Continuous berth video acquisition device deployment method, network device and storage medium

Abstract

The application discloses a continuous parking space video acquisition device deployment method, network equipment and a storage medium. The method comprises the following steps: acquiring the farthest license plate recognition distance of a gun-type camera; acquiring the nearest end video acquisition distance and the farthest end video acquisition distance of the gun-type camera under the maximum focal length; determining the video acquisition range of the gun-type camera according to the farthest license plate recognition distance, the nearest end video acquisition distance and the farthest end video acquisition distance; calculating the continuous parking space in the video acquisition range of the gun-type camera according to the predetermined parking space size, and deploying the gun-type camera at the road traffic pole for the continuous parking space. The application solves the problems of high deployment cost of the parking space video acquisition system and poor accuracy of the parking space video acquisition result in the prior art, and can effectively reduce the deployment cost of the parking space video acquisition system and improve the accuracy of the parking space video acquisition result.

Description

Technical Field

[0001] This application relates to the field of intelligent parking management, specifically to a method for deploying video acquisition equipment for continuous parking spaces, network equipment, and storage media. Background Technology

[0002] Traditional smart parking solutions often rely on manual labor and handheld devices to input vehicle entry and exit orders. This approach is inefficient and cumbersome in the face of rapidly evolving smart parking trends. To address the inefficiency and cumbersome nature of traditional smart parking solutions, a scheme using a combination of PTZ (polygonal) and bullet cameras for video capture has been proposed. This scheme utilizes a telephoto bullet camera to cover a wide-angle area. When a vehicle enters the frame, the PTZ camera zooms in to track the vehicle and capture its license plate. When a vehicle leaves a parking space, the bullet camera, in conjunction with the PTZ camera, rotates to a preset coordinate to track and capture the license plate. However, this high-position PTZ / bullet camera video capture solution suffers from data loss when multiple vehicles enter and exit simultaneously, and unstable license plate recognition, ultimately leading to poor accuracy. Additionally, a scheme using low-position video stakes for video detection of parking space status and complete parking events (vehicle entry and exit) has also been proposed. This low-position video stake solution can adapt to parking environments under various lighting conditions, such as heavy rain, heavy snow, fog, and night. It has high recognition accuracy. However, because the video stake is installed at a low position on the road, it is easily obstructed and damaged by external forces. Summary of the Invention

[0003] This application provides a method for deploying continuous berth video acquisition equipment, a network device, and a storage medium to at least solve the problems of high deployment cost and poor accuracy of berth video acquisition results in related technologies.

[0004] According to one aspect of this application, a method for deploying a continuous parking space video acquisition device is provided, comprising: obtaining the farthest license plate recognition distance of a bullet camera; obtaining the nearest and farthest video acquisition distances of the bullet camera at its maximum focal length; determining the video acquisition range of the bullet camera based on the farthest license plate recognition distance, the nearest video acquisition distance, and the farthest video acquisition distance; calculating the continuous parking spaces for which the bullet camera can acquire video within the video acquisition range based on a predetermined parking space size; and deploying the bullet camera at a road traffic pole for the continuous parking spaces.

[0005] According to another aspect of this application, a network device is also provided, including a memory and a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the above-described method steps.

[0006] According to another aspect of this application, a readable storage medium is also provided, on which computer instructions are stored, wherein the computer instructions, when executed by a processor, implement the above-described method steps.

[0007] In this embodiment, the farthest license plate recognition distance of the bullet camera is obtained; the nearest and farthest video acquisition distances of the bullet camera at its maximum focal length are obtained; the video acquisition range of the bullet camera is determined based on the farthest license plate recognition distance, the nearest video acquisition distance, and the farthest video acquisition distance; the continuous parking spaces within the video acquisition range of the bullet camera are calculated based on the predetermined parking space size, and bullet cameras are deployed at road traffic poles for the continuous parking spaces. Through the continuous parking space video acquisition equipment deployment scheme provided in this application, firstly, by using the imaging effect of the physical parking spaces in the bullet camera lens and the recognition effect of artificial intelligence, the number of effectively covered parking spaces in the image can be obtained, which can improve the accuracy of vehicle recognition results on the effectively covered parking spaces; secondly, one bullet camera can capture video from multiple continuous parking spaces, which saves manpower and financial costs compared to setting a low-level video stake for each parking space in the prior art; thirdly, since the bullet camera is deployed above the road traffic poles, it can effectively avoid external human-caused obstruction and damage. Attached Figure Description

[0008] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0009] Figure 1 This is a flowchart of a continuous berth video acquisition device deployment method according to an embodiment of this application;

[0010] Figure 2 This is a schematic diagram of the video acquisition distance range of a bullet camera according to an embodiment of this application;

[0011] Figure 3 This is a schematic diagram of the deployment of a bullet camera according to an embodiment of this application;

[0012] Figure 4 This is a schematic diagram of the structure of a continuous berth video acquisition system according to an embodiment of this application. Detailed Implementation

[0013] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0014] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0015] Traditional smart parking solutions often rely on manual labor and handheld devices to input vehicle entry and exit orders. This approach is inefficient and cumbersome in the face of rapidly evolving smart parking trends. To address these issues, a high-position camera-based video capture solution has been proposed. However, this solution suffers from data loss due to multiple vehicles entering and exiting simultaneously, and unstable license plate recognition. To resolve the unstable license plate recognition issue, existing technologies have proposed installing low-position video surveillance poles to monitor parking space status and complete vehicle entry and exit events. However, this solution is costly in terms of deployment manpower and resources, and the video surveillance poles are easily damaged.

[0016] To address the aforementioned problems in existing technologies, this application proposes a rapid deployment scheme for roadside parking in complex road conditions, based on bullet cameras and AI video deep learning technology. The scheme calculates and analyzes influencing factors such as the angle between the parking space and the roadway, the number of consecutive parking spaces, the size parameters of the parking spaces, and the focal length and field of view parameters of the bullet camera to determine the installation position and height of the high-position video system. Ultimately, a standardized design reference is provided, significantly shortening the required on-site manpower input period. This offers a more stable, accurate, and cost-effective design method for smart parking solutions.

[0017] The deployment scheme of the continuous berth video acquisition equipment provided in this application is described below with reference to the accompanying drawings. Figure 1 This is a flowchart of the steps of the continuous berth video acquisition device deployment method according to an embodiment of this application. The following is a description of the steps. Figure 1 The steps included in the process shown in the diagram will be explained.

[0018] Step S102: Obtain the furthest license plate recognition distance of the bullet camera;

[0019] Roadside parking is increasingly being incorporated into urban parking management in various cities. Based on the angle of inclination between the parking space and the driving lane, roadside parking can be categorized into three types: parallel parking, perpendicular parking, and angled parking. Parallel parking is the most common. This application proposes a continuous parking space video acquisition equipment deployment scheme, using parallel parking as an example. To obtain a highly practical and accurate continuous parking space video acquisition equipment deployment scheme, the factors affecting video image quality and algorithm analysis accuracy are first analyzed, followed by a detailed study of the characteristics and actual influence coefficients of each factor. Then, through linear programming fitting of a large amount of experimental data, the optimal range of imaging quality and algorithm recognition rate for bullet cameras under different environments is derived. Finally, combining parameters such as the height of road traffic poles, the field of view of the bullet camera, and the size of the parking space, the continuous parking spaces within the video acquisition range of the bullet camera are calculated, ultimately outputting the continuous parking space video acquisition equipment deployment scheme.

[0020] The specific model of the bullet camera can be flexibly selected by those skilled in the art based on factors such as camera parameters, cost, and application scenarios. In this embodiment, the method for selecting the target model of the bullet camera is not specifically limited. Once the target model of the bullet camera is selected, the field of view of the bullet camera can be determined.

[0021] The size of license plates is fixed and uniform. According to national standards, the size of license plates for small cars is 440 mm * 140 mm. Common license plates generally consist of 7 characters: 1 Chinese character and 6 letters or numbers. The distance between the second and third characters is half a character's width. The width of a single character is 60 mm, and the height is 120 mm.

[0022] In my country's urban areas, the height of road traffic poles and signs is generally 6-8 meters, and the length of the horizontal arm is also 6-8 meters. Therefore, this embodiment uses an 8-meter height and 8-meter horizontal arm as an example for illustrating the installation space of the bullet camera. According to the spatial geometric calculation formula, when the diagonal of a cuboid is 70 meters and the two horizontal sides are 8 meters, the third horizontal side is approximately 68 meters. Combining this with the requirements of the national standard "Specifications for Setting Up Parking Spaces on Urban Roads," the length of parking spaces for small cars is between 6 and 6.5 meters; using the maximum of 6.5 meters, the maximum license plate recognition distance of the bullet camera is 68 meters, equivalent to the 10th parking space.

[0023] Step S104: Obtain the closest and furthest video acquisition distances of the bullet camera at its maximum focal length;

[0024] After determining the maximum license plate recognition distance of the bullet camera, the actual field of view range of the bullet camera needs to be determined based on the field of view and focal length range of the bullet camera.

[0025] Because bullet cameras are installed on road traffic poles at a certain height, and license plates are perpendicular to the road surface, when the license plate is too close to the bullet camera and not on the same horizontal plane, license plate distortion will occur due to the imaging angle. Therefore, after determining the actual field of view, it is necessary to import license plate samples with different distorted angles into the algorithm terminal for recognition and statistical analysis, and further calculate the nearest and farthest video acquisition distances of the bullet camera at its maximum focal length based on the height of the pole where the bullet camera is installed.

[0026] Step S106: Determine the video acquisition range of the bullet camera based on the farthest license plate recognition distance, the nearest video acquisition distance, and the farthest video acquisition distance.

[0027] Based on the bullet camera's maximum license plate recognition distance, field of view, and focal length range, the actual field of view range of the bullet camera was determined. Then, the nearest and farthest video acquisition distances at the maximum focal length were further determined. Finally, by combining the results of finding the two farthest and nearest subsets, the video acquisition range of the bullet camera was determined.

[0028] Step S108: Calculate the continuous parking spaces within the video capture range of the bullet camera based on the predetermined parking space size, and deploy bullet cameras at the road traffic poles for the continuous parking spaces.

[0029] It should be noted that the aforementioned bullet camera can be installed at a predetermined position on a road traffic pole, and this predetermined position can be flexibly adjusted according to actual needs; the aforementioned road traffic pole is a pole used to install the bullet camera.

[0030] Furthermore, regarding berth layout, the arrangement methods include, but are not limited to, parallel berths, perpendicular berths, and angled berths. Subsequently, the number of consecutive berths captured by the bullet camera within its video capture range can be calculated based on the berth layout. The specific implementation method is as follows:

[0031] After determining the video capture range, if the parking spaces are arranged in parallel, the number of consecutive parking spaces can be obtained by dividing the difference between the upper and lower limits of the video capture range by the width of a single parking space.

[0032] After determining the video capture range, if the parking spaces are arranged in a vertical parking pattern, the number of consecutive parking spaces can be obtained by dividing the difference between the upper and lower limits of the video capture range by the length of a single parking space.

[0033] After determining the video capture range, if the parking spaces are arranged in an angled manner, a distance value can be determined by the difference between the upper and lower limits of the video capture range and the angle of the parking space. Then, by dividing this distance value by the width of a single parking space, the number of consecutive parking spaces can be obtained.

[0034] It should be noted that the above is only a plan for deploying a single bullet camera to identify continuous berths within the video capture range. In actual implementation, multiple bullet cameras can be deployed to capture video from continuous berths within a larger video capture range.

[0035] Through the above steps, firstly, by using the imaging effect of the physical parking spaces in the lens of the bullet camera and the recognition effect of artificial intelligence, the number of effectively covered parking spaces in the image can be obtained, which can improve the accuracy of vehicle recognition results on the effectively covered parking spaces; secondly, one bullet camera can collect video from multiple consecutive parking spaces, which can save manpower and financial costs compared to the existing technology of setting a low-level video stake for each parking space; thirdly, since the bullet camera is deployed above the road traffic poles, it can effectively avoid external human-caused obstruction and damage.

[0036] In one optional embodiment, the farthest license plate recognition distance of the bullet camera is obtained, and the specific implementation steps are as follows: within the depth of field of the bullet camera, the farthest license plate recognition distance corresponding to the preset license plate recognition accuracy is obtained.

[0037] It should be noted that the aforementioned maximum license plate recognition distance refers to the furthest straight-line distance for license plate recognition by the bullet camera while ensuring a preset license plate recognition accuracy rate. This furthest straight-line distance is the farthest straight-line distance between the bullet camera and the license plate. Within the depth of field of the bullet camera, the image clarity can be effectively guaranteed. Based on this, the corresponding furthest license plate recognition distance is obtained in conjunction with the preset license plate recognition accuracy rate. Furthermore, the preset license plate recognition accuracy rate may include, but is not limited to, 90%, 95%, 96%, and 98%.

[0038] In the embodiments of this application, a more accurate maximum license plate recognition distance is obtained by combining the depth range of the bullet camera with a preset license plate recognition accuracy.

[0039] In one optional embodiment, within the depth of field of the bullet camera, the furthest license plate recognition distance corresponding to a preset license plate recognition accuracy is obtained. The specific implementation steps are as follows:

[0040] Step S11: Based on the field of view, pixel size, and license plate size of the bullet camera, determine the number of pixels in the image area occupied by license plates at different distances from the bullet camera.

[0041] Step S13: Determine the correspondence between the number of pixels within the image area occupied by each license plate and the license plate recognition accuracy.

[0042] Step S15: Based on the correspondence and the preset license plate recognition accuracy, the farthest license plate recognition distance of the bullet camera is obtained by filtering.

[0043] In other words, firstly, using the field of view, pixel size, and license plate size of the bullet camera, the number of pixels within the frame occupied by license plates at different distances from the camera is calculated. Then, the correspondence between the number of pixels within the frame occupied by each license plate and the license plate recognition accuracy is determined. Finally, using these correspondences and a preset license plate recognition accuracy rate, the maximum license plate recognition distance for the bullet camera is selected. It should be noted that the pixel size of the bullet camera affects its depth of field to some extent.

[0044] In one optional embodiment, based on the field of view, pixel size, and license plate size of the bullet camera, the number of pixels within the image area occupied by license plates at different distances from the bullet camera is determined. The specific implementation steps are as follows:

[0045] Step S21: Calculate the size of each license plate after imaging by the bullet camera based on the field of view of the bullet camera, the size of the license plate, and the distance from each license plate to the bullet camera.

[0046] Using the known field of view of the bullet camera, the size of the license plate, and the distance from each license plate to the bullet camera, the size of each license plate after being imaged by the unknown bullet camera can be calculated.

[0047] Step S23: Calculate the number of pixels within the frame area occupied by each license plate based on the size of each license plate after imaging by the bullet camera and the pixel size of the bullet camera.

[0048] Since the size of each license plate after imaging by a bullet camera is equal to the product of the number of pixels in the frame occupied by each license plate and the pixel size of the bullet camera, the number of pixels in the frame occupied by each license plate can be calculated by dividing the size of each license plate after imaging by the pixel size of the bullet camera.

[0049] In the embodiments of this application, by utilizing parameters such as the field of view of the bullet camera, pixel size, distance from each license plate to the bullet camera, and license plate size, the number of pixels within the frame occupied by each license plate can be accurately obtained.

[0050] In one optional embodiment, the correspondence between the number of pixels within the image area occupied by each license plate and the license plate recognition accuracy is determined, and the specific implementation steps are as follows:

[0051] Step S31: Recognize the content of each license plate within the frame area to obtain the license plate content recognition result for each license plate;

[0052] Step S33: Based on the license plate content recognition results of each license plate, obtain the license plate recognition accuracy of each license plate;

[0053] Step S35: Establish the correspondence between the number of pixels within the image area occupied by each license plate and the license plate recognition accuracy.

[0054] In other words, for the number of pixels occupied by a license plate in the image, a preset number of license plate samples are adjusted to the size indicated by the number of pixels occupied by the license plate; secondly, the adjusted license plate samples are input into the license plate recognition model to obtain the license plate content recognition result; finally, based on the license plate content recognition result corresponding to each license plate sample, the correspondence between the number of pixels occupied by the license plate in the image and the license plate recognition accuracy is determined.

[0055] It should be noted that the license plate recognition model can extract the license plate area based on the license plate texture and the projection method, then segment and recognize the characters in the license plate area, and finally output the license plate content recognition result.

[0056] In the embodiments of this application, the license plate content within the screen area occupied by each license plate is identified to obtain the license plate content recognition result of each license plate. Then, the license plate recognition accuracy corresponding to the license plate content recognition result of each license plate is obtained. Finally, the correspondence between the number of pixels within the screen area occupied by each license plate and the license plate recognition accuracy is established.

[0057] In one optional embodiment, the maximum license plate recognition distance of the bullet camera is obtained by filtering based on the correspondence and the preset license plate recognition accuracy. The specific implementation steps are as follows:

[0058] Step S41: Construct the trend of license plate pixel count as a function of license plate recognition accuracy based on the correspondence relationships;

[0059] The correspondence between the number of pixels within the image area occupied by each license plate and the license plate recognition accuracy is converted into a trend of the number of license plate pixels (corresponding to the number of pixels within the image area occupied by the license plate) changing with the license plate recognition accuracy (corresponding to the license plate recognition accuracy). The trend of the number of license plate pixels changing with the license plate recognition accuracy is represented by a curve graph. For example, in the curve graph, the horizontal axis represents the number of license plate pixels and the vertical axis represents the license plate recognition accuracy.

[0060] Step S43: Compare the trend of change with the preset license plate recognition accuracy to determine the farthest straight-line distance for license plate recognition by the bullet camera;

[0061] The aforementioned longest straight-line distance for license plate recognition is the longest straight-line distance between the bullet camera and the license plate.

[0062] Step S45: Based on the height of the road traffic poles and traffic signs, determine the pole height and crossarm dimensions for mounting the bullet camera;

[0063] Step S47: Determine the farthest license plate recognition distance of the bullet camera based on the pole height, crossarm size, and the farthest license plate recognition straight-line distance.

[0064] In the embodiments of this application, the farthest license plate recognition distance of the bullet camera can be accurately selected based on the correspondence and the preset license plate recognition accuracy.

[0065] Table 1 shows an example of how the number of license plate pixels changes with license plate recognition accuracy:

[0066] Table 1

[0067] Distance from license plate to camera Number of pixels in the width of the license plate License plate recognition rate 1 20 meters 369.6 99.20％ 2 30 meters 246.7 98.80％ 3 40 meters 184.3 98.10％ 4 50 meters 147.2 97.1％ 5 60 meters 123.8 96.00％ 6 70 meters 105.3 94.70％ 7 80 meters 92.3 91.30％ 8 90 meters 82.1 87.50％ 9 100 meters 73.8 80.70％ 10 110 meters 67.6 70.40％ 11 120 meters 61.4 58.90％

[0068] Analysis of the trends in Table 1 shows that the recognition rate reaches 95% when the license plate width reaches 100 pixels. Furthermore, the improvement in recognition rate with increasing the number of pixels tends to plateau. Since the sample data already includes features such as blurriness and occlusion that prevent license plate recognition, a 95% recognition rate is chosen as the preset license plate recognition accuracy. A 95% recognition rate corresponds to a maximum license plate recognition distance of approximately 70 meters for the bullet camera in the table; therefore, the maximum license plate recognition distance for the bullet camera can be determined to be 70 meters.

[0069] It should be noted that the data in Table 1 is only an example. In actual implementation, the number of license plate pixels and the trend of license plate recognition accuracy will vary depending on the model of the selected camera, the size of the license plate, etc.

[0070] In addition, by analyzing the trend of the relationship between the recognition distance and the recognition accuracy of the bullet camera through the license plate recognition model and a large number of discrete license plate samples, the determined maximum license plate recognition distance of the bullet camera is highly reliable.

[0071] In one optional embodiment, the nearest and farthest video acquisition distances of the bullet camera at its maximum focal length are obtained. The specific implementation steps are as follows: the nearest and farthest video acquisition distances of the bullet camera are calculated based on the field of view of the bullet camera and the height of the pole used to mount the bullet camera.

[0072] The aforementioned field of view determines the range of the video image captured by the bullet camera. The field of view is the angle formed by the two edges of the maximum range through which the image of the target object can pass through the lens. The field of view determines the camera's field of view; a larger field of view means a wider field of view. The field of view can be measured using three data indicators: horizontal angle (from the left to the right of the image), vertical angle (from the top to the bottom of the image), and oblique angle (from one corner of the image to the opposite corner).

[0073] In the embodiments of this application, by combining the field of view of the bullet camera and the height of the pole used to mount the bullet camera, a more accurate calculation of the nearest and farthest video acquisition distances of the bullet camera is obtained.

[0074] In one optional embodiment, the nearest and farthest video acquisition distances of the bullet camera are calculated based on the field of view of the bullet camera and the height of the pole used to mount the bullet camera. The specific implementation steps are as follows:

[0075] Step S51: Calculate the vertical field of view range of the bullet camera within the standard focal length range;

[0076] Step S53: Based on the height of the pole on which the bullet camera is installed and the vertical field of view, calculate the nearest and farthest video acquisition distances of the bullet camera.

[0077] In other words, by using the field of view of the bullet camera as the vertical field of view range of the bullet camera within the standard focal length range, and combining the height of the pole on which the bullet camera is installed with the vertical field of view range, the nearest and farthest video acquisition distances of the bullet camera can be calculated.

[0078] In one optional embodiment: the above-mentioned gun-type camera is an infrared zoom gun-type camera, the standard focal length range of the lens of the infrared zoom gun-type camera is 8-56mm, and the resolution of the lens is not less than 4 million pixels.

[0079] The lenses mentioned above can be categorized into six types based on focal length: standard lenses, wide-angle lenses, fisheye lenses, telephoto lenses, reflective telephoto lenses, and macro lenses. Different lens structures and focal lengths are applied in different environments depending on actual needs and cost requirements. In practical implementation, a mature and stable infrared zoom bullet camera with automatic aperture adjustment, widely used in the market, can be selected. The corresponding lens should have a 1 / 1.8' sensor area and an f / 1.4 aperture, with a standard focal length range of 8-56mm. Furthermore, from the perspective of image clarity, the number of pixels in a license plate image at a fixed distance depends on the resolution of the bullet camera. Higher resolution results in a larger number of pixels for the same license plate area, leading to greater clarity. However, higher resolution also requires more computation for license plate recognition, resulting in longer processing times and higher chip computing power consumption. This significantly reduces the algorithm's concurrent computation capability when handling multiple vehicles, potentially causing vehicles to be missed. Meanwhile, taking into account the cost issues of large-scale operation, this solution selects an infrared zoom gun-type camera with a lens resolution of 4 megapixels.

[0080] In an optional embodiment, after calculating the continuous parking spaces within the video capture range of the bullet camera based on the predetermined parking space size, and deploying the bullet camera at the road traffic pole for the continuous parking spaces, the continuous parking space video capture device deployment method provided in this application embodiment further includes the following steps:

[0081] Step S110: If the number of berths to be video-captured exceeds the number of consecutive berths for video capture, determine the remaining berths to be video-captured, wherein the remaining berths to be video-captured are the berths to be video-captured other than the consecutive berths for video capture.

[0082] Step S112: For the remaining parking spaces to be video-captured, at least one bullet camera is deployed at at least one road traffic pole.

[0083] In the embodiments of this application, if the deployed bullet cameras are insufficient to capture video of all parking spaces, one or more bullet cameras can be added at the corresponding road traffic poles for the remaining parking spaces to be captured, thereby ensuring that all parking spaces can be captured video. Furthermore, this optional continuous parking space video capture equipment deployment scheme is flexible in deployment and ensures high accuracy in identifying vehicles within the parking spaces.

[0084] The following is combined Figures 2 to 4 The deployment method of the continuous berth video acquisition equipment provided in this application will be illustrated with a specific example.

[0085] Roadside parking is increasingly being incorporated into urban parking management in various cities. This specific example first analyzes the factors affecting video image quality and algorithm analysis accuracy; secondly, it delves into the characteristics and actual impact coefficients of each factor to lay the foundation for constructing a design scheme model; thirdly, through linear programming fitting of a large amount of experimental data, it derives the optimal range of camera imaging quality and algorithm recognition rate under different environments; finally, combining the actual pole height, field of view, and parking space size, it calculates the continuous parking spaces within the effective video acquisition range, ultimately outputting a deployment scheme for continuous parking space video acquisition equipment. The overall process is as follows:

[0086] S1: Research the factors that affect license plate clarity and algorithm recognition rate.

[0087] Factors affecting license plate clarity and algorithm recognition rate include, but are not limited to, lens focal length and lens parameters. Once these factors are determined, the selection of the bullet camera can be determined.

[0088] This example utilizes the theory of greedy algorithms. First, a mathematical model is established to describe the problem. The problem is divided into multiple subproblems. Each subproblem is solved separately to obtain a locally optimal solution. Finally, all locally optimal solutions are merged into the global answer.

[0089] Here we first introduce two key concepts: depth of field (DOF) and field of view (FOV).

[0090] DOF (Depth of Field): Depth of field determines the sharpness of video imaging. In this solution, the vehicle is moving in real time, and the focal point is constantly changing. Therefore, a clear image can be output within a certain depth of field range for algorithm analysis. Depth of field consists of two segments: foreground depth of field and background depth of field. The calculation formula is as follows:

[0091] Foreground depth = F*δ*L² / (f² + F*δ*L)

[0092] Depth of field = F*δ*L² / (f² - F*δ*L)

[0093] Formula 1: Depth of field = Foreground depth of field + Background depth of field = 2f² * F * δ * L² / (f⁴ - F² * δ² * L²)

[0094] Where δ is the diameter of the circle of confusion, f is the lens focal length, F is the lens aperture value, and L is the focusing distance.

[0095] FOV (Field of View): The field of view determines the range of the video image. FOV is the angle between the two edges of the maximum area through which the image of the target object can pass through the lens. FOV determines the camera's field of view; a larger FOV means a wider field of view. FOV can be measured using three data indicators: horizontal angle (from left to right of the image), vertical angle (from top to bottom of the image), and diagonal angle (from one corner to the opposite corner of the image). Each field of view (α) can be calculated from the selected size (d) and the effective focal length (f). It is related to the CCD sensor size and lens focal length. This solution requires calculating the vertical field of view, and the calculation formula is as follows:

[0096] Formula 2: FOV = 2arctan(d / 2f)

[0097] Where d is the height of the CCD (charge-coupled device) and f is the focal length of the lens.

[0098] S2: Calculate the pixel performance of license plates under different factors, including common focal lengths and lens selections.

[0099] In this step, after identifying the factors affecting license plate clarity and algorithm recognition rate, the required lens specifications and type are confirmed, i.e., the target model of bullet camera is determined.

[0100] Lenses can be categorized into three types based on their structure: First, fixed-aperture, manually adjustable, and automatically adjustable aperture prime lenses; second, manually adjustable and automatically adjustable aperture zoom lenses; and third, motorized triple-variable lenses. Lenses can also be categorized into six types based on their focal length: standard lenses, wide-angle lenses, fisheye lenses, telephoto lenses, reflective telephoto lenses, and macro lenses. Different lens structures and focal lengths are used in different environments depending on actual needs and cost requirements. This solution selects a mature and stable infrared zoom camera with automatic aperture adjustment, widely used in the market, and a lens with a 1 / 1.8" sensor area and an f / 1.4 aperture, with a standard focal length range of 8-56mm.

[0101] Furthermore, from the perspective of image clarity, the number of pixels a license plate has after being imaged by a camera at a fixed distance depends on the camera's resolution. Higher resolution means a larger number of pixels for the same license plate area, resulting in greater clarity. However, higher resolution also requires more computation for license plate recognition, leading to longer processing times and higher chip processing power consumption. This significantly reduces the algorithm's concurrent computational capabilities when handling multiple vehicles, potentially causing vehicles to be missed. Considering both cost considerations for large-scale implementation, a 4-megapixel resolution camera was chosen for this example.

[0102] The focal length of a lens refers to the measurement of how well light converges or diverges across the beam in an optical system. In camera applications, it refers to the distance from the center of the lens to the imaging plane, such as the film or CCD. The focal length determines the size of the image of the subject on the CCD. The longer the focal length, the larger the image; the shorter the focal length, the smaller the image. When studying the effective imaging distance of 8-56mm focal lengths, it is only necessary to break it down into studying the furthest imaging range of a 56mm focal length and the closest imaging range of an 8mm focal length. By combining the conclusions of these two subsets, a global conclusion can be derived.

[0103] Specifically, this step calculates the pixel representation of the license plate under common focal lengths, lens selections, and various factors, thus determining the number of pixels the license plate occupies within the frame corresponding to the distance between different license plates and the bullet camera. First, the license plate size must be determined. According to national standards, the size of a small car license plate is 440 mm * 140 mm. Common license plates typically consist of 7 characters: 1 Chinese character and 6 letters or numbers. The spacing between the second and third characters is half a character's width. Each character is 60 mm wide and 120 mm high.

[0104] According to the formula: object width W = 2 * D * tan(FOV / 2), and combining the FOV formula FOV = 2arctan(d / 2f), substituting the width of the license plate (440mm) and the distance D between the license plate and the camera as variables, we can obtain the width W after imaging. Then, according to the formula: pixel width W = pixel size (px) * number of pixels, substituting the pixel size of 1 / 1.8', 4mp (2.9px), we can finally obtain the number of pixels occupied by the license plate.

[0105] By substituting different distances D into the formula, the number of pixels occupied by the license plate corresponding to different distances D (the distance between the license plate and the bullet camera) can be obtained.

[0106] S3: By importing a training set of license plates with different pixel values, output statistical data on the algorithm's recognition rate.

[0107] This step utilizes data clustering analysis to group a collection of physical or abstract objects into multiple classes composed of similar objects. In the following tests, the labeling information of the training samples is unknown. The goal is then to analyze the unlabeled training samples to ultimately obtain the inherent patterns in the data, providing a foundation for further data analysis. In this scheme, 1000 license plate samples containing various features are randomly obtained from the sample material library. These 1000 samples are then processed into discrete samples corresponding to the length pixels of the license plate, packaged separately, and imported into the algorithm recognition terminal in batches to obtain license plate recognition accuracy data for different pixel occupancy numbers. The specific data obtained are shown in Table 1.

[0108] S4: The farthest license plate recognition distance for license plate imaging is determined by depth of field and recognition rate analysis.

[0109] Using the data in Table 1, a graph was constructed with the horizontal axis representing the number of pixels occupied and the vertical axis representing the recognition accuracy. Analysis of the graph's trend shows that when the license plate length reaches 100 pixels, the recognition rate reaches 95%. Furthermore, the improvement in recognition rate with increasing pixel count tends to plateau. Since the sample material already includes features such as blurriness and occlusion that prevent license plate information recognition, a 95% recognition rate is sufficient for application requirements. A 95% recognition rate corresponds to a maximum effective license plate recognition distance of approximately 70 meters for the bullet camera in the table.

[0110] The calculated 70 meters is a straight-line distance. In actual implementation, this straight-line distance needs to be converted into a perpendicular parking distance. In my country, the height of poles and traffic signs used in urban roads is generally 6-8 meters, and the length of the horizontal arm is also 6-8 meters. Therefore, this scheme also uses a height and horizontal arm of 8 meters as the spatial location for the camera. According to the spatial solid geometry calculation formula, when the diagonal of a cuboid is 70 meters and the two horizontal sides are 8 meters, the third horizontal side is approximately 68 meters. Combining this with the requirements of the national standard "Specifications for Setting Up Parking Spaces on Urban Roads," the length of parking spaces for small cars is between 6 meters and 6.5 meters; using the maximum of 6.5 meters, the conversion yields a maximum license plate recognition distance of 68 meters for the camera's perpendicular parking space, equivalent to... Figure 2 The 10th berth in the series.

[0111] S5: Calculate the closest and furthest video acquisition distances of the bullet camera at its maximum focal length based on the field of view formula and the pole height.

[0112] The field of view (FOV) and actual range of an 8-56mm focal length are calculated using formulas. At a standard F1.4 lens aperture, substituting the 8-56mm focal length range yields approximately the following values: horizontal FOV: 41°–10°, vertical FOV: 23°–5°, diagonal FOV: 48°–11°. The vertical FOV is a key parameter for analyzing the actual visual range distance. When the furthest FOV distance at a 56mm focal length is within 68 meters, the near-end boundary is calculated to be 36 meters based on the vertical FOV. The final video capture range for a 56mm focal length is... Figure 2 Berths 7 through 10.

[0113] In practical implementation, the installation height of the bullet camera also needs to be considered. Since license plates are perpendicular to the road surface, when the license plate is too close to the bullet camera and not on the same horizontal plane, distortion of the license plate due to the imaging angle will occur. By importing license plate samples with different distorted angles into the algorithm terminal for recognition and statistical analysis, the results show that the license plate recognition rate is below 95% when the angle is less than 25 degrees. Using the Pythagorean theorem to calculate tan25°, when the height of one right-angled side is 6 meters, the other right-angled side is approximately 12 meters. Introducing the FOV calculation formula, at a 56mm focal length and with the closest video acquisition distance adjusted to 12 meters, the farthest video acquisition distance is 39 meters. Substituting this into the parking space dimensions, the final video acquisition range for the 56mm focal length is... Figure 2 Berths 3 through 6.

[0114] S6: Based on the actual physical dimensions of the parking spaces, calculate the continuous parking spaces within the video capture range of the bullet camera.

[0115] like Figure 2 As shown, combining the conclusions from the farthest and nearest subsets above, it is concluded that an infrared zoom camera with a standard focal length of 8-56mm can capture video from any four consecutive parking spaces between 12 meters and 60 meters in outdoor environments. Additionally, the parking location is the location of the road traffic pole.

[0116] S7: Deployment plan for video acquisition equipment for continuous berths.

[0117] By combining the conclusions obtained from S5 and S6 with practical applications, the design problems of continuous berths can be solved by deploying 8-56mm bullet cameras in single points or in clusters.

[0118] Before deployment, introduce the following variables: the closest video acquisition distance of the bullet camera is a, the farthest video acquisition distance of the bullet camera is b, and the installation height of the bullet camera is 6 meters.

[0119] When a < 12 meters, due to license plate distortion, there are many license plate recognition errors, making video capture impossible. Figure 3 Berths 09 and 10 could not be captured via video from point 1.

[0120] When a ≥ 12 meters and b ≤ 35 meters, a single bullet-shaped camera can be designed to capture video from four parking spaces. Figure 3 Berth 05-08.

[0121] When a ≥ 35 meters and b ≤ 60 meters, a single bullet camera can be designed to capture video from four parking spaces. Figure 3 Berths 01-04.

[0122] When b > 60 meters, video capture is not possible due to insufficient license plate pixels.

[0123] Based on the above conclusions, the final deployment scheme for the bullet camera is shown in the diagram below. Figure 3 As shown. When video capture is impossible in the blind spot of a gun-type camera lens, the following complementary solutions can be used to address this issue. This allows for optimized output planning and design for scenarios with 24 consecutive berths. Figure 3 The video of berths 09-12 was captured by a gun-type camera at point 2, while the video of berths 13-16 was captured by a camera at point 1.

[0124] Figure 4 This is a schematic diagram of the structure of a continuous berth video acquisition system according to an embodiment of this application, as shown below. Figure 4 As shown, after deploying bullet cameras at each parking location (corresponding to the aforementioned road traffic poles), each bullet camera can send the captured data to a switch via a connected router. The switch then uploads the data to the cloud platform via the internet. It should be noted that the algorithm terminal can be a standalone device or integrated into other devices; no limitations are imposed here.

[0125] This example demonstrates a deployment scheme for continuous parking space video acquisition equipment. Utilizing high-position video equipment, algorithm terminals, and planning methods, it enables rapid deployment in various roadside parking scenarios and complex road conditions. Based on influencing factors such as the angle between the parking space and the roadway, the number of consecutive parking spaces, the size parameters of the parking spaces, the focal length and field of view parameters of the bullet camera, and the effective supplementary lighting distance, a method is used to calculate and analyze the installation position and height of the high-position video equipment. Finally, the camera video footage is captured, and artificial intelligence is used to analyze the real-time trajectories and license plates of passing vehicles to generate parking event data. Ultimately, this enables intelligent and refined management of the entire roadside parking process.

[0126] The continuous berth video acquisition equipment deployment method provided in this example only requires configuring the focal length per unit to simultaneously capture the real-time status of multiple vehicles at multiple berths in a single frame, without losing vehicle entry and exit data. This effectively solves the concurrency problem caused by the rotation and zoom of PTZ cameras when the status of multiple vehicles changes simultaneously. Furthermore, the bullet camera's image output is stable and less affected by external environmental factors; shaking and changes in ambient brightness have a lower impact on the bullet camera's image output quality. Moreover, the bullet camera's supplementary light distance is matched to its focal length parameters, meeting the brightness requirements for license plate recognition at night outdoors, thus ensuring recognition accuracy. From a deployment cost perspective, the high-position video capture ratio can reach 1 channel for 4 berths, far lower than the deployment cost of 1 channel for 1 berth for low-position video.

[0127] In this embodiment, a network device is provided, including a memory and a processor. The memory stores a computer program, and the processor is configured to run the computer program to perform the methods described in the above embodiments.

[0128] The aforementioned program can run on a processor or be stored in memory (or computer-readable medium). Computer-readable medium includes both permanent and non-permanent, removable and non-removable media, and information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable medium does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0129] These computer programs may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes can be implemented using different modules, and different steps can be implemented using different modules.

[0130] This embodiment also provides an apparatus. This apparatus, referred to as a continuous berth video acquisition equipment deployment apparatus, includes:

[0131] The first acquisition module is used to acquire the farthest license plate recognition distance of the bullet camera;

[0132] The second acquisition module is used to acquire the nearest and farthest video acquisition distances of the bullet camera at its maximum focal length.

[0133] The first determining module is used to determine the video acquisition range of the bullet camera based on the farthest license plate recognition distance, the nearest video acquisition distance, and the farthest video acquisition distance.

[0134] The first deployment module is used to calculate the continuous parking spaces within the video capture range of the bullet camera based on the predetermined parking space size, and to deploy the bullet camera at the road traffic pole for the continuous parking spaces.

[0135] The above embodiments solve the problems of high deployment cost and poor accuracy of berth video acquisition results in related technologies. The continuous berth video acquisition equipment deployment scheme provided by the above embodiments has the following advantages: First, by using the imaging effect of the physical berth in the lens of the bullet camera and the recognition effect of artificial intelligence, the number of berths effectively covered in the image can be obtained, which can improve the accuracy of vehicle recognition results on the effectively covered berths. Second, one bullet camera can collect video from multiple continuous berths, which can save manpower and financial costs compared to the existing technology of setting a low-level video stake for each berth. Third, since the bullet camera is deployed above the road traffic poles, it can effectively avoid external human-caused obstruction and damage.

[0136] The system or apparatus is used to implement the functions of the methods in the above embodiments. Each module in the system or apparatus corresponds to each step in the method, as has been described in the method and will not be repeated here.

[0137] Optionally, the first acquisition module includes an acquisition unit, used to acquire the farthest license plate recognition distance corresponding to a preset license plate recognition accuracy within the depth of field range of the bullet camera.

[0138] Optionally, the acquisition unit includes: a first processing subunit, used to determine the number of pixels in the frame area occupied by license plates at different distances from the bullet camera based on the field of view, pixel size, and license plate size of the bullet camera; a second processing subunit, used to determine the correspondence between the number of pixels in the frame area occupied by each license plate and the license plate recognition accuracy; and a third processing subunit, used to filter based on the correspondence and a preset license plate recognition accuracy to obtain the farthest license plate recognition distance of the bullet camera.

[0139] Optionally, the first processing subunit includes: a first calculation subunit, used to calculate the size of each license plate after imaging by the bullet camera based on the field of view of the bullet camera, the size of the license plate, and the distance from each license plate to the bullet camera; and a second calculation subunit, used to calculate the number of pixels within the frame area occupied by each license plate based on the size of each license plate after imaging by the bullet camera and the pixel size of the bullet camera.

[0140] Optionally, the second processing subunit includes: a recognition subunit, used to recognize the license plate content within the frame area occupied by each license plate, and obtain the license plate content recognition result of each license plate; an acquisition subunit, used to obtain the license plate recognition accuracy of each license plate based on the license plate content recognition result of each license plate; and an establishment subunit, used to establish the correspondence between the number of pixels within the frame area occupied by each license plate and the license plate recognition accuracy.

[0141] Optionally, the third processing subunit includes: a construction subunit, used to construct a trend of license plate pixel count changing with license plate recognition accuracy based on each correspondence; a first determination subunit, used to compare the trend of change with a preset license plate recognition accuracy to determine the farthest straight-line distance for license plate recognition by the bullet camera; a second determination subunit, used to determine the pole height and crossarm size for mounting the bullet camera based on the height of road traffic poles and traffic signs; and a third determination subunit, used to determine the farthest license plate recognition distance of the bullet camera based on the pole height, crossarm size, and the farthest straight-line distance for license plate recognition.

[0142] Optionally, the second acquisition module includes a calculation unit, used to calculate the nearest and farthest video acquisition distances of the gun-type camera based on the field of view of the gun-type camera and the height of the pole used to mount the gun-type camera.

[0143] Optionally, the above calculation unit includes: a third calculation subunit for calculating the vertical field of view range of the bullet camera in the standard focal length range; and a fourth calculation subunit for calculating the nearest and farthest video acquisition distance of the bullet camera based on the height of the pole on which the bullet camera is installed and the vertical field of view range.

[0144] Optionally, the aforementioned bullet camera is an infrared zoom bullet camera, the standard focal length range of the lens of the infrared zoom bullet camera is 8-56mm, and the resolution of the lens is not less than 4 million pixels.

[0145] Optionally, the above device further includes: a second determining module, configured to, after calculating the continuous parking spaces within the video acquisition range of the bullet camera according to the predetermined parking space size, and deploying the bullet camera at the road traffic pole for the continuous parking spaces, determine the remaining parking spaces to be video acquired when the number of parking spaces to be video acquired exceeds the number of continuous parking spaces to be video acquired, wherein the remaining parking spaces to be video acquired are parking spaces to be video acquired other than the continuous parking spaces to be video acquired; and a second deployment module, configured to adjust and deploy at least one bullet camera at at least one road traffic pole for the remaining parking spaces to be video acquired.

[0146] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for deploying video acquisition equipment for continuous berths, comprising: Obtain the furthest license plate recognition distance of the bullet camera; Obtain the nearest and farthest video acquisition distances of the bullet-shaped camera at its maximum focal length; The video capture range of the bullet camera is determined based on the farthest license plate recognition distance, the nearest video capture distance, and the farthest video capture distance. The bullet camera calculates the continuous parking spaces within the video capture range based on the predetermined parking space size, and deploys the bullet camera at the road traffic pole for the continuous parking spaces. The process of obtaining the furthest license plate recognition distance of the bullet camera includes: Based on the field of view, pixel size, and license plate size of the bullet camera, determine the number of pixels within the image area occupied by license plates at different distances from the bullet camera; Determine the correspondence between the number of pixels within the image area occupied by each license plate and the license plate recognition accuracy; Based on the aforementioned correspondences and the preset license plate recognition accuracy, the farthest license plate recognition distance of the bullet camera is obtained through screening.

2. The method according to claim 1, characterized in that, The step of determining the number of pixels within the frame area occupied by license plates at different distances from the camera, based on the field of view, pixel size, and license plate size of the bullet camera, includes: The size of each license plate after imaging by the gun-type camera is calculated based on the field of view of the gun-type camera, the size of the license plate, and the distance from each license plate to the gun-type camera. Based on the size of each license plate after imaging by the bullet camera and the pixel size of the bullet camera, the number of pixels within the image area occupied by each license plate is calculated.

3. The method according to claim 1, characterized in that, The step of determining the correspondence between the number of pixels within the image area occupied by each license plate and the license plate recognition accuracy includes: The content of the license plate within the area occupied by each license plate is identified to obtain the license plate content recognition result of each license plate; Based on the license plate content recognition results of each license plate, the license plate recognition accuracy of each license plate is obtained; Establish the correspondence between the number of pixels within the image area occupied by each license plate and the license plate recognition accuracy.

4. The method according to claim 1, characterized in that, The step of obtaining the farthest license plate recognition distance of the bullet camera by filtering based on the aforementioned correspondences and the preset license plate recognition accuracy includes: Based on the aforementioned correspondences, a trend is constructed regarding the change in the number of license plate pixels as a function of license plate recognition accuracy. The trend of the change is compared with the preset license plate recognition accuracy to determine the farthest straight-line distance for license plate recognition by the bullet camera; Based on the height of road traffic poles and traffic signs, determine the pole height and crossarm dimensions for mounting the bullet camera; The farthest license plate recognition distance of the bullet camera is determined based on the pole height, the crossarm size, and the farthest license plate recognition straight-line distance.

5. The method according to claim 1, characterized in that, The steps for obtaining the nearest and farthest video acquisition distances of the bullet camera at its maximum focal length include: Based on the field of view of the gun-type camera and the height of the pole used to mount the gun-type camera, the nearest and farthest video acquisition distances of the gun-type camera are calculated.

6. The method according to claim 5, characterized in that, The steps for calculating the nearest and farthest video acquisition distances of the bullet camera based on its field of view and the height of the pole used to mount it include: Calculate the vertical field of view range of the bullet camera within the standard focal length range; Based on the height of the pole on which the bullet camera is installed and the vertical field of view range, the nearest and farthest video acquisition distances of the bullet camera are calculated.

7. The method according to claim 1, characterized in that: The gun-type camera is an infrared zoom gun-type camera, and the standard focal length range of the lens of the infrared zoom gun-type camera is 8-56mm, and the resolution of the lens is not less than 4 million pixels.

8. The method according to claim 1, characterized in that, After calculating the consecutive parking spaces within the video capture range of the bullet camera based on the predetermined parking space size, and deploying the bullet camera at the road traffic pole for the consecutive parking spaces, the method further includes: If the number of berths to be video-captured exceeds the number of consecutive berths for video capture, the remaining berths to be video-captured are determined, wherein the remaining berths to be video-captured are berths to be video-captured other than the consecutive berths for video capture. For the remaining parking spaces to be video-captured, at least one of the bullet-shaped cameras is deployed at at least one of the road traffic poles.

9. A network device, comprising a memory and a processor; wherein, The memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the steps of the continuous berth video acquisition device deployment method according to any one of claims 1 to 8.

10. A readable storage medium having computer instructions stored thereon, wherein, When executed by a processor, the computer instructions implement the steps of the continuous berth video acquisition device deployment method according to any one of claims 1 to 8.

Images