Mechanical storage location detection method and system and storage medium

Abstract

The invention provides a mechanical storage location detection method and system and a storage medium. The method comprises the following steps: acquiring a view around a vehicle; in the cruising stage, a parking space is searched, a view around a vehicle is input into a parking space detection model, the parking space detection model is used for detecting coordinates of four corner points of the parking space, and the corner points comprise two entrance corner points; on the basis of a storage location detection result output by the storage location detection model, when the storage location is located in the optimal observation area of the entrance corner points, the coordinates of the two entrance corner points are optimized; inputting the view around the vehicle into a travelable area model, wherein the travelable area model is used for detecting whether an obstacle exists in the parking space or not; and based on the optimized coordinates of the two entrance corner points and a travelable area detection result output by the travelable area model, whether the parking space can be normally parked or not is judged. According to the invention, the accuracy of parking space corner point coordinate and parking space width detection can be improved.

Description

Technical Field

[0001] The present application relates to the field of vehicle parking technology, and in particular to a mechanical parking location detection method and a system and storage medium thereof. Background Art

[0002] Mechanical parking spaces usually refer to parking spaces in mechanical parking garages. They are non-fixed parking spaces that use mechanical equipment to park and retrieve vehicles. Mechanical parking spaces can significantly improve land utilization and space utilization, and are an effective way to solve the problem of difficult urban parking. Mechanical parking spaces are movable, non-independent spaces that must rely on the overall operation of machinery and the exchange of space to achieve the use of a single parking space. Mechanical parking equipment can flexibly and conveniently convert empty spaces to store and retrieve vehicles through actions such as raising, lowering or lateral movement of the vehicle loading plate. This type of parking space is different from traditional fixed parking spaces, and it can create more parking spaces in a limited space.

[0003] However, since the mechanical storage location is higher than the ground, the image coordinates after perspective transformation will be mapped back to the real-world coordinates inconsistently, which will lead to inaccurate coordinates of the storage location corners and inaccurate detection of the storage location width. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a mechanical storage location detection method and its system, and storage medium, which can improve the accuracy of storage location corner point coordinates and storage location width detection.

[0005] One aspect of an embodiment of the present application provides a mechanical storage location detection method. The method includes: obtaining a view around a vehicle; in the cruising stage, searching for a storage location, inputting the view around the vehicle into a storage location detection model, the storage location detection model is used to detect the coordinates of the four corner points of the storage location, the corner points include two entrance corner points; based on the storage location detection result output by the storage location detection model, when the storage location is within the optimal observation area of ​​the entrance corner point, optimizing the coordinates of the two entrance corner points; inputting the view around the vehicle into an operative area model, the operative area model is used to detect whether there are obstacles in the storage location; based on the optimized coordinates of the two entrance corner points and the operative area detection result output by the operative area model, judging whether the storage location can be normally parked.

[0006] Furthermore, the method also includes: obtaining multiple views around the vehicle in advance; marking the four corner points of each mechanical storage location in the multiple views to obtain a first labeled training data set; constructing an initial storage location detection model based on a deep learning target detection model; and using the first labeled training data set to train the initial storage location detection model to obtain the final trained storage location detection model.

[0007] Furthermore, the method also includes: obtaining multiple views around the vehicle in advance; marking whether there are obstacles in each mechanical storage location in the multiple views to obtain a labeled second training data set; constructing an initial model for movable area detection based on a deep learning image segmentation model; and using the labeled second training data set to train the initial model for movable area detection to obtain the final trained movable area detection model.

[0008] Furthermore, when the following conditions are met at the same time, it is determined that the storage location is in the optimal observation area of ​​the entrance corner point: the angle between the longitudinal direction of the vehicle body and the line connecting the two entrance corner points of the storage location is within a predetermined angle; and the distance between the fisheye lens on the side of the vehicle close to the storage location and the storage location entrance is less than a first predetermined distance.

[0009] Furthermore, the two entrance corner points include a left entrance corner point and a right entrance corner point, and the optimization of the coordinates of the two entrance corner points includes: in the longitudinal direction of the vehicle body, when the fisheye lens on the side of the vehicle close to the storage location is close to the left entrance corner point or the right entrance corner point within the range of a second predetermined distance, calculating the coordinates of the left entrance corner point or the right entrance corner point in the world coordinate system; within the range, at each moment, respectively calculating the average value of the left entrance corner point or the right entrance corner point in the world coordinate system at all moments before the current moment, and respectively using them as the coordinates of the left entrance corner point or the right entrance corner point in the world coordinate system at the current moment; when the vehicle leaves the optimal observation area of ​​the entrance corner points, locking the coordinates of the left entrance corner point and the right entrance corner point, and using them as the coordinates of the two optimized entrance corner points.

[0010] Further, judging whether the storage location can be normally parked based on the coordinates of the two optimized entrance corner points and the movable area detection result output by the movable area model includes: calculating the entrance width of the storage location based on the coordinates of the two optimized entrance corner points; determining that the storage location can be normally parked when the following conditions are met at the same time: the entrance width of the storage location is larger than the vehicle width by a third predetermined distance; there are no obstacles in the area of ​​the four corner points of the storage location.

[0011] Furthermore, the method also includes: calculating the coordinates of the center point of the storage location entrance based on the coordinates of the two optimized entrance corner points; in the parking stage, when the storage location is in the optimal observation area for the vehicle's parking stage, updating the coordinates of the center point of the storage location entrance; and guiding the vehicle to park based on the updated coordinates of the center point of the storage location entrance to prevent the vehicle from parking off-center.

[0012] Furthermore, when the following conditions are met at the same time, it is determined that the storage location is in the optimal observation area of ​​the vehicle's parking stage: the body of the vehicle does not invade the storage location; the two entrance corner points of the storage location are both within the rearview fisheye lens of the vehicle; the angle between the longitudinal direction of the vehicle body and the line connecting the two entrance corner points of the storage location is within a predetermined angle; the rearview fisheye lens is within a predetermined distance range from the entrance center point of the storage location; and the intersection of the central axis of the vehicle and the line connecting the two entrance corner points of the storage location is within a fourth predetermined distance from the entrance center point.

[0013] Further, the updating of the coordinates of the center point of the entrance of the storage location includes: in the optimal observation area of ​​the parking stage, based on the storage location detection model, detecting the coordinates of two entrance corner points of the storage location, the two entrance corner points including a left entrance corner point and a right entrance corner point; in the optimal observation area of ​​the parking stage, at each moment, respectively calculating the average values ​​of the left entrance corner point and the right entrance corner point at all moments before the current moment in the world coordinate system, and using them as the coordinates of the left entrance corner point and the right entrance corner point at the current moment in the world coordinate system respectively; at each moment, calculating the average value of the coordinates of the entrance center point based on the coordinates of the left entrance corner point and the coordinates of the right entrance corner point to obtain the final coordinates of the entrance center point; when the vehicle leaves the optimal observation area of ​​the parking stage, locking the coordinates of the entrance center point, and using them as the updated coordinates of the entrance center point of the storage location.

[0014] Furthermore, the method also includes: determining a storage location stop line based on a storage location stop line detection model; and completing final parking of the vehicle based on the storage location stop line.

[0015] Furthermore, the view around the vehicle includes a bird's-eye view around the vehicle, and a front fisheye lens, a rear fisheye lens, a left fisheye lens and a right fisheye lens are respectively installed on the front, back, left and right sides of the vehicle. The acquisition of the view around the vehicle includes: acquiring four surround fisheye images of the vehicle through the front fisheye lens, the rear fisheye lens, the left fisheye lens and the right fisheye lens; and obtaining the bird's-eye view based on the four surround fisheye images and using a stitching algorithm to stitch together.

[0016] Another aspect of an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the mechanical storage location detection method as described above are implemented.

[0017] Another aspect of the embodiments of the present application provides a mechanical storage location detection system, which includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the mechanical storage location detection method described above.

[0018] The mechanical storage location detection method and its system and storage medium of one or more embodiments of the present application optimize the coordinates of the two entrance corner points when the storage location is in the optimal observation area of ​​the entrance corner point based on the storage location detection results output by the storage location detection model, thereby solving the problem of inaccurate detection of storage location corner point coordinates and storage location width caused by the image coordinates after perspective transformation mapped back to the real-world coordinates being inconsistent due to the mechanical storage location being higher than the ground, and can improve the accuracy of storage location corner point coordinates and storage location width detection.

[0019] In addition, the mechanical storage location detection method and its system and storage medium of one or more embodiments of the present application use an executable area model to determine whether there are obstacles in the storage location, without the need for ultrasonic and depth prediction models. The model is simpler and requires less cost; it does not rely on high-precision maps, saving costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of a mechanical storage location detection method according to an embodiment of the present application.

[0021] Figure 2 A schematic diagram of a storage location detection result output by a storage location detection model according to an embodiment of the present application.

[0022] Figure 3 A schematic diagram of the optimal observation area of ​​the entrance corner point according to an embodiment of the present application.

[0023] Figure 4 This is a flow chart of a mechanical storage location detection method according to an embodiment of the present application.

[0024] Figure 5 A schematic diagram of the optimal observation area during the docking phase of an embodiment of the present application.

[0025] Figure 6 This is a schematic block diagram of a mechanical storage location detection system according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are merely examples of devices consistent with some aspects of the present application as detailed in the appended claims.

[0027] The mechanical storage location detection method and its system and storage medium of the present application are described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the features of the following embodiments and implementations can be combined with each other.

[0028] Figure 1 A flow chart of a mechanical storage location detection method according to an embodiment of the present application is disclosed. Figure 1 As shown, a mechanical storage location detection method according to an embodiment of the present application may include steps S11 to S16.

[0029] In step S11 , a view around the vehicle is acquired.

[0030] The view around the vehicle may include, for example, a bird's-eye view (BEV) around the vehicle.

[0031] In some embodiments, a front fisheye lens, a rear fisheye lens, a left fisheye lens, and a right fisheye lens are installed on the front, rear, left, and right sides of the vehicle, respectively. Therefore, the front fisheye lens, the rear fisheye lens, the left fisheye lens, and the right fisheye lens can be used to obtain the front surround fisheye image, the rear surround fisheye image, the left surround fisheye image, and the right surround fisheye image of the vehicle, respectively. Then, the bird's-eye view can be obtained by stitching these four surround fisheye images using a stitching algorithm.

[0032] In step S12, during the cruising phase, the storage location is searched, and then, steps S13 and S14 are entered respectively.

[0033] In step S13, the view around the vehicle is input into the storage location detection model.

[0034] The storage location detection model can output storage location detection results. Figure 2 A schematic diagram of a storage location detection result output by a storage location detection model according to an embodiment of the present application is disclosed. Figure 2 As shown, the storage location detection model can be used to detect the coordinates of the four corner points P, N, L, and R of the storage location, wherein the four corner points include two entrance corner points P and N and two tail corner points L and R. The storage location detection result includes the coordinates of the four corner points P, N, L, and R of the storage location.

[0035] In some embodiments, the mechanical storage location detection method of the present application also includes: pre-establishing a storage location detection model.

[0036] The following will introduce in detail how to pre-establish a storage location detection model.

[0037] First, multiple views around the vehicle are obtained in advance. For example, 50,000 BEV views can be collected and annotated. Then, each mechanical storage location in multiple BEV views can be annotated as follows: Figure 2 The four corner points P, N, L, and R of the mechanical storage location are shown, so that the first labeled training data set can be obtained.

[0038] The initial model for warehouse location detection can be constructed based on the target detection model of deep learning. In one embodiment, it can be improved based on the yolov5 model so that the model can simultaneously detect the four corner points P, N, L, and R of the warehouse location, thereby obtaining the initial model for warehouse location detection.

[0039] Next, the labeled first training data set may be used to train the constructed initial model for storage location detection, thereby obtaining a finally trained storage location detection model.

[0040] Return to reference Figure 1 In step S14, the view around the vehicle is input into the drivable area model, and the drivable area model can be used to detect whether there are obstacles in the storage location.

[0041] In some embodiments, the mechanical storage location detection method of the present application also includes: pre-establishing an executable area detection model.

[0042] The following will introduce in detail how to pre-establish a storage location detection model.

[0043] First, multiple views around the vehicle are obtained in advance, for example, 50,000 BEV views can be collected and annotated. Then, for each mechanical storage location in the multiple BEV views, it is marked whether there are obstacles in the mechanical storage location, so as to obtain a second annotated training data set;

[0044] An initial model for detecting an executable area can be constructed based on an image segmentation model of deep learning. In one embodiment, an improvement can be made based on a U-Net model so that the model can predict whether there are obstacles in a storage location, thereby obtaining an initial model for detecting an executable area.

[0045] Next, the labeled second training data set may be used to train an initial model for detecting an executable region, thereby obtaining a finally trained executable region detection model.

[0046] In step S15, it is determined whether the storage location is in the best observation area of ​​the entrance corner point. If the result of the determination is "yes", the process proceeds to step S16. Otherwise, it returns to step S15 to continue the determination.

[0047] In step S16, based on the storage location detection result output by the storage location detection model, when the storage location is within the optimal observation area of ​​the entrance corner point, the coordinates of the two entrance corner points P and N may be optimized.

[0048] Figure 3 A schematic diagram of the best observation area of ​​the entrance corner point of an embodiment of the present application is disclosed. Figure 3 As shown, in some embodiments, when the following conditions are met at the same time, it is determined that the storage location is in the best observation area of ​​the entrance corner:

[0049] (1) The angle between the longitudinal direction of the vehicle body and the line connecting the two entrance corner points P and N of the storage location is within the predetermined angle;

[0050] (2) The distance S between the fisheye lens F on the side of the vehicle close to the storage location and the storage location entrance is less than the first predetermined distance.

[0051] For example, when the angle between the longitudinal direction of the vehicle body and the line connecting the two entrance corner points P and N of the storage location is within 20°, and the distance S between the fisheye lens F on the side of the vehicle close to the storage location and the storage location entrance is less than 4m (meters), it is determined that the storage location is in the best observation area of ​​the entrance corner points.

[0052] The two entry corner points P and N include a left entry corner point P and a right entry corner point N. Optimizing the coordinates of the two entry corner points P and N in step S15 may further include steps S151 to S153.

[0053] In step S151, in the longitudinal direction of the vehicle body (for example Figure 3 When the fisheye lens of the vehicle close to the storage location is close to the left entrance corner point P or the right entrance corner point N and is within a second predetermined distance, for example, within a range of 20 cm (centimeter), the coordinates of the left entrance corner point P or the right entrance corner point N in the world coordinate system are calculated.

[0054] In step S152, within the above range, at each moment, the average value of the left entrance corner point P or the right entrance corner point N at all moments before the current moment in the world coordinate system is calculated, and used as the coordinates of the left entrance corner point P or the right entrance corner point N at the current moment in the world coordinate system.

[0055] In step S153, when the vehicle leaves the optimal observation area of ​​the entrance corner point, the coordinates of the left entrance corner point P and the coordinates of the right entrance corner point N are locked, that is, the coordinates of the left entrance corner point P and the coordinates of the right entrance corner point N no longer change, and the locked coordinates of the left entrance corner point P and the coordinates of the right entrance corner point N are used as the coordinates of the two optimized entrance corner points P and N.

[0056] Continue to return Figure 1 In step S17, it can be determined whether the storage location can be normally docked based on the coordinates of the two entrance corner points P and N optimized in step S16 and the feasible area detection result output by the feasible area model in step S14.

[0057] Furthermore, the entrance width of the storage location can be calculated based on the coordinates of the two optimized entrance corner points P and N. The entrance width of the storage location can be obtained by solving the Euclidean distance between the coordinates of the left entrance corner point P and the right entrance corner point N.

[0058] When the following conditions are met at the same time, it is determined that the storage location can be docked normally:

[0059] (1) The entrance width of the storage location is greater than the vehicle width by a third predetermined distance, for example, the entrance width of the storage location is greater than the vehicle width + 0.5 m;

[0060] (2) There are no obstacles in the area of ​​the four corner points of the storage location.

[0061] If it is determined that the storage location can be parked normally, the process proceeds to step S18. In step S18, the vehicle enters the parking stage. If it is determined that the storage location cannot be parked, the process returns to step S12 to continue searching for the storage location.

[0062] The mechanical storage location detection method of the present application optimizes the coordinates of the two entrance corner points when the storage location is in the optimal observation area of ​​the entrance corner point based on the storage location detection results output by the storage location detection model, thereby solving the problem of inaccurate detection of storage location corner point coordinates and storage location width caused by the mechanical storage location being higher than the ground, resulting in inconsistent mapping of image coordinates after perspective transformation back to real-world coordinates.

[0063] In addition, the mechanical storage location detection method of the present application uses an executable area model to determine whether there are obstacles in the storage location, without the need for ultrasonic and depth prediction models. The model is simpler and requires less cost; it does not rely on high-precision maps, saving costs.

[0064] Figure 4 A flow chart of a mechanical storage location detection method according to another embodiment of the present application is disclosed. Figure 4 As shown, in some embodiments, the mechanical storage location detection method of the present application may further include step S21.

[0065] In step S21, the coordinates of the storage location entrance center point C may be further calculated based on the coordinates of the two optimized entrance corner points.

[0066] In step S22, during the parking phase, it is determined whether the storage location is in the optimal observation area for the parking phase of the vehicle. If the result of the determination is "yes", the process proceeds to step S23. Otherwise, the process returns to step S22 to continue the determination.

[0067] In step S23, when the storage location is in the optimal observation area for the parking phase of the vehicle, the coordinates of the storage location entrance center point C are updated.

[0068] Figure 5 A schematic diagram of the best observation area during the docking phase of an embodiment of the present application is disclosed. Figure 5 As shown, in some embodiments, when the following conditions are met at the same time, it is determined that the storage location is in the optimal observation area of ​​the vehicle parking stage:

[0069] (1) The vehicle body does not intrude into the storage space;

[0070] (2) The two entrance corner points P and N of the storage location are both within the rear-view fisheye lens M of the vehicle;

[0071] (3) The angle between the longitudinal direction of the vehicle body (i.e., the x-direction) and the line connecting the two entrance corner points P and N of the storage location is within a predetermined angle. For example, the angle between the longitudinal direction of the vehicle body and the line connecting the two entrance corner points P and N of the storage location is within 15°;

[0072] (4) The distance between the rear-view fisheye lens M and the entrance center point C of the storage location is within a predetermined distance range, for example, within a range of 0.1 m to 1.5 m;

[0073] (5) The intersection point O of the central axis of the vehicle and the line connecting the two entrance corner points P and N of the storage location is within a fourth predetermined distance from the entrance center point C, for example, within 0.5 m.

[0074] In some embodiments, updating the coordinates of the center point of the storage location entrance in step S23 may further include steps S231 to S234.

[0075] In step S231, within the optimal observation area of ​​the docking stage, the coordinates of the two entrance corner points P and N of the storage location are detected based on the storage location detection model, and the two entrance corner points P and N include a left entrance corner point P and a right entrance corner point N.

[0076] In step S232, within the optimal observation area during the parking phase, at each moment, the average values ​​of the left entrance corner point P and the right entrance corner point N in the world coordinate system at all moments before the current moment are calculated and used as the coordinates of the left entrance corner point P and the right entrance corner point N in the world coordinate system at the current moment.

[0077] In step S233, at each moment, the average value of the coordinates of the entrance center point C is calculated based on the coordinates of the left entrance corner point P and the coordinates of the right entrance corner point N to obtain the final coordinates of the entrance center point C.

[0078] In step S234, when the vehicle leaves the optimal observation area in the parking phase, the coordinates of the entrance center point are locked and used as the updated coordinates of the entrance center point of the storage location.

[0079] return Figure 4 In step S24, the vehicle can be guided to park based on the updated coordinates of the center point of the storage location entrance to prevent the vehicle from parking in an offset manner.

[0080] The mechanical storage location detection method of the present application can solve the problem that the mechanical storage location is prone to parking deviation by updating the coordinates of the center point of the storage location entrance when the storage location is in the optimal observation area during the parking stage, and using this to guide the vehicle to park.

[0081] In some embodiments, the mechanical storage location detection method of the present application may further include step S25 and step S26.

[0082] In step S25 , the storage location stop line may be determined based on the storage location stop line detection model.

[0083] In step S26, the final parking of the vehicle is completed based on the parking stop line.

[0084] The mechanical storage location detection method of the present application can accurately detect the storage location stop line based on the storage location stop line detection model, thereby preventing rear-end collisions and effectively solving the problem of easy rear-end collisions during the mechanical storage location parking process.

[0085] The present application also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the mechanical storage location detection method as described above are implemented.

[0086] The present application also provides a mechanical storage location detection system. Figure 6 A schematic block diagram of a mechanical storage location detection system 600 according to an embodiment of the present application is disclosed. Figure 6 As shown, a mechanical storage location detection system 600 of an embodiment of the present application includes a processor 601, an internal bus 602, a network interface 603, a memory 604, and a non-volatile memory 605, and may also include hardware required for other services. The processor 601 can read the corresponding computer program from the non-volatile memory 605 into the memory 604 and then run it to implement the steps of the mechanical storage location detection method as described above. Of course, in addition to the software implementation, the present application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic components.

[0087] The mechanical storage location detection system 600 of the present application can have similar beneficial technical effects as the above mechanical storage location detection method, so it will not be repeated here.

[0088] The mechanical storage location detection method and its system and storage medium provided in the embodiment of the present application are introduced in detail above. Specific examples are used herein to illustrate the mechanical storage location detection method and its system and storage medium in the embodiment of the present application. The description of the above embodiments is only used to help understand the core idea of ​​the present application and is not intended to limit the present application. It should be pointed out that for ordinary technicians in this technical field, without departing from the spirit and principles of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications should also fall within the scope of protection of the claims attached to the present application.

Claims

1. A mechanical storage location detection method, characterized in that: include: Get a view around the vehicle; In the cruising phase, a storage location is searched, and the view around the vehicle is input into a storage location detection model, wherein the storage location detection model is used to detect the coordinates of four corner points of the storage location, wherein the corner points include two entry corner points; Based on the storage location detection results output by the storage location detection model, when the storage location is within the optimal observation area of ​​the entrance corner point, optimizing the coordinates of the two entrance corner points; Inputting the view around the vehicle into a drivable area model, wherein the drivable area model is used to detect whether there are obstacles in the storage location; Based on the coordinates of the two optimized entrance corner points and the movable area detection result output by the movable area model, it is determined whether the storage location can be docked normally.

2. The method according to claim 1, characterized in that The method further comprises: Obtain multiple views of the vehicle's surroundings in advance; For each mechanical storage location in the plurality of views, four corner points of the mechanical storage location are respectively marked to obtain a marked first training data set; Build an initial model for location detection based on a deep learning-based target detection model; The labeled first training data set is used to train the initial model for storage location detection to obtain the finally trained storage location detection model.

3. The method according to claim 1, characterized in that The method further comprises: Obtain multiple views of the vehicle's surroundings in advance; Marking each mechanical storage location in the multiple views respectively whether there is an obstacle in the mechanical storage location to obtain a marked second training data set; Build an initial model for executable region detection based on a deep learning-based image segmentation model; The labeled second training data set is used to train the initial model for detecting the executable area, so as to obtain the finally trained model for detecting the executable area.

4. The method according to claim 1, characterized in that When the following conditions are met at the same time, the storage location is determined to be in the best observation area of ​​the entrance corner: The angle between the longitudinal direction of the vehicle body and the line connecting the two entrance corner points of the storage location is within the predetermined angle; and The distance between the fisheye lens on the side of the vehicle close to the storage location and the storage location entrance is less than the first predetermined distance.

5. The method according to claim 1, characterized in that The two entry corner points include a left entry corner point and a right entry corner point, and the optimization of the coordinates of the two entry corner points includes: In the longitudinal direction of the vehicle body, when the fisheye lens on the side of the vehicle close to the storage location is close to the left entrance corner point or the right entrance corner point within the range of the second predetermined distance, the coordinates of the left entrance corner point or the right entrance corner point in the world coordinate system are calculated; Within the range, at each moment, respectively calculate the average value of the left entrance corner point or the right entrance corner point at all moments before the current moment in the world coordinate system, and use it as the coordinate of the left entrance corner point or the right entrance corner point at the current moment in the world coordinate system; When the vehicle leaves the optimal observation area of ​​the entrance corner point, the coordinates of the left entrance corner point and the coordinates of the right entrance corner point are locked and used as the coordinates of the two optimized entrance corner points.

6. The method according to claim 1, characterized in that The step of judging whether the storage location can be normally docked based on the coordinates of the two optimized entrance corner points and the executable area detection result output by the executable area model includes: Calculate the entrance width of the storage location based on the coordinates of the two optimized entrance corner points; When the following conditions are met at the same time, it is determined that the storage location can be docked normally: The entrance width of the storage location is greater than the vehicle width by a third predetermined distance; There are no obstacles in the area of ​​the four corner points of the storage location.

7. The method according to claim 1, characterized in that Also includes: Calculate the coordinates of the center point of the storage location entrance based on the coordinates of the two optimized entrance corner points; During the parking phase, when the parking space is in the optimal observation area for the vehicle during the parking phase, the coordinates of the center point of the parking space entrance are updated; The updated coordinates of the center point of the entrance to the storage location are used to guide the vehicle to park in order to prevent the vehicle from parking in the wrong direction.

8. The method according to claim 7, characterized in that When the following conditions are met at the same time, it is determined that the storage location is in the optimal observation area of ​​the vehicle during the parking phase: The vehicle body does not intrude into the storage space; The two entrance corners of the storage location are both within the vehicle’s rear-view fisheye lens; The angle between the longitudinal direction of the vehicle body and the line connecting the two entrance corner points of the storage location is within the predetermined angle; The rear-view fisheye lens is within a predetermined distance range from the center point of the entrance of the storage location; and The intersection of the central axis of the vehicle and the line connecting the two entrance corner points of the storage location is within a fourth predetermined distance from the entrance center point.

9. The method according to claim 7, characterized in that The updating of the coordinates of the center point of the storage location entrance includes: In the optimal observation area of ​​the docking stage, the coordinates of two entrance corner points of the storage location are detected based on the storage location detection model, and the two entrance corner points include a left entrance corner point and a right entrance corner point; In the optimal observation area of ​​the parking stage, at each moment, respectively calculate the average values ​​of the left entrance corner point and the right entrance corner point at all moments before the current moment in the world coordinate system, and use them as the coordinates of the left entrance corner point and the right entrance corner point at the current moment in the world coordinate system; At each moment, the average value of the coordinates of the entrance center point is calculated based on the coordinates of the left entrance corner point and the coordinates of the right entrance corner point to obtain the final coordinates of the entrance center point; When the vehicle leaves the optimal observation area during the parking phase, the coordinates of the entrance center point are locked and used as the updated coordinates of the storage location entrance center point.

10. The method according to claim 1, characterized in that Also includes: Determine the storage location stop line based on the storage location stop line detection model; Complete the final parking of the vehicle based on the parking stop line.

11. The method according to claim 1, characterized in that The view around the vehicle includes a bird's-eye view around the vehicle. A front fisheye lens, a rear fisheye lens, a left fisheye lens and a right fisheye lens are respectively installed on the front, back, left and right sides of the vehicle. The view around the vehicle includes: Acquire four surround fisheye images of the vehicle through the front fisheye lens, the rear fisheye lens, the left fisheye lens and the right fisheye lens; The bird's-eye view is obtained by stitching the four surround fisheye images using a stitching algorithm.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the mechanical storage location detection method as described in any one of claims 1 to 11 are implemented.

13. A mechanical storage location detection system, characterized in that: It includes a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the mechanical storage location detection method as described in any one of claims 1 to 11.

Images