Method, apparatus, device and storage medium for determining vehicle access to a parking space

Abstract

The application provides a method, device, equipment and storage medium for determining vehicle access to a parking space, which comprises: acquiring a first image collected by a camera, determining a vehicle detection frame of a target vehicle and a position of a preset center point of the vehicle body in the first image; if it is determined that the vehicle detection frame and the position of the preset center point of the vehicle body meet a parking space occupation condition of a target parking space frame, acquiring vehicle tracking information of the target vehicle in a plurality of second images collected within a preset time before the first image, and if it is determined according to the vehicle tracking information that the target vehicle appears in a warning frame, determining that the target vehicle drives into a parking space corresponding to the target parking space frame. A warning area outside a plurality of parking spaces covered by the camera is defined to assist in judging vehicle access to the parking spaces, and the accuracy of the determination result of the parking space access is improved through the dual judgment of the parking space occupation condition and the passing warning area.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method, apparatus, device, and storage medium for determining the entry and exit of vehicles from parking spaces. Background Technology

[0002] As my country's urbanization process accelerates, the number of vehicles is also increasing daily, leading to a severe parking problem in cities. For city administrators, optimizing parking resources and improving the intelligence level of parking management have become important goals in developing new smart cities. Therefore, automated fee collection management for roadside parking spaces is particularly crucial.

[0003] Currently, automated toll collection strategies for roadside parking spaces include the following methods: handheld terminal devices, geomagnetic sensors, and video monitoring stations. Among these, the handheld terminal method requires continuous manual monitoring with the device, leading to potential omissions and high operational costs due to heavy reliance on human labor and low technological sophistication. This method is clearly unsuitable for the technological goals of modern smart cities. The second method is the geomagnetic sensor solution, which uses sensors within the geomagnetic field to detect obstacles above the field. However, since it cannot determine whether an obstacle is a parked car or another non-motorized vehicle, this method is susceptible to interference and has a high false alarm rate.

[0004] The prerequisite for automated toll collection strategies for roadside parking spaces is the ability to automatically and accurately identify vehicles entering and leaving the parking spaces. Therefore, how to achieve automatic and accurate identification of vehicles entering and leaving parking spaces is a problem that needs to be solved. Summary of the Invention

[0005] This invention provides a method, apparatus, device, and storage medium for determining vehicle entry and exit from parking spaces, thereby improving the accuracy of the determination results.

[0006] In a first aspect, embodiments of the present invention provide a method for determining whether a vehicle is entering or exiting a parking space, the method comprising:

[0007] Acquire the first image captured by the camera, which covers multiple berths;

[0008] The first image is processed for vehicle detection to obtain the vehicle detection box and the position of the preset center point of the vehicle body corresponding to the target vehicle in the first image.

[0009] if it is determined that the position of the vehicle detection frame and the preset center point of the vehicle body meets the parking space occupation condition of the target parking space boundary, obtaining vehicle tracking information of the target vehicle in a plurality of second images collected within a preset time before the first image, the vehicle tracking information including a vehicle identifier of the target vehicle in a respective corresponding positioning position in the plurality of second images; wherein the target parking space boundary is an image area boundary corresponding to any of the parking spaces in the camera shooting picture;

[0010] if it is determined that the vehicle identifier appears in a preset warning boundary according to the vehicle tracking information, it is determined that the target vehicle parks in the parking space corresponding to the target parking space boundary; wherein the warning boundary is an image area boundary corresponding to a warning area required to pass through the plurality of parking spaces in the camera shooting picture, and the warning area is a road area surrounding the plurality of parking spaces.

[0011] In a second aspect, an embodiment of the present application provides a device for determining vehicle parking and leaving, comprising:

[0012] a obtaining module, configured to obtain a first image collected by a camera, wherein the camera covers a plurality of parking spaces;

[0013] a detection module, configured to perform vehicle detection processing on the first image to obtain a position of a vehicle detection frame and a preset center point of a vehicle body corresponding to a target vehicle in the first image;

[0014] a processing module, configured to, if it is determined that the position of the vehicle detection frame and the preset center point of the vehicle body meets a parking space occupation condition of a target parking space boundary, obtain vehicle tracking information of the target vehicle in a plurality of second images collected within a preset time before the first image, the vehicle tracking information including a vehicle identifier of the target vehicle in a respective corresponding positioning position in the plurality of second images; and if it is determined that the vehicle identifier appears in a preset warning boundary according to the vehicle tracking information, it is determined that the target vehicle parks in the parking space corresponding to the target parking space boundary; wherein the warning boundary is an image area boundary corresponding to a warning area required to pass through the plurality of parking spaces in the camera shooting picture, and the warning area is a road area surrounding the plurality of parking spaces, and the target parking space boundary is an image area boundary corresponding to any of the parking spaces in the camera shooting picture.

[0015] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a communication interface; wherein the memory stores executable code, when the executable code is executed by the processor, the processor can at least implement the method for determining vehicle parking and leaving according to the first aspect.

[0016] In a fourth aspect, an embodiment of the present application provides a non-transitory machine readable storage medium, which stores executable code, and when the executable code is executed by a processor of an electronic device, the processor can implement at least the method for determining whether a vehicle enters or exits a parking space according to the first aspect.

[0017] In a fifth aspect, an embodiment of the present application provides a method for determining whether a vehicle enters or exits a roadside parking space, comprising:

[0018] obtaining a first image collected by a roadside camera, wherein the roadside camera comprises a camera arranged on the same side of the roadside parking space, and the roadside camera covers a plurality of roadside parking spaces;

[0019] performing vehicle detection processing on the first image to obtain a vehicle detection frame corresponding to a target vehicle in the first image and a position of a preset center point of the vehicle body;

[0020] if it is determined that the vehicle detection frame and the position of the preset center point of the vehicle body meet a parking space occupancy condition of a target parking space boundary, obtaining vehicle tracking information of the target vehicle in a plurality of second images collected within a preset time before the first image, wherein the vehicle tracking information comprises a respective positioning position of a vehicle identifier of the target vehicle in the plurality of second images; wherein the target parking space boundary is a boundary of an image region corresponding to any of the roadside parking spaces in a shooting picture of the roadside camera;

[0021] if it is determined according to the vehicle tracking information that the vehicle identifier appears in a preset warning boundary, determining that the target vehicle parks in a roadside parking space corresponding to the target parking space boundary; wherein the warning boundary is a boundary of an image region corresponding to a warning region required to pass through when entering or exiting the plurality of roadside parking spaces in the shooting picture of the roadside camera, and the warning region is a road region surrounding the plurality of roadside parking spaces.

[0022] For the case that a vehicle enters a parking space set at a roadside, a camera can be set at the roadside on the same side of the parking space, which is called a roadside camera. One roadside camera can be configured to cover multiple fixed roadside parking spaces. The roadside camera is not rotatable, so the roadside camera has a fixed shooting range, that is, for the multiple roadside parking spaces covered by one roadside camera, the multiple roadside parking spaces always correspond to a fixed image area in the picture taken by the roadside camera. The closed curve surrounded by the boundary of the image area is called a parking space border. For the multiple roadside parking spaces covered by one roadside camera, a warning area is defined on the same side of the roadside parking spaces. The warning area is a road area surrounding the multiple roadside parking spaces, and the vehicle entering or exiting the multiple roadside parking spaces must pass through the warning area. Based on the definition, it can be understood that the warning area also has a fixed image area in the picture taken by the roadside camera. The closed curve surrounded by the boundary of the image area is called a warning border.

[0023] In the embodiment of the present application, for any camera (such as the above-mentioned roadside camera), the camera takes a video picture, and a frame of image can be obtained by sampling the video picture taken by the camera. For each frame of image obtained by sampling, vehicle detection and vehicle tracking processing can be performed. The purpose of vehicle detection is to detect the vehicle detection frame of each vehicle contained in the image and the position of a preset center point on the vehicle, which is the corresponding pixel position in the image. The purpose of vehicle tracking is to determine the same vehicle in different frames of image and determine the vehicle identification of the vehicle. The vehicle identification can be the recognized license plate number or the same tracking number assigned to the same vehicle.

[0024] For any frame image (referred to as a first image) collected by the camera, after vehicle detection is performed on the first image, the vehicle detection frame corresponding to the target vehicle (referring to any vehicle contained therein) in the first image and the position of the preset center point on the vehicle body are obtained, and on the premise that the plurality of parking spaces covered by the camera each correspond to a parking space frame in the image, first, it is determined whether the target vehicle satisfies the parking space occupation condition, i.e., whether the target vehicle currently occupies the parking space corresponding to the target parking space frame is determined in combination with the detected vehicle detection frame of the target vehicle and the position of the preset center point on the vehicle body. If it is determined that the vehicle detection frame and the position of the preset center point on the vehicle body meet the parking space occupation condition of the target parking space frame, time is traced back to obtain the vehicle tracking information of the target vehicle in a plurality of second images collected within a preset time (such as 10 seconds) before the first image. The vehicle tracking information includes the positioning position of the vehicle identifier of the target vehicle in each of the plurality of second images, that is, based on the previous vehicle tracking process, the position of the target vehicle corresponding to the same vehicle identifier in the previous frames of images can be obtained. Further, it is determined whether the target vehicle entering the parking space corresponding to the target parking space frame has appeared in the warning area before entering the parking space according to the obtained vehicle tracking information, i.e., whether the vehicle identifier of the target vehicle appears in the warning frame in the plurality of second images, if the vehicle identifier appears in one or several second images, it is finally determined that the target vehicle drives into the parking space corresponding to the target parking space frame and stops.

[0025] In the above scheme, a warning area is defined outside the continuous several parking spaces covered by the same camera to assist in determining the vehicle entering and leaving the parking space. Only when it is first determined that the vehicle meets the parking space occupation condition and then it is determined that the vehicle meets the condition of passing through the warning area, it is finally determined that the vehicle drives into the parking space and stops, thereby improving the accuracy of the determination result of driving in. Moreover, when determining the parking space occupation, the vehicle detection frame and the specific center point on the vehicle body are combined to determine whether the vehicle occupies a parking space, thereby improving the accuracy of the determination result of the parking space occupation. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0027] Figure 1 A schematic diagram of a parking space parking scene provided by an embodiment of the present application;

[0028] Figure 2A flow chart of a method for determining that a vehicle has entered a parking space according to an embodiment of the present application is provided.

[0029] Figure 3 A schematic diagram of a vehicle detection frame and a vehicle chassis detection frame according to an embodiment of the present application is provided.

[0030] Figure 4 A schematic diagram of a vehicle detection frame and a vehicle chassis center point determination principle according to an embodiment of the present application is provided.

[0031] Figure 5 A flow chart of a method for determining that a vehicle has exited a parking space according to an embodiment of the present application is provided.

[0032] Figure 6 A structural schematic diagram of a device for determining that a vehicle has entered or exited a parking space according to an embodiment of the present application is provided.

[0033] Figure 7 A structural schematic diagram of an electronic device according to the embodiment shown in Figure 6 A structural schematic diagram of an electronic device according to the embodiment shown in DETAILED DESCRIPTION

[0034] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0035] Some embodiments of the present application will be described in detail below with reference to the drawings. The following embodiments and features in the embodiments can be combined with each other without conflict. In addition, the step sequence in each method embodiment is only an example and is not strictly limited.

[0036] The method for determining that a vehicle has entered or exited a parking space according to the embodiments of the present application is suitable for application scenarios of parking management of roadside parking spaces and indoor and outdoor parking lots. The roadside parking space refers to a parking space allowed to be parked on one side of a carriageway close to a sidewalk.

[0037] In the traditional artificial and geomagnetic-based parking fee management scheme, in addition to the large labor cost, a geomagnetic device needs to be independently set up on each parking space, which also has a large equipment overhead, and this scheme also does not have a picture or video evidence chain of parking, which is easy to cause parking fee disputes. Therefore, for roadside parking, a video stake scheme can be used, which mainly installs a camera on the road sidewalk near the roadside parking at the same height as the vehicle. Although this scheme can preserve the parking evidence chain, it has high maintenance costs and is easy to be damaged by manual operation. In addition, in many northern cities, the road dust is relatively high, and within 3 meters of the road, the dust can easily block the camera, so it also needs to be cleaned regularly by manual operation.

[0038] In view of the disadvantages of the above scheme, for roadside parking, an embodiment of the present application proposes a high-position video scheme, which needs to install a high pole with a relatively high height (generally more than 6 meters) on the sidewalk of the road, and then install a camera on the high pole, each camera is responsible for shooting several (generally 3-4) parking spaces, and through artificial intelligence algorithm, the entering and exiting vehicles of the parking spaces are identified to generate corresponding entering and exiting events, such as recording the parking start and end time of the vehicle and the corresponding vehicle identification (such as license plate number), and saving the picture / video evidence chain of the vehicle entering and exiting, so as to realize the automatic charging management of the roadside parking. This scheme can effectively solve the problems of the previous scheme, such as being easily disturbed, no evidence chain, and high labor operation cost. The high pole can be installed by borrowing a pole (such as borrowing a street lamp pole), which reasonably utilizes the existing municipal facilities and reduces the impact on the city landscape. Similarly, for indoor and outdoor parking lots, cameras can also be installed at a higher position.

[0039] For convenience of description, an embodiment of the present application takes the roadside parking scene as an example for description. For the roadside parking demarcated on one side of the road, the camera (referred to as a roadside camera) for shooting the roadside parking can be arranged on the opposite side of the roadside parking or on the same side of the roadside parking. For example, a row of roadside parking is demarcated on the left edge of a road, and the roadside camera can be arranged on the left sidewalk of the road (on the same side of the roadside parking) or on the right sidewalk of the road (on the opposite side of the roadside parking). Of course, in actual application, the roadside cameras arranged on the same side and the opposite side can both exist.

[0040] It should be noted that a roadside camera has a limited shooting range, therefore, a roadside camera is generally configured to cover several roadside parking spaces. In addition, the roadside camera can be configured not to rotate, that is, it has a fixed shooting angle, so that the multiple roadside parking spaces covered by the same roadside camera have a fixed and unchanged image area in all pictures collected by the roadside camera.

[0041] When the roadside camera is arranged on the same side of the roadside parking space, as shown in Figure 1 the occlusion of the vehicles on the front and rear parking spaces is often serious, and the occlusion rate is generally greater than 60%, and sometimes close to 100%. Therefore, in the case of the roadside camera arranged on the same side of the roadside parking space, it is a great challenge to accurately identify the vehicles entering and leaving the parking space. The vehicles entering and leaving the parking space refer to the vehicles that actually enter the parking space for parking and then leave the parking space after parking for a period of time.

[0042] In the embodiment of the present application, in order to accurately identify the vehicles entering and leaving the parking space, a warning area is defined. As shown in Figure 1 on the same side of the same roadside camera covering a plurality of roadside parking spaces, a carriageway area surrounding the plurality of roadside parking spaces in the same direction is set, which is called a warning area. This warning area represents an area through which all vehicles pass from appearing in the shooting picture of the roadside camera to entering or leaving a certain roadside parking space. As shown in Figure 1 The warning area "surrounds" the plurality of roadside parking spaces, which does not mean that the plurality of roadside parking spaces are located inside the warning area (belonging to the inclusive relationship), but from the positional relationship, the warning area is separate from the plurality of roadside parking spaces, that is, there is no inclusive and intersection relationship.

[0043] It can be understood that the warning area does not need to be actually drawn on the real carriageway, and the warning area is only a region defined to assist in accurately identifying the vehicles entering and leaving the parking space. Since the plurality of roadside areas surrounded by the warning area have fixed image areas in the corresponding shooting picture of the roadside camera, and the positional relationship between the warning area and the plurality of roadside areas is also fixed, the warning area also has a fixed image area in the shooting picture of the roadside camera.

[0044] In the embodiment of the present application, the image area boundary corresponding to the roadside parking space in the shooting picture of the corresponding roadside camera is called a parking space frame, and the image area boundary corresponding to the warning area in the shooting picture of the roadside camera is called a warning frame.

[0045] The parking space frame can be marked in the image area based on the real roadside parking boundary line drawn on the carriageway under the condition that no vehicle enters the plurality of roadside parking spaces covered by the roadside camera.

[0046] The warning frame can be obtained by mapping the spatial positional relationship (i.e. the positional and distance relationship of the two in the real road scene) between the pre-defined warning area and the plurality of roadside parking spaces, and the image area corresponding to the plurality of roadside parking spaces in the shooting picture of the roadside camera.

[0047] The method for determining the entry and exit of a vehicle into a roadside parking space provided by the embodiments of the present application can be executed by an electronic device, which can be a server or a terminal device in communication connection with a roadside camera. The server can be a physical server or a virtual server (virtual machine) in the cloud. Of course, the method provided by the embodiments of the present application can also be completed by the roadside camera in cooperation with the electronic device in the actual execution process. For example, the vehicle detection and vehicle tracking of the image can be completed locally in the roadside camera, and the electronic device receives the data related to the generation of a parking record transmitted by the roadside camera to generate and store the parking record.

[0048] The implementation process of the scheme for determining the entry and exit of a vehicle into a roadside parking space provided by the embodiments of the present application is described below.

[0049] Figure 2 A flowchart of a method for determining the entry of a vehicle into a parking space provided by the embodiments of the present application is shown in FIG. 1, which includes the following steps: Figure 2

[0050] 201. A first image collected by a roadside camera is acquired. The roadside camera includes a camera arranged on the same side of a roadside parking space, and the roadside camera covers multiple roadside parking spaces.

[0051] 202. Vehicle detection processing is performed on the first image to obtain a vehicle detection frame corresponding to a target vehicle in the first image and the position of a preset center point of the vehicle body.

[0052] 203. If it is determined that the vehicle detection frame and the position of the preset center point of the vehicle body meet the parking space occupation condition of a target parking space frame, vehicle tracking information of the target vehicle in multiple second images collected within a preset time before the first image is acquired. The vehicle tracking information includes the positioning position of the vehicle identifier of the target vehicle in each of the multiple second images.

[0053] The target parking space frame is the image area boundary corresponding to one of the multiple roadside parking spaces in the shooting picture of the roadside camera.

[0054] 204. If it is determined according to the vehicle tracking information that the vehicle identifier appears in a preset warning frame, it is determined that the target vehicle parks in the roadside parking space corresponding to the target parking space frame.

[0055] As described above, the warning frame is the image area boundary of the warning area required to pass through the multiple roadside parking spaces in the shooting picture of the roadside camera, and the warning area is the road area surrounding the multiple roadside parking spaces.

[0056] ​For the roadside camera X covering the plurality of roadside parking spaces, the roadside camera X can continuously shoot video pictures. In the face of the demand for accurate identification of vehicles entering and exiting the roadside parking spaces, the video pictures shot by the roadside camera X can be sampled to obtain a frame of image, and for each frame of image, vehicle detection and vehicle tracking processing can be performed. The sampling frequency can be pre-set, such as 5 frames per second.

[0057] The purpose of vehicle detection is to detect the vehicle detection frame of each vehicle contained in the image and the position of a preset center point on the vehicle body, which refers to the corresponding pixel position in the image. The purpose of vehicle tracking is to determine the same vehicle in different frames of images and determine the vehicle identification. The above-mentioned preset center point can be the vehicle chassis center point.

[0058] For vehicle detection processing, the above-mentioned first image can be any frame of image collected by the roadside camera X. Taking the first image as an example, assuming that the above-mentioned vehicle body preset center point is the vehicle chassis center point, the vehicle detection processing of the first image can be implemented as follows:

[0059] The first image is processed using a pre-trained deep neural network model to obtain the vehicle detection frame and the vehicle chassis detection frame corresponding to the target vehicle in the first image, and the position of the center point of the vehicle chassis detection frame is determined as the position of the vehicle chassis center point. The target vehicle refers to any vehicle detected from the first image.

[0060] In the embodiment of the application, the vehicle detection frame refers to a detection frame containing a complete vehicle, and the vehicle chassis detection frame refers to a detection frame containing a vehicle chassis area, as shown in Figure 3

[0061] The vehicle detection frame and the vehicle chassis detection frame are both rectangular frames, and when the vehicle chassis detection frame is obtained, the pixel coordinates of its four vertices in the image can be obtained, and then the pixel coordinates of the center point can be obtained based on the pixel coordinates of the four vertices, that is, the position of the vehicle chassis center point.

[0062] In fact, vehicle detection is also a target detection task for images of a specific target, i.e., vehicles. Therefore, the above-mentioned deep neural network model can be trained using the training idea of the target detection task. In short, a large number of training samples can be pre-collected, which can include training sample images of vehicles in various postures and occlusion situations in images, and the training of the deep neural network model can be performed based on the labeled supervision information. The labeled supervision information is the vehicle detection frame and the vehicle chassis detection frame corresponding to the vehicle in the corresponding training sample image.

[0063] ​In fact, in the case where the roadside camera is arranged on the same side of the roadside parking space, there may be a serious occlusion between the vehicles parked in the roadside parking space from the perspective of the roadside camera, as shown in Figure 1 The vehicle with license plate A occludes most of the vehicle with license plate B, at this time, it may be unreliable to determine the position of the chassis center point completely depending on the detection result of the deep neural network model, because the deep neural network model may not output the vehicle chassis detection frame at this time.

[0064] For this case, the vehicle detection processing on the first image can also be implemented as follows:

[0065] performing vehicle detection processing on the first image using a pre-trained deep neural network model to obtain a vehicle detection frame corresponding to the target vehicle in the first image;

[0066] determining a sub-detection frame of a preset height ratio of the bottom of the vehicle detection frame, the preset height ratio being set according to the height of the vehicle chassis from the ground;

[0067] determining the position of the top center point of the sub-detection frame as the position of the vehicle chassis center point.

[0068] That is, at this time, the deep neural network model can be trained only for vehicle detection frame recognition, and the position coordinates of the vehicle chassis center point in the first image are indirectly determined based on the vehicle detection frame obtained thereby.

[0069] For ease of understanding, the case will be exemplarily described in combination with Figure 4 Figure 4 In the case shown in the figure, it is assumed that the deep neural network model detects one vehicle detection frame Q from the first image, and it is assumed that the preset height ratio is 1 / 5, that is, the height H of the vehicle detection frame is divided into 5 parts, and the bottom 1 / 5H forms a sub-detection frame q, and the center point of the top width of the sub-detection frame q (the circle point in the figure) is the chassis center point. The height of 1 / 5H is obtained by statistically analyzing the ratio of the height of the vehicle chassis from the ground to the height of the vehicle top from the ground.

[0070] ​In the embodiments of the present application, in addition to performing vehicle detection processing on each frame of image obtained by sampling to obtain the vehicle detection frame corresponding to each vehicle contained in each frame of image and the position of the center point set on the vehicle, vehicle tracking processing is also performed on continuous different frames of image. Because the vehicle detection processing can only determine whether a vehicle is contained in a frame of image and the corresponding position of the contained vehicle in the image, the identification information of the vehicle and the motion trajectory information of the vehicle cannot be obtained, and thus the identity of the vehicle between different frames of image cannot be known. Through the vehicle tracking processing, the identification information of the vehicle and the corresponding position of the same vehicle in different frames of image can be determined.

[0071] The vehicle detection processing and the vehicle tracking processing can cooperate with each other. For example, when a vehicle detection frame is obtained through the vehicle detection processing in a frame of image, the vehicle tracking processing can determine whether the vehicle is the same vehicle as a vehicle appearing in some previous images by recognizing the visual features (such as color, vehicle model, size, contour, etc.) of the vehicle and combining the moving speed and other feature information. The vehicle tracking processing can be implemented by referring to the related prior art, and thus detailed description is not given in the embodiments.

[0072] In actual application, the vehicle detection processing and the vehicle tracking processing can be executed by different processes. When a vehicle appears in the image of the roadside camera X, the vehicle detection processing and the vehicle tracking processing detect the vehicle for the first time, and the vehicle tracking processing assigns a tracking number to the vehicle as a kind of identification information of the vehicle. Meanwhile, the vehicle tracking processing can also perform license plate recognition processing to recognize the license plate number of the vehicle. If the license plate number can be recognized (in fact, the license plate number can not be recognized in some frames of image due to factors such as long distance, high moving speed, and occlusion between vehicles), the license plate number and the tracking number can be used as the identification information of the vehicle.

[0073] Optionally, in specific implementation, if the vehicle detection processing detects a vehicle in a frame of image, the vehicle tracking processing extracts the features such as visual features and moving speed of the vehicle and determines the vehicle identification of the vehicle, and then the vehicle identification and the corresponding vehicle detection frame can be associated and marked. In this way, through the joint processing of the vehicle detection and the vehicle tracking, the vehicle detection frame and the vehicle identification corresponding to the vehicle in different frames of image can be obtained, and thus the moving trajectory of the vehicle can be formed based on the positions of the vehicle detection frames corresponding to the same vehicle identification in continuous multiple frames of image.

[0074] The above introduces the process of vehicle detection and tracking for each frame of image collected by roadside camera X. Based on the above example of detecting the vehicle detection frame corresponding to the target vehicle in the first image and the position of the preset center point of the vehicle body (such as the center point of the vehicle chassis), in order to determine whether the target vehicle enters a certain roadside parking space covered by roadside camera X, first, taking the roadside parking space as the core, the position of the vehicle detection frame and the position of the vehicle chassis center point are compared with the parking space frame corresponding to each of the multiple roadside parking spaces covered by roadside camera X to determine whether the target vehicle meets the parking space occupation condition, in other words, to determine whether the state of the roadside parking space is in the occupied state.

[0075] The parking space occupation condition includes that the coincidence degree of the vehicle detection frame and the target parking space frame is greater than a set threshold, and the position of the vehicle chassis center point is located in the target parking space frame. Wherein, the target parking space frame is any one of the parking space frames corresponding to the multiple roadside parking spaces. If the above condition is met, it is determined that the target vehicle occupies the roadside parking space corresponding to the target parking space frame. The set threshold corresponding to the coincidence degree is, for example, a preset value such as 30%.

[0076] In actual application, the coincidence degree of the vehicle detection frame and the target parking space frame can be judged first, if the coincidence degree is greater than the set threshold, then it is judged whether the vehicle chassis center point is located in the target parking space frame, only when both conditions are met, the subsequent judgment process is executed. If the coincidence degree is less than the set threshold, the subsequent judgment is not performed, and it is determined that the target vehicle does not occupy the roadside parking space corresponding to the target parking space frame.

[0077] In the embodiment of the application, through the above-mentioned double judgment of coincidence degree and vehicle chassis center point, it can be more accurately determined whether the target vehicle occupies a roadside parking space, that is, to realize more accurate judgment of parking space entry. Because in fact, there may be a situation that a vehicle only passes through a roadside parking space without parking, and the passing vehicle generally cannot meet the above two judgment conditions. For example: large vehicles (such as buses, passenger cars) may meet the judgment condition of the above-mentioned coincidence degree when passing through the roadside parking space due to the large vehicle body, but may not meet the judgment condition of the chassis center point.

[0078] When the vehicle detection frame of the target vehicle and the position of the vehicle chassis center point satisfy the parking space occupation condition of the target parking space frame, in order to more accurately determine that the target vehicle is a vehicle that has parked in the corresponding roadside parking space, it is further necessary to determine whether the target vehicle has appeared in the warning area before entering the roadside parking space. If it has appeared in the warning area, it is finally determined that the target vehicle is a vehicle that has parked in the roadside parking space corresponding to the target parking space frame, i.e., the target vehicle has parked in the roadside parking space. If it has not appeared in the warning area, it is determined that the target vehicle has not parked in the roadside parking space.

[0079] In order to realize the above-mentioned determination of whether the target vehicle has appeared in the warning area, vehicle tracking information of the target vehicle in multiple second images collected within a preset time before the first image needs to be obtained. The preset time is, for example, 10 seconds, 15 seconds, etc. The multiple second images are multiple images sampled in sequence within the preset time, and are referred to as multiple second images for distinction from the first image. In combination with the above-mentioned introduction of vehicle detection and vehicle tracking processing, it can be understood that the vehicle tracking information includes the respective positioning positions of the vehicle identifier of the target vehicle in the multiple second images. The positioning position can be the position of the vehicle detection frame of the target vehicle in the corresponding second image, of course, a certain feature point (such as a center point, a corner point) on the vehicle detection frame can also be used as the positioning position, and the positioning position has been obtained based on the vehicle tracking processing and the vehicle detection processing. If it is determined according to the vehicle tracking information that the vehicle identifier of the target vehicle appears in the warning frame, it is determined that the target vehicle has parked in the roadside parking space corresponding to the target parking space frame.

[0080] Among them, as described above, the vehicle identifier can include a license plate number and a tracking number. In the determination process, it can be first determined whether the license plate number has appeared in the warning frame. If the license plate number is not successfully recognized due to a long distance, a high vehicle moving speed, etc., it can be further determined whether the tracking number appears in the warning frame.

[0081] The dual confirmation mode of the license plate number and the tracking number is used to determine whether the target vehicle has appeared in the warning area, which can improve the anti-interference performance of the determination result and overcome the interference of complex environments (such as shielding, high-speed movement, and long distance).

[0082] In summary, in the above scheme, a pre-warning area is defined outside the same side of the several continuous roadside parking spaces covered by the same roadside camera to assist in determining whether the vehicle enters or exits the parking space. Only when it is determined that the vehicle meets the conditions for occupying the roadside parking space and the vehicle meets the conditions for passing through the pre-warning area, it is finally determined that the vehicle enters the roadside parking space for parking, thereby improving the accuracy of the determination result. Moreover, when determining whether the vehicle occupies the parking space, the vehicle detection frame and a specific center point on the vehicle are combined to determine whether the vehicle occupies the parking space, thereby improving the accuracy of the determination result.

[0083] In an optional embodiment, to further improve the accuracy of the determination result of the vehicle entering the parking space for parking, based on the above examples of the first image and the plurality of second images, in combination with the vehicle detection frame and the vehicle chassis center point of the target vehicle detected in the first image, and the positioning position of the vehicle identifier of the target vehicle in the corresponding second image determined by tracking in the plurality of second images, it is further determined whether the target vehicle is stably parked in the roadside parking space after entering the roadside parking space corresponding to the target parking space frame. If yes, it is finally determined that the target vehicle enters the roadside parking space for parking. Otherwise, it is determined that the target vehicle only passes through the roadside parking space.

[0084] Based on this, determining that the target vehicle enters the roadside parking space corresponding to the target parking space frame can be implemented as follows:

[0085] Obtaining vehicle tracking information of the target vehicle in a plurality of third images collected within a preset time after the first image is obtained;

[0086] Respectively performing vehicle detection processing on the plurality of third images to obtain the positions of the vehicle detection frame and the vehicle body preset center point of the target vehicle in the plurality of third images in combination with the vehicle tracking information of the target vehicle in the plurality of third images;

[0087] If the positions of the vehicle detection frame and the vehicle body preset center point of the target vehicle in the plurality of third images meet the parking space occupation conditions of the target parking space frame, it is determined that the target vehicle enters the roadside parking space corresponding to the target parking space frame for parking.

[0088] The preset time is, for example, a preset value such as 10 seconds, 15 seconds, etc. The plurality of third images are similar to the plurality of second images, and do not mean that there are multiple images in the same frame, but different frames of images are respectively sampled within a corresponding time period.

[0089] In general, the vehicle tracking information here is mainly used to identify the same target vehicle. The vehicle detection frame contained in each frame of image and the position of the vehicle body preset center point (such as the vehicle chassis center point) are obtained by performing vehicle detection processing on multiple frames of third images respectively, and the vehicle identification corresponding to the vehicle detection frame contained in each frame of image can be determined in combination with the vehicle tracking processing result, so that the vehicle detection frame of the target vehicle can be determined. If it is determined that the target vehicle satisfies the condition of occupying the roadside parking space corresponding to the target parking space frame within the above-mentioned preset time based on the vehicle detection frame and the position of the vehicle chassis center point corresponding to the target vehicle in multiple frames of third images respectively, it is considered that the target vehicle is stably parked in the roadside parking space corresponding to the target parking space frame.

[0090] When it is finally determined that the target vehicle is parked in the roadside parking space corresponding to the target parking space frame, an entry record corresponding to the target vehicle can be generated, which includes the identification of the roadside parking space corresponding to the target parking space frame, the vehicle identification of the target vehicle, the entry time, and the entry video. The entry video at least includes the video segment sampled from the multiple frames of second images and first images, and of course, the video segment sampled from the multiple frames of third images can also be included. Further, the entry record can also include the entry trajectory of the target vehicle: based on the positioning positions corresponding to the vehicle identification of the target vehicle in the images obtained by sampling the entry video, the entry trajectory is generated.

[0091] In an optional embodiment, in order to save computing power, the following parking space occupation condition judgment strategy is provided:

[0092] For any frame of image i (such as the first image described above) sampled, if a vehicle detection frame and a vehicle chassis center point of a vehicle (referred to as a target vehicle) are detected in the image, and it is found that the vehicle detection frame is relatively far from the target parking space frame when comparing the coincidence degree of the vehicle detection frame and the target parking space frame, the image j for next time parking space occupation condition judgment for the target vehicle can be determined based on the sampling frequency. Generally, at least one frame of image is included between image j and image i. Assuming that it is determined that the above-mentioned judgment for the target vehicle needs to be performed after m frames of images, and m is greater than or equal to 1, after the vehicle detection frame and the vehicle chassis center point of the target vehicle are obtained in the image j after m frames apart based on the vehicle tracking and the vehicle detection processing, the coincidence degree of the vehicle detection frame and the target parking space frame is compared again, and it is determined whether the vehicle chassis center point is located within the target parking space frame.

[0093] Since the vehicle detection and the vehicle tracking are realized based on the determined deep neural network model and the vehicle tracking algorithm, the two processing processes have stable computing power consumption, and the parking space occupation condition judgment also needs computing power consumption. Through the above-mentioned strategy, the execution times of the judgment can be reduced, and the computing power consumption can be reduced.

[0094] Figure 5 A flow chart of a method for determining a vehicle leaving a parking space according to an embodiment of the present application is shown in FIG. 5, which can include the following steps: Figure 5

[0095] 501. Perform vehicle detection on the last one of the plurality of fourth images captured after the first image.

[0096] 502. If the target vehicle does not meet the parking space occupancy condition of the target parking space bounding box according to the vehicle detection result of the plurality of fourth images, obtain vehicle tracking information of the target vehicle in the plurality of fourth images.

[0097] 503. If the vehicle identification of the target vehicle appears in the early warning bounding box according to the vehicle tracking information of the target vehicle in the plurality of fourth images, determine that the target vehicle has left the road side parking space corresponding to the target parking space bounding box.

[0098] Based on the above example in the preceding embodiment, it is assumed that the target vehicle has parked in the road side parking space corresponding to the target parking space bounding box based on the first image and the plurality of second images captured before the first image.

[0099] It can be understood that in the case where the target vehicle has parked in the road side parking space corresponding to the target parking space bounding box based on the plurality of third images in the preceding embodiment, the first image in step 501 is replaced by the last one of the plurality of third images.

[0100] In summary, the "first image" in this step refers to the image finally captured when it is determined that the target vehicle has parked stably in the road side parking space corresponding to the target parking space bounding box.

[0101] For ease of description, the road side parking space corresponding to the target parking space bounding box is referred to as the target road side parking space. Then, in order to determine when the target vehicle has left the target road side parking space, vehicle detection and vehicle tracking need to be performed on the images captured by the road side camera X after the first image.

[0102] ​Assuming that in a frame of image collected at a later T1 moment, based on the vehicle detection result, it is found that the vehicle detection box and the vehicle chassis center point originally satisfying the above parking space occupation condition with the target parking space border no longer satisfy the parking space occupation condition, i.e., the coincidence degree of the vehicle detection box and the target parking space border is lower than the set threshold, and / or the vehicle chassis center point is no longer located within the target parking space border, it can be preliminarily considered that the target vehicle originally parked in the target roadside parking space may have driven out of the target roadside parking space. In order to ensure the reliability of the determination result, it is continued to be observed whether the vehicle detection box and the vehicle chassis center point of the target vehicle are not detected within the target parking space border in multiple frames of images collected after the T1 moment (assuming the time to T2), if so, it is finally considered that the target vehicle does not satisfy the occupation condition of the target roadside parking space. In the above example, the multiple frames of images collected between T1 and T2 are the above-mentioned multiple frames of fourth images.

[0103] It can be understood that, as described above, while the vehicle detection processing is performed on these frames of images, the vehicle tracking processing is also performed, so that the vehicle detection box and the vehicle chassis center point corresponding to the same vehicle identification in each frame of image of the target vehicle can be obtained.

[0104] After determining that the target vehicle no longer satisfies the occupation condition of the target roadside parking space based on the vehicle detection result and the vehicle tracking result of the multiple frames of fourth images, it is further determined whether the target vehicle appears in the warning area between T1 and T2 based on the vehicle tracking information of the target vehicle obtained from the multiple frames of fourth images, i.e., whether the vehicle identification of the target vehicle appears in the warning border in some images. Among them, the vehicle tracking information includes the respective positioning positions of the vehicle identification of the target vehicle in the multiple frames of fourth images. If it appears, it is determined that the target vehicle drives out of the target roadside parking space, at this time, a driving-out record corresponding to the target vehicle can be generated, and the driving-out record includes the identification of the target roadside parking space, the vehicle identification of the target vehicle, the driving-out time and the driving-out video, wherein the driving-out video at least includes a video segment sampled from the multiple frames of fourth images. The correspondence between the identification of the target roadside parking space and the corresponding target parking space border is pre-stored and can be determined by querying.

[0105] Since the roadside camera is arranged on the same side as the roadside parking space, after the target vehicle parks in the target roadside parking space, due to the shielding of other vehicles before and after the target vehicle, it is more difficult to determine whether the target vehicle leaves the target roadside parking space where it parks. Therefore, in the embodiment, the double confirmation mode of the vehicle detection box and the vehicle chassis center point is combined to determine whether the target vehicle no longer occupies the target roadside parking space, and whether the target vehicle passes through the warning area after not occupying the target roadside parking space is further combined for judgment, and multiple judgments can accurately determine the behavior of the target vehicle driving out of the target roadside parking space.

[0106] Based on the corresponding entry and exit records of the target vehicle, not only the target vehicle can be charged, but also an accurate evidence chain can be formed.

[0107] In the above embodiments, the judgment process of whether the vehicle enters or exits the parking space is illustrated in the scenario of the vehicle entering or exiting the roadside parking space. As described above, the scheme is also applicable to indoor and outdoor parking lot scenarios, that is, any scenario that requires determination of vehicle entry and exit of parking spaces. In summary, in these indoor and outdoor parking lot scenarios, a camera can be configured to capture multiple parking spaces (i.e., parking spaces). For each camera, the aforementioned warning area in the foregoing embodiments can be set. Referring to the processing process of the video image collected by each camera in the foregoing embodiments, the entry and exit of the vehicle in each parking space can be determined.

[0108] Therefore, in summary, the present application provides a scheme suitable for any parking space scenario to determine the entry and exit of the vehicle in the parking space, as follows:

[0109] Obtain a first image collected by a camera, wherein the camera covers multiple parking spaces;

[0110] Perform vehicle detection processing on the first image to obtain a vehicle detection frame corresponding to a target vehicle in the first image and a position of a preset center point of the vehicle body;

[0111] If it is determined that the vehicle detection frame and the position of the preset center point of the vehicle body meet the parking space occupation condition of a target parking space frame, obtain vehicle tracking information of the target vehicle in multiple second images collected within a preset time before the first image, wherein the vehicle tracking information includes a vehicle identifier of the target vehicle and a respective positioning position of the vehicle identifier in the multiple second images; wherein the target parking space frame is an image area boundary corresponding to any of the parking spaces in the camera image.

[0112] If it is determined according to the vehicle tracking information that the vehicle identifier appears in a preset warning frame, it is determined that the target vehicle enters the parking space corresponding to the target parking space frame; wherein the warning frame is an image area boundary corresponding to a warning area required to pass through the multiple parking spaces in the camera image, and the warning area is a road area surrounding the multiple parking spaces.

[0113] The specific implementation process of the scheme can refer to the related descriptions in the foregoing embodiments, which will not be repeated here.

[0114] The following will describe in detail the device for determining the entry and exit of the vehicle in the parking space according to one or more embodiments of the present application. Those skilled in the art can understand that these devices can be configured using commercially available hardware components through the steps taught by the present scheme.

[0115] Figure 6 A structural schematic diagram of a device for determining vehicle access to a parking space is provided for an embodiment of the present application, as shown in the figure, the device comprises an acquisition module 11, a detection module 12, and a processing module 13. Figure 6

[0116] The acquisition module 11 is configured to acquire a first image captured by a camera, wherein the camera covers a plurality of parking spaces.

[0117] The detection module 12 is configured to perform vehicle detection processing on the first image to obtain a vehicle detection frame corresponding to a target vehicle in the first image and a position of a vehicle body preset center point.

[0118] The processing module 13 is configured to, if it is determined that the vehicle detection frame and the position of the vehicle body preset center point meet a parking space occupation condition of a target parking space boundary, acquire vehicle tracking information of the target vehicle in a plurality of second images captured within a preset time before the first image, wherein the vehicle tracking information includes a vehicle identifier of the target vehicle and a respective positioning position of the vehicle identifier in the plurality of second images; and if it is determined according to the vehicle tracking information that the vehicle identifier appears in a preset warning boundary, determine that the target vehicle has parked in a parking space corresponding to the target parking space boundary; wherein the warning boundary is a boundary of an image region corresponding to a warning area required to access the plurality of parking spaces in a shooting picture of the camera, the warning area is a road area surrounding the plurality of parking spaces, and the target parking space boundary is a boundary of an image region corresponding to any of the parking spaces in the shooting picture of the camera.

[0119] Optionally, the parking space occupation condition comprises that an overlap degree between the vehicle detection frame and the target parking space boundary is greater than a set threshold, and the position of the vehicle body preset center point is located within the target parking space boundary.

[0120] Optionally, the vehicle body preset center point comprises a vehicle chassis center point.

[0121] Optionally, the detection module 12 is specifically configured to perform vehicle detection processing on the first image using a pre-trained deep neural network model to obtain a vehicle detection frame corresponding to a target vehicle in the first image and a vehicle chassis detection frame, wherein the vehicle detection frame refers to a detection frame containing a complete vehicle; and determine a position of a center point of the vehicle chassis detection frame as the position of the vehicle chassis center point.

[0122] ​Optionally, the detection module 12 is specifically configured to: perform vehicle detection processing on the first image by using a pre-trained deep neural network model to obtain a vehicle detection frame corresponding to the target vehicle in the first image; determine a sub-detection frame of a bottom preset height ratio of the vehicle detection frame, the preset height ratio being set according to the height of a vehicle chassis from the ground; and determine the position of a top center point of the sub-detection frame as the position of the vehicle chassis center point.

[0123] Optionally, the vehicle identifier includes a license plate number and a tracking number, the tracking number being assigned in a vehicle tracking processing process on the image, and the same vehicle having the same tracking number.

[0124] Optionally, the processing module 13 is specifically configured to: obtain vehicle tracking information of the target vehicle in a plurality of third images collected within a preset time after the first image; perform vehicle detection processing on the plurality of third images respectively to obtain vehicle detection frames and positions of vehicle body preset center points corresponding to the target vehicle in the plurality of third images respectively in combination with the vehicle tracking information of the target vehicle in the plurality of third images; and determine that the target vehicle parks in a parking space corresponding to the target parking space frame if the vehicle detection frames and the positions of the vehicle body preset center points corresponding to the target vehicle in the plurality of third images respectively meet the parking space occupation condition.

[0125] Optionally, the device further includes a recording module configured to generate a parking-in record corresponding to the target vehicle, the parking-in record including an identifier of a parking space corresponding to the target parking space frame, the vehicle identifier, a parking-in time, and a parking-in video, wherein the parking-in video at least includes a video segment sampled from the plurality of second images and the first image.

[0126] Optionally, the detection module 12 is further configured to perform vehicle detection processing on a plurality of fourth images collected after the first image respectively. The processing module 13 is further configured to: if it is determined that the target vehicle does not meet the parking space occupation condition of the target parking space frame according to a vehicle detection result of the plurality of fourth images, obtain vehicle tracking information of the target vehicle in the plurality of fourth images; and if it is determined that the vehicle identifier appears in the early warning frame according to the vehicle tracking information of the target vehicle in the plurality of fourth images, determine that the target vehicle drives away from the parking space corresponding to the target parking space frame.

[0127] Optionally, the recording module is further configured to generate a parking-out record corresponding to the target vehicle, the parking-out record including an identifier of a parking space corresponding to the target parking space frame, the vehicle identifier, a parking-out time, and a parking-out video, wherein the parking-out video at least includes a video segment sampled from the plurality of fourth images.

[0128] Figure 6 The apparatus shown can perform the steps provided in the foregoing embodiments, and the detailed execution process and technical effects refer to the descriptions in the foregoing embodiments, which will not be repeated here.

[0129] In one possible design, the above Figure 6 The structure of the apparatus for determining vehicle access to a berth shown can be implemented as an electronic device. As Figure 7 The electronic device can include a processor 21, a memory 22, and a communication interface 23. The memory 22 stores executable code, which, when executed by the processor 21, causes the processor 21 to implement at least the method for determining vehicle access to a berth as provided in the foregoing embodiments.

[0130] In addition, the embodiment of the present application provides a non-transitory machine readable storage medium, which stores executable code. When the executable code is executed by a processor of an electronic device, the processor can implement at least the method for determining vehicle access to a berth as provided in the foregoing embodiments.

[0131] The apparatus embodiments described above are only schematic, and the network elements illustrated as separate components can or can not be physically separate. Some or all of the modules can be implemented in accordance with the actual needs, and those of ordinary skill in the art can understand and implement without creative labor.

[0132] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be implemented by means of an appropriate general-purpose hardware platform, and of course can also be implemented by means of a combination of hardware and software. Based on such understanding, the above technical solutions can be embodied in the form of a computer program product, and the present application can be implemented in the form of a computer program product containing computer usable program code in one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.).

[0133] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for determining the entry and exit of vehicles from parking spaces, characterized in that, include: Acquire the first image captured by the camera, which covers multiple berths; The first image is processed for vehicle detection to obtain the vehicle detection box and the position of the preset center point of the vehicle body corresponding to the target vehicle in the first image. If it is determined that the position of the vehicle detection frame and the preset center point of the vehicle body meets the parking space occupancy condition of the target parking space border, then the vehicle tracking information of the target vehicle in multiple frames of second images collected within a preset time before the first image is obtained. The vehicle tracking information includes the positioning position of the vehicle identifier of the target vehicle in each of the multiple frames of second images. The target parking space border is the image area boundary corresponding to any parking space in the camera's captured image. The target parking space border is marked in advance based on the image area occupied by the roadside parking space boundary line actually drawn on the roadway in the captured image when no vehicle enters the multiple parking spaces covered by the camera. If the vehicle identification is determined to appear within a preset warning border based on the vehicle tracking information, then the target vehicle is determined to have entered and parked in the parking space corresponding to the target parking space border. The warning border is the image area boundary in the camera's view corresponding to the warning area that the vehicle must traverse to enter or exit the multiple parking spaces. The warning area is the driveway area surrounding the multiple parking spaces. The positional relationship between the warning area and the driveway area of ​​the multiple parking spaces is that there is no inclusion relationship and no intersection relationship.

2. The method according to claim 1, characterized in that, The parking space occupancy conditions include: the overlap between the vehicle detection frame and the target parking space border is greater than a set threshold, and the position of the vehicle body's preset center point is located within the target parking space border.

3. The method according to claim 1, characterized in that, The step of determining that the target vehicle drives into the parking space corresponding to the border of the target parking space includes: Vehicle tracking information of the target vehicle in multiple frames of third images collected within a preset time after acquiring the first image; Vehicle detection processing is performed on the multiple frames of the third image respectively, so as to obtain the position of the vehicle detection box and the preset center point of the vehicle body of the target vehicle in the multiple frames of the third image by combining the vehicle tracking information of the target vehicle in the multiple frames of the third image respectively; If the positions of the vehicle detection frame and the preset center point of the vehicle body corresponding to the target vehicle in the multi-frame third image meet the parking space occupancy conditions, then it is determined that the target vehicle drives into the parking space corresponding to the target parking space frame.

4. The method according to claim 1, characterized in that, The method further includes: An entry record corresponding to the target vehicle is generated. The entry record includes the identifier of the parking space corresponding to the border of the target parking space, the vehicle identifier, the entry time, and the entry video. The entry video includes at least a video segment of the sampled multi-frame second image and the first image.

5. The method according to claim 1, characterized in that, The method further includes: Vehicle detection processing is performed on multiple fourth images acquired after the first image; If it is determined from the vehicle detection results of the multi-frame fourth image that the target vehicle does not meet the parking space occupancy condition of the target parking space border, then the vehicle tracking information of the target vehicle in the multi-frame fourth image is obtained. If, based on the vehicle tracking information of the target vehicle in the fourth frame of the multi-frame image, it is determined that the vehicle identifier appears within the warning border, then it is determined that the target vehicle has left the parking space corresponding to the target parking space border.

6. The method according to claim 5, characterized in that, The method further includes: Generate a departure record corresponding to the target vehicle. The departure record includes the identifier of the parking space corresponding to the border of the target parking space, the vehicle identifier, the departure time, and the departure video. The departure video includes at least a video segment of the sampled multi-frame fourth image.

7. The method according to any one of claims 1 to 6, characterized in that, The preset center point of the vehicle body includes the center point of the vehicle chassis.

8. The method according to claim 7, characterized in that, The vehicle detection processing of the first image includes: The first image is processed by a pre-trained deep neural network model to detect vehicles, so as to obtain the vehicle detection box and vehicle chassis detection box corresponding to the target vehicle in the first image. The vehicle detection box refers to the detection box containing the complete vehicle. The position of the center point of the vehicle chassis detection frame is determined as the position of the center point of the vehicle chassis.

9. The method according to claim 7, characterized in that, The vehicle detection processing of the first image includes: The first image is processed by a pre-trained deep neural network model to detect vehicles, so as to obtain the vehicle detection box corresponding to the target vehicle in the first image. A sub-detection frame with a preset height percentage at the bottom of the vehicle detection frame is determined, wherein the preset height percentage is set according to the height of the vehicle chassis from the ground. The position of the top center point of the sub-detection frame is determined as the position of the vehicle chassis center point.

10. The method according to any one of claims 1 to 6, characterized in that, The vehicle identification includes a license plate number and a tracking number. The tracking number is assigned during the vehicle tracking process of the image, and the same vehicle has the same tracking number.

11. An electronic device, characterized in that, include: The system includes a memory, a processor, and a communication interface; wherein the memory stores executable code that, when executed by the processor, causes the processor to perform the method for determining vehicle entry and exit from parking spaces as described in any one of claims 1 to 10.

12. A non-transitory machine-readable storage medium, characterized in that, The non-transitory machine-readable storage medium stores executable code that, when executed by a processor of an electronic device, causes the processor to perform the method for determining a vehicle's entry and exit from a parking space as described in any one of claims 1 to 10.

13. A method for determining the entry and exit of vehicles from roadside parking spaces, characterized in that, include: Acquire a first image captured by a roadside camera, wherein the roadside camera includes a camera installed on the same side as a roadside parking space, and the roadside camera covers multiple roadside parking spaces; The first image is processed for vehicle detection to obtain the vehicle detection box and the position of the preset center point of the vehicle body corresponding to the target vehicle in the first image. If it is determined that the position of the vehicle detection frame and the preset center point of the vehicle body meets the parking space occupancy condition of the target parking space border, then the vehicle tracking information of the target vehicle in multiple frames of second images collected within a preset time before the first image is obtained. The vehicle tracking information includes the positioning position of the vehicle identifier of the target vehicle in each of the multiple frames of second images. The target parking space border is the image area boundary corresponding to any of the roadside parking spaces in the image captured by the roadside camera. The target parking space border is marked in advance based on the image area occupied by the roadside parking space boundary line actually drawn on the roadway in the image captured by the camera when no vehicle enters the multiple parking spaces it covers. If the vehicle identification is determined to appear within a preset warning border based on the vehicle tracking information, then the target vehicle is determined to have entered and parked in the roadside parking space corresponding to the target parking space border. The warning border is the image area boundary of the warning area that must be traversed to enter or exit the multiple roadside parking spaces, as captured by the roadside camera. The warning area is the driveway area surrounding the multiple roadside parking spaces. The positional relationship between the warning area and the driveway area of ​​the multiple parking spaces is that there is no inclusion relationship and no intersection relationship.

Images