on image recognition, the system includes:

[0057] A parking environment model building module, which builds a three-dimensional environment model around the target vehicle by collecting the environmental information around the vehicle in real time through an in-vehicle radar and a high-definition camera built in the target vehicle;

[0058] The environmental information includes images and videos collected by the in-vehicle radar and the high-definition camera built in the target vehicle within a preset radius area around the target vehicle;

[0059] The three-dimensional environment model around the target vehicle is constructed based on the environmental information collected within a preset radius area around the target vehicle, and through image recognition technology and 3D generation technology for 2D images.

[0060] In this embodiment, the 3D generation technology for the collected 2D images includes SV3D (StableVideo3D) released by Stability AI or the Make-It-3D method (which creates high-fidelity 3D objects from a single image by using a 2D diffusion model as a 3D-aware prior. This framework does not require multi-view images for training and can be applied to any input image).

[0061] Parking area dynamic locking module, which identifies the position information of obstacles and the position information of parking markings in the three-dimensional environment model around the target vehicle, and locks the parking calibration target area of the target vehicle;

[0062] The position information of the obstacles includes the vector formed by each position point in the three-dimensional environment model around the target vehicle where the difference in height corresponding to the reference plane is greater than the preset height, based on the center point of the target vehicle; the preset height is the minimum height difference between the target vehicle and the ground; the reference plane is a plane parallel to the plane where the bottom crossbeam of the target vehicle is located and with a corresponding height lower than the plane where the bottom crossbeam of the target vehicle is located by a preset height;

[0063] The position information of the parking markings includes each parking marking area identified by image recognition technology in the three-dimensional environment model around the target vehicle and the vector formed by each position point in each parking marking area based on the center point of the target vehicle;

[0064] When locking the parking calibration target area of the target vehicle, obtain the minimum parking space area corresponding to the target vehicle preset in the database, denoted as the reference contour area; calibrate the position information of the parking markings in the three-dimensional environment model around the target vehicle according to each parking space marking contour model preset in the database to obtain the parking marking position calibration information;

[0065] Randomly enclose the reference contour area at different positions in the three-dimensional environment model around the target vehicle, and the intersection of each random enclosure scheme for the reference contour area and the corresponding obstacle position points in the obstacle position information is empty; obtain the calibration matching value corresponding to each random enclosure scheme for the reference contour area, denoted as BD; the locking result of the parking calibration target area of the target vehicle is the reference contour area in the random enclosure scheme with the largest calibration matching value;

[0066] The BD = MU / MZ + r·L,

[0067] Among them, MU represents the area of the intersection region between the reference contour region in the corresponding randomly defined scheme and the region corresponding to the parking line position calibration information; MZ represents the area of the region corresponding to the parking line position calibration information; L represents the shortest distance between the reference contour region in the corresponding randomly defined scheme and the obstacle position points in the three-dimensional environment model around the target vehicle; r represents a preset calibration matching conversion factor;

[0068] During the process of calibrating the parking line position information in the three-dimensional environment model around the target vehicle according to each preset parking line contour model in the database, calculate the contour model matching value of the parking line position information in the three-dimensional environment model around the target vehicle based on the parking line contour model, denoted as P. The region covered by each preset parking line contour model in the three-dimensional environment model around the target vehicle when the value in the corresponding each P is the largest is used as the calibrated parking line position information;

[0069] The P is equal to the ratio of the number of position points belonging to the preset corresponding parking line contour model in the parking line position information in the three-dimensional environment model around the target vehicle to the total number of position points in the corresponding parking line position information when assuming that the parking line contour model corresponding to the parking line position information in the three-dimensional environment model around the target vehicle is the preset corresponding parking line contour model.

[0070] Querying the minimum parking space area corresponding to the target vehicle preset in the database is to consider that different vehicle models of the target vehicle will have different body parameters, which will in turn lead to differences in the corresponding minimum parking space areas when the corresponding vehicles are parked.

[0071] Parking demand condition extraction module, the parking demand condition extraction module combines the current vehicle position area corresponding to the target vehicle in the constructed three-dimensional environment model to obtain the relative position information of the parking calibration target area of the target vehicle based on the current vehicle position area, and generates a parking demand condition;

[0072] The relative position information of the parking calibration target area of the target vehicle based on the current vehicle position area obtained is the set composed of the current vehicle position area corresponding to the three-dimensional environment model and the parking calibration target area of the target vehicle;

[0073] The parking demand condition is the relative obstacle area within the parking demand reference area when the parking environment risk of the target vehicle based on the locked parking calibration target area of the target vehicle is 0;

[0074] When obtaining the target vehicle parking environment risk of the target vehicle parking calibration target area based on locking, extract the union area of the area in the three-dimensional environment model around the target vehicle where the distance to the locked target vehicle parking calibration area is less than or equal to the first associated distance and the locked target vehicle parking calibration area, and denote it as the parking demand reference area;

[0075] The first associated distance is the maximum value of the distances between the front and rear ends of the vehicle and the wheels of the vehicle closest in distance;

[0076] Denote the position point with the lowest horizontal height in the parking demand reference area as the obstacle reference point, and extract the area composed of all position points in the parking demand reference area where the difference between the corresponding horizontal height and the horizontal height of the obstacle reference point is greater than the minimum height difference between the target vehicle and the ground, and denote it as the relative obstacle area;

[0077] If the relative obstacle area in the parking demand reference area is an empty set, determine that the target vehicle parking environment risk of the locked target vehicle parking calibration target area is 0; otherwise, determine that the target vehicle parking environment risk of the locked target vehicle parking calibration target area is 1;

[0078] When the target vehicle parking environment risk of the locked target vehicle parking calibration target area is 1, mark the corresponding random enclosure scheme for the target vehicle parking calibration target area, jump back to the parking area dynamic locking module, and after removing the marked random scheme, re-lock the target vehicle parking calibration target area.

[0079] Parking path intelligent planning module, the parking path intelligent planning module uses a multi-process method to parallelly analyze the parking path schemes corresponding to the first parking path planning method and the second parking path planning method, generate parking planning path feedback information and feedback it to the target vehicle display end. After the driver confirms, automatically park the target vehicle according to the parking planning path confirmed by the driver;

[0080] The parking planning path feedback information includes the feedback time, the parking path scheme corresponding to the first parking path planning method, and the parking path scheme corresponding to the second parking path planning method;

[0081] The parking planning path feedback information adopts a real-time feedback method, and the content of the parking planning path feedback information corresponding to different feedback times is different;

[0082] There may be a situation where the parking path scheme corresponding to the first parking path planning method and the parking path scheme corresponding to the second parking path planning method in the parking planning path feedback information are empty sets.

[0083] The first parking path planning method obtains the historical parking records of the target vehicle, calculates the adaptation values between the parking demand conditions and each historical parking record, and obtains the historical parking path planning scheme according to the adaptation values;

[0084] When calculating the adaptation values between the parking demand conditions and each historical parking record,

[0085] obtain the historical parking records of the target vehicle, extract the vehicle parking routes corresponding to each parking record in the historical data, and denote the adaptation value between the parking demand conditions and the k-th parking record as SPk,

[0086] SPk = PCk · YCk · YZk,

[0087] where PCk represents the parking area deviation value between the parking area after the target vehicle executes according to the vehicle parking route corresponding to the k-th parking record and the locked target vehicle parking calibration target area; the value of PCk is equal to the quotient of the intersection area between the parking area after the target vehicle executes according to the vehicle parking route corresponding to the k-th parking record and the locked target vehicle parking calibration target area divided by the union area between the parking area after the target vehicle executes according to the vehicle parking route corresponding to the k-th parking record and the locked target vehicle parking calibration target area;

[0088] YCk represents the influence value of the relative obstacle area in the area passed by the target vehicle after executing according to the vehicle parking route corresponding to the k-th parking record based on the parking demand reference area;

[0089] If the intersection of the area passed by the target vehicle after executing according to the vehicle parking route corresponding to the k-th parking record and the relative obstacle area in the parking demand reference area is an empty set, then YCk = 0; otherwise, YCk = 1;

[0090] YZk represents the influence value of the obstacle position points in the obstacle position information in the three-dimensional environment model around the target vehicle based on the area passed by the target vehicle after executing according to the vehicle parking route corresponding to the k-th parking record;

[0091] If the intersection of the set of obstacle position points in the obstacle position information in the three-dimensional environment model around the target vehicle and the area passed by the target vehicle after executing according to the vehicle parking route corresponding to the k-th parking record is an empty set, then YZk = 0; otherwise, YZk = 1;

[0092] The vehicle parking routes corresponding to each parking record whose adaptation value to the parking demand conditions is greater than the preset adaptation value are recorded as a historical parking path planning alternative;

[0093] The historical parking path planning solution is the historical parking path planning alternative with the largest adaptation value to the parking demand conditions among all the obtained historical parking path planning alternatives;

[0094] If there is no historical parking path planning alternative, the historical parking path planning solution is empty.

[0095] The second parking path planning method generates the best reverse path planning solution for the target vehicle position based on the obstacle position information by obtaining the parking demand conditions and the obstacle position information in the three-dimensional environment model around the target vehicle through an iterative method, and obtains the best parking planning path for the target vehicle;

[0096] When obtaining the best parking planning path for the target vehicle,

[0097] Obtain the parking demand conditions and the obstacle position information in the three-dimensional environment model around the target vehicle, simulate the parking planning path when the target vehicle moves from the initial state based on the locked target vehicle parking calibration target area to the current vehicle position state, and use the obtained simulated parking planning path as the best parking planning path for the target vehicle;

[0098] The initial state means that the center point of the target vehicle is aligned with the center point of the locked target vehicle parking calibration target area, and the central axis from the front of the target vehicle to the direction of the parking space is parallel to the long marking line in the target vehicle parking calibration target area;

[0099] During the process of simulating the parking planning path when the target vehicle moves from the initial state based on the locked target vehicle parking calibration target area to the current vehicle position state, each time the intersection between the forward route area simulated by the target vehicle and the parking demand conditions or the obstacle position information in the three-dimensional environment model around the target vehicle is not an empty set, a reverse operation is triggered in advance when the distance between the target vehicle and the obstacle is the preset obstacle distance; each time the intersection between the reverse route area simulated by the target vehicle and the parking demand conditions or the obstacle position information in the three-dimensional environment model around the target vehicle is not an empty set, a forward operation is triggered in advance when the distance between the target vehicle and the obstacle is the preset obstacle distance; the best reverse path planning solution for the target vehicle position based on the obstacle position information is generated through an iterative method;

[0100] When the target vehicle performs a forward operation or a reverse operation during the simulation process, the deflection angle of the target vehicle is obtained by matching the forward deflection angle and the reverse deflection angle corresponding to the relative position between the simulated vehicle position in the database and the actual position of the current vehicle.

[0101] It should be noted that in this text, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0102] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An automatic parking path planning system based on image recognition, characterized in that: The system includes: a parking environment model building module, a parking area dynamic locking module, a parking requirement condition extraction module and a parking path intelligent planning module. The parking environment model building module collects environmental information around the target vehicle in real time through the built-in vehicle radar and high-definition camera on the target vehicle, and builds a three-dimensional environmental model around the target vehicle; The parking area dynamic locking module identifies obstacle position information and parking line position information in a three-dimensional environment model around the target vehicle, and locks the parking demarcation target area of ​​the target vehicle; When the parking calibration target area of ​​the target vehicle is locked, the minimum parking space area corresponding to the target vehicle preset in the database is obtained and recorded as the reference contour area; the parking marking position information in the three-dimensional environment model around the target vehicle is calibrated according to the contour models of each parking space marking preset in the database to obtain the parking marking position calibration information; Randomly delineate reference contour areas at different positions in the three-dimensional environment model around the target vehicle, and the intersection of each random delineation scheme for the reference contour area and the corresponding obstacle position point in the obstacle position information is empty; obtain the calibration matching value corresponding to each random delineation scheme for the reference contour area, recorded as BD; the locking result of the parking calibration target area of ​​the target vehicle is the reference contour area in the random delineation scheme with the largest calibration matching value; The parking requirement condition extraction module obtains the relative position information of the parking calibration target area of ​​the target vehicle based on the current vehicle position area in combination with the current vehicle position area corresponding to the target vehicle in the constructed three-dimensional environment model, and generates the parking requirement condition; the parking requirement condition is the relative obstacle area in the parking requirement reference area corresponding to the target vehicle parking environment risk of 0 based on the locked parking calibration target area of ​​the target vehicle; The parking path intelligent planning module adopts a multi-process method to analyze the parking path solutions corresponding to the first parking path planning method and the second parking path planning method in parallel, generates parking plan path feedback information and feeds it back to the target vehicle display terminal, and after the driver confirms, automatically parks the target vehicle according to the parking plan path confirmed by the driver; The first parking path planning method obtains the historical parking records of the target vehicle, calculates the adaptation value between the parking requirement condition and each historical parking record, and obtains the historical parking path planning scheme according to the adaptation value; The second parking path planning method obtains parking requirements and obstacle position information in a three-dimensional environment model around the target vehicle, and iteratively generates an optimal reverse path planning solution for the target vehicle position based on the obstacle position information to obtain the optimal parking planning path for the target vehicle.

2. The automatic parking path planning system based on image recognition according to claim 1, characterized in that: The environmental information includes images and videos collected by the built-in vehicle radar and high-definition camera on the target vehicle within a preset radius around the target vehicle; The three-dimensional environment model around the target vehicle is constructed based on the collected environmental information within a preset radius area around the target vehicle through image recognition technology and 3D generation technology of 2D images.

3. The automatic parking path planning system based on image recognition according to claim 1, characterized in that: The obstacle position information includes a vector formed based on the center point of the target vehicle for each position point in the three-dimensional environment model around the target vehicle whose corresponding height difference with the reference plane is greater than a preset height; the preset height is the minimum height difference between the target vehicle and the ground; the reference plane is a plane parallel to the plane where the bottom crossbeam of the target vehicle is located and whose corresponding height is lower than the preset height of the plane where the bottom crossbeam of the target vehicle is located; The position information of the stop line includes each stop line area identified by image recognition technology in the three-dimensional environment model around the target vehicle and a vector formed by each position point in each stop line area based on the center point of the target vehicle; The BD=MU / MZ+r·L, Wherein, MU represents the intersection area between the reference contour area in the corresponding random delineation scheme and the area corresponding to the parking mark position calibration information; MZ represents the area corresponding to the parking mark position calibration information; L represents the shortest distance between the reference contour area in the corresponding random delineation scheme and the obstacle position point in the three-dimensional environment model around the target vehicle; r represents the preset calibration matching conversion factor; In the process of calibrating the parking line position information in the three-dimensional environment model around the target vehicle according to each parking line contour model preset in the database, the contour model matching value of the parking line position information in the three-dimensional environment model around the target vehicle based on the parking line contour model is calculated, which is recorded as P, and the area covered by each preset parking line contour model in the three-dimensional environment model around the target vehicle when the corresponding value in each P is the largest is used as the calibrated parking line position information; The P is equal to the ratio of the number of position points in the parking line position information in the three-dimensional environment model around the target vehicle that belong to the preset corresponding parking line contour model to the total number of position points in the corresponding parking line position information when assuming that the parking line contour model corresponding to the parking line position information in the three-dimensional environment model around the target vehicle is the preset corresponding parking line contour model.

4. The automatic parking path planning system based on image recognition according to claim 3, characterized in that: The parking calibration target area of ​​the target vehicle is obtained based on the relative position information of the current vehicle position area, which is a set of the current vehicle position area corresponding to the three-dimensional environment model and the parking calibration target area of ​​the target vehicle; When obtaining the target vehicle parking environment risk based on the locked target vehicle parking calibration target area, extract the union area of ​​the area whose distance from the locked target vehicle parking calibration area is less than or equal to the first associated distance and the locked target vehicle parking calibration area in the three-dimensional environment model around the target vehicle, and record it as the parking demand reference area; The first associated distance is the maximum value of the distances between the front and rear ends of the vehicle and the nearest vehicle wheel respectively; The position point with the lowest horizontal height in the parking demand reference area is recorded as the obstacle reference point, and the area consisting of all the position points in the parking demand reference area whose difference between the corresponding horizontal height and the horizontal height of the obstacle reference point is greater than the minimum height difference between the target vehicle and the ground is extracted and recorded as the relative obstacle area; If the relative obstacle area in the parking demand reference area is an empty set, the target vehicle parking environment risk in the locked target vehicle parking calibration target area is determined to be 0; Otherwise, it is determined that the target vehicle parking environment risk of the locked target vehicle parking calibration target area is 1; When the target vehicle parking environment risk of the locked target vehicle parking calibration target area is 1, the random demarcation scheme corresponding to the target vehicle parking calibration target area is marked, and the dynamic locking module of the parking area is jumped again. After the marked random scheme is eliminated, the parking calibration target area of ​​the target vehicle is locked again.

5. The automatic parking path planning system based on image recognition according to claim 4, characterized in that: The parking planning path feedback information includes feedback time, a parking path plan corresponding to the first parking path planning method, and a parking path plan corresponding to the second parking path planning method; The parking planning path feedback information adopts a real-time feedback mode, and the parking planning path feedback information corresponding to different feedback times has different contents; There is a case where the parking path solution corresponding to the first parking path planning method and the parking path solution corresponding to the second parking path planning method in the parking planning path feedback information have an empty set.

6. The automatic parking path planning system based on image recognition according to claim 1, characterized in that: When calculating the adaptation value of the parking demand condition and each historical parking record, Obtain the historical parking records of the target vehicle, extract the vehicle parking route corresponding to each parking record in the historical data, and record the adaptation value between the parking requirement condition and the kth parking record as SPk. SPk=PCk·YCk·YZk, Wherein, PCk represents the parking area deviation value between the parking area after the target vehicle executes the vehicle parking route corresponding to the kth parking record and the locked target vehicle parking calibration target area; the value of PCk is equal to the quotient of the intersection area of ​​the parking area after the target vehicle executes the vehicle parking route corresponding to the kth parking record and the locked target vehicle parking calibration target area divided by the quotient of the union area of ​​the parking area after the target vehicle executes the vehicle parking route corresponding to the kth parking record and the locked target vehicle parking calibration target area; YCk represents the impact value of the relative obstacle area in the parking demand reference area of ​​the target vehicle after executing the vehicle parking route corresponding to the kth parking record; If the intersection of the area passed by the target vehicle after executing the vehicle parking route corresponding to the kth parking record and the relative obstacle area in the parking demand reference area is an empty set, then YCk=0; otherwise, YCk=1; YZk represents the influence value of the obstacle position point in the obstacle position information in the three-dimensional environment model around the target vehicle after the target vehicle executes the vehicle parking route corresponding to the kth parking record; If the intersection of the set of obstacle position points in the obstacle position information in the three-dimensional environment model around the target vehicle and the area passed by the target vehicle after executing the vehicle parking route corresponding to the kth parking record is an empty set, then YZk=0; otherwise, YZk=1; When the adaptation value to the parking requirement condition is greater than the preset adaptation value, the vehicle parking route corresponding to each parking record is recorded as a historical parking route planning alternative; The historical parking path planning solution is the historical parking path planning solution with the largest adaptation value to the parking requirement condition among all the historical parking path planning solution obtained; If there is no historical parking path planning alternative, the historical parking path planning solution is empty.

7. The automatic parking path planning system based on image recognition according to claim 1, characterized in that: When the optimal parking planning path of the target vehicle is obtained, Obtaining parking demand conditions and obstacle position information in a three-dimensional environment model around the target vehicle, simulating a parking planning path of the target vehicle from an initial state based on a locked parking calibration target area to a current vehicle position state, and using the obtained simulated parking planning path as an optimal parking planning path for the target vehicle; The initial state indicates that the center point of the target vehicle is aligned with the center point of the parking calibration target area of ​​the locked target vehicle, and the central axis of the target vehicle from the front to the parking space is parallel to the long marking line in the parking calibration target area of ​​the target vehicle; In the process of simulating the parking path planning of the target vehicle from the initial state based on the locked target vehicle parking calibration target area to the current vehicle position state, each time the intersection between the forward route area simulated by the target vehicle and the parking requirement condition or the obstacle position information in the three-dimensional environment model around the target vehicle is not an empty set, the reverse operation is triggered in advance when the distance between the target vehicle and the obstacle is a preset obstacle distance; each time the intersection between the reverse route area simulated by the target vehicle and the parking requirement condition or the obstacle position information in the three-dimensional environment model around the target vehicle is not an empty set, the forward operation is triggered in advance when the distance between the target vehicle and the obstacle is a preset obstacle distance; and the optimal reverse path planning solution of the target vehicle position based on the obstacle position information is generated in an iterative manner; When the target vehicle performs a forward operation or a reverse operation during the simulation, the deflection angle of the target vehicle is obtained by matching the forward deflection angle and the reverse deflection angle corresponding to the relative position between the simulated vehicle position in the database and the actual position of the current vehicle.

Images