A monitoring point management method and device
By filtering the target image in the video frame sequence, attribute evaluation is carried out for key parts of the monitoring point, and using image semantic description information to determine the quality level, solving the problem that the existing technology cannot fully evaluate the installation quality of the monitoring point, achieving high Accurate monitoring point control and early warning functions.
Patent Information
- Application Number
- CN202110865338.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-07-29
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2041-07-29
AI Technical Summary
The prior art cannot comprehensively evaluate the impact of the installation angle, height and position of video surveillance points on the quality of video analysis. High-precision monitoring point quality analysis cannot be achieved through quality diagnosis of image dimensions alone.
By obtaining video frame sequences, filtering target images with image quality meet the requirements, performing attribute evaluation for key parts of the evaluation target, using image semantic description information to determine the quality level, and determining whether manual verification is required based on the quality level in multiple time periods .
It realizes a comprehensive and accurate assessment of the quality of monitoring points, can conduct comprehensive evaluation in the installation and deployment dimension, provide early warning, and improves the accuracy of monitoring point governance and automated analysis capabilities.
Smart Images

Figure CN113674224B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of video image analysis, and in particular to a monitoring point management method and device. Background Art
[0002] With the rapid development of smart cities and artificial intelligence, the penetration rate of video surveillance points in cities, counties or enterprises is increasing. Whether the installation and deployment of video surveillance points is reasonable plays a decisive role in the quality of video acquisition.
[0003] The existing technology often uses the original information of video images to diagnose global image quality anomalies such as clarity, color cast, noise, blur, jitter, brightness, occlusion, etc. of video scene images through traditional machine learning feature extraction, pattern classification and image processing methods.
[0004] However, the inventors of this application have found in their long-term research and development that the existing technology can only detect abnormalities in the image dimension of monitoring points. Since the installation angle, height, and reasonableness of the position of the monitoring points are key factors that directly affect the quality of video analysis, quality diagnosis using only the image dimension cannot fully describe the quality of video monitoring points. In view of this, how to quickly and effectively perform high-precision automated analysis and diagnosis of the quality of video points has become an urgent problem to be solved. Summary of the invention
[0005] The main technical problem solved by the present application is to provide a monitoring point management method and device, which can effectively improve the accuracy of monitoring point management.
[0006] In order to solve the above technical problems, a technical solution adopted in the present application is: to provide a monitoring point management method, including: obtaining a video frame sequence collected by the monitoring point within a preset time period, and screening out a target image whose image quality meets the set image quality requirements from the video frame sequence; wherein the target image contains an evaluation target; performing attribute evaluation on at least one key part of the evaluation target in the target image to obtain a corresponding attribute evaluation result, wherein the attribute evaluation result is related to the image semantic description information of the corresponding key part; determining the quality level of the evaluation target using the attribute evaluation result; and judging whether it is necessary to perform manual verification of the monitoring point based on the quality levels of the evaluation targets within multiple preset time periods.
[0007] Among them, the step of performing attribute evaluation on at least one key part of the evaluation target in the target image to obtain a corresponding attribute evaluation result includes: obtaining all the key parts to be evaluated based on the type of the evaluation target, and obtaining the corresponding image semantic description information for each key part; judging whether all the image semantic description information is greater than or equal to the corresponding preset threshold; if so, determining that the attribute evaluation result of the evaluation target is the first quality target; otherwise, determining that the attribute evaluation result of the evaluation target is the second quality target.
[0008] Among them, the step of obtaining all the key parts to be evaluated based on the type of the evaluation target, and obtaining the corresponding image semantic description information for each key part includes: in response to the type of the evaluation target being a human target, obtaining the key parts to be evaluated including at least one of the torso, head and limbs, and obtaining the corresponding image semantic description information for at least one of the torso, head and limbs, wherein the image semantic description information includes human image size, human image completeness, human image clarity, human image detection confidence and the proportion of human images with associated faces; and / or, in response to the type of the evaluation target being a face target, obtaining the key parts to be evaluated including facial features, and obtaining the corresponding image semantic description information for the facial features, wherein the image semantic description information includes facial image size, facial image completeness, human image clarity, human image detection confidence and the proportion of human images with associated faces. Face image clarity, face angle, face image detection confidence and whether a mask is worn; and / or, in response to the type of the evaluation target being a motor vehicle target, the key parts to be evaluated include at least one of the vehicle logo, headlights, windows and license plates, and for at least one of the vehicle logo, headlights, windows and license plates, the corresponding image semantic description information is obtained, wherein the image semantic description information includes the motor vehicle image size, the motor vehicle image completeness, the motor vehicle image clarity, the motor vehicle image detection confidence and whether there is an associated license plate; and / or, in response to the type of the evaluation target being a license plate target, the key parts to be evaluated include printed characters, and for the printed characters, the corresponding image semantic description information is obtained, wherein the image semantic description information includes the license plate image size, the license plate image completeness, the license plate image clarity and the license plate image detection confidence.
[0009] Among them, the step of using the attribute evaluation results to determine the quality level of the evaluation target includes: obtaining the total number of target images of the evaluation target within a preset time period; obtaining the number of the evaluation targets whose attribute evaluation results are second quality targets within a preset time period; and determining the quality level of the evaluation target based on the ratio of the number of the second quality targets to the total number of target images.
[0010] Among them, the step of determining the quality level of the evaluation target based on the ratio of the number of the second quality targets to the total number of target images includes: comparing the ratio with the first threshold and the second threshold; if the ratio is less than the first threshold, then determining that the quality level of the evaluation target is a high quality level; if the ratio is greater than or equal to the first threshold and less than the second threshold, then determining that the quality level of the evaluation target is a medium quality level; if the ratio is greater than or equal to the second threshold, then determining that the quality level of the evaluation target is a low quality level, wherein the first threshold is less than the second threshold.
[0011] Among them, the step of judging whether it is necessary to manually check the monitoring point based on the quality level of the evaluation target within multiple preset time periods includes: judging whether the quality level of the evaluation target within multiple preset time periods is reduced; if so, triggering an early warning and manually checking the monitoring point; otherwise, obtaining the video frame sequence of the monitoring point in the next preset time period.
[0012] Among them, the types of evaluation targets include at least two of human target type, face target type, motor vehicle target type, and license plate target type; the step of judging whether the quality level of the evaluation targets within the multiple preset times is reduced includes: judging whether the quality level of any one of the evaluation targets within the multiple preset times is reduced.
[0013] Among them, the step of obtaining a video frame sequence collected by the monitoring point within a preset time period, and filtering out a target image whose image quality meets the set image quality requirements from the video frame sequence, includes: performing target tracking on each video frame in the video frame sequence to obtain the video frame where the evaluation target is located and its coordinate information in the current video frame; using the coordinate information of all evaluation targets of the evaluation target in the current video frame to obtain the corresponding image quality score; and using the video frame with the highest image quality score as the target image.
[0014] To solve the above technical problems, another technical solution adopted in the present application is: to provide a monitoring point management device, comprising a memory and a processor coupled to each other, the memory storing program instructions, and the program instructions are used to be executed by the processor to implement the monitoring point management method mentioned in any of the above embodiments.
[0015] In order to solve the above technical problems, another technical solution adopted in the present application is: providing a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is used to implement the monitoring point management method mentioned in any of the above embodiments.
[0016] Different from the prior art, the beneficial effects of the present application are as follows: a monitoring point management method is provided in the present application, which makes full use of the image semantic description information of key parts of the evaluation target to evaluate the attributes of the target image, thereby determining the quality level of the evaluation target type, and judging whether it is necessary to manually check the current monitoring point based on the quality level in different time periods; wherein, the image semantic description information can not only evaluate the quality-related dimensions of the target image, but also comprehensively evaluate the installation and deployment dimensions of the monitoring point, such as the installation height, angle, and scene rationality of the monitoring point, comprehensively and accurately feedback the quality of the monitoring point and issue early warnings for abnormal situations, thereby achieving more accurate monitoring point management. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:
[0018] Figure 1 It is a flowchart of an implementation method of the monitoring point management method of the present application;
[0019] Figure 2 yes Figure 1 A schematic flow chart of an implementation method of step S101;
[0020] Figure 3 yes Figure 1 A schematic flow chart of an implementation method of step S102;
[0021] Figure 4 yes Figure 1 A schematic flow chart of an implementation method of step S103;
[0022] Figure 5 yes Figure 1 A structural diagram of an implementation method of step S104;
[0023] Figure 6 It is a schematic diagram of the framework of an implementation method of the monitoring point management device of the present application;
[0024] Figure 7 This is a structural diagram of an implementation method of the monitoring point management device of the present application;
[0025] Figure 8 It is a schematic diagram of a framework of an implementation scheme of a computer-readable storage medium of the present application. DETAILED DESCRIPTION
[0026] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0027] See also Figure 1 , Figure 1 This is a flow chart of an implementation method of the monitoring point management method of the present application. Specifically, it may include the following steps:
[0028] S101: Obtain a video frame sequence collected by a monitoring point within a preset time period, and select a target image whose image quality meets a set image quality requirement from the video frame sequence; wherein the target image includes an evaluation target.
[0029] Optionally, the front-end equipment at the monitoring point is used to collect and acquire monitoring video within a preset time range, and the monitoring video is framed to obtain a video frame sequence. In this embodiment, the frame rate of the video frame is set to 8-25 frames per second, and the preset time is set to 24 hours. The target detection and tracking algorithm is used to extract all frame images containing the evaluation target from the video frame sequence, and then each evaluation target in each frame image is subjected to image quality analysis. Finally, when the evaluation target disappears in the video frame sequence, a frame image whose image quality meets the preset image quality requirements in the life cycle of the scoring target is selected and used as a frame target image with the best image quality. It should be noted that the image quality here is for each evaluation target, not for each frame image.
[0030] In a specific implementation scenario, please refer to Figure 2 , Figure 2 yes Figure 1 The above step S101 specifically includes the following steps:
[0031] S201: Tracking a target on each video frame in a video frame sequence to obtain the video frame where the evaluation target is located and its coordinate information in the current video frame.
[0032] Optionally, a target tracking algorithm is used to obtain the coordinates of the evaluation target on each video frame image, and the coordinates of the evaluation target between the previous and next frame images are temporally and spatially associated; at the same time, a unique ID identifier is generated for each evaluation target, and the associated coordinate information is matched one-to-one with the ID identifier.
[0033] S202: Obtain corresponding image quality scores using coordinate information of all evaluation targets in the current video frame.
[0034] Optionally, the target scoring capture algorithm is used to queue and cache the coordinate information of the evaluation target and the corresponding ID identification, and the historical frame trajectory of each ID target is maintained. Based on the coordinate information of all evaluation targets in each video frame image, the size of the current evaluation target to be scored, whether the position is blocked, etc., is mainly concerned, and each evaluation target in each frame image is captured and scored.
[0035] S203: Taking the video frame with the highest image quality score as the target image.
[0036] Optionally, when the evaluation target disappears in the video frame sequence, a frame with the highest score is selected from the video frame image period of the evaluation target and is used as the target image.
[0037] Through the above implementation, an image frame with the best image quality of the evaluation target in the video frame sequence can be obtained, which can be used as an evaluation basis in the subsequent attribute evaluation process, and can reflect the highest level of the evaluation target obtained at the current monitoring point.
[0038] S102: Performing attribute evaluation on at least one key part of the evaluation target in the target image to obtain a corresponding attribute evaluation result, wherein the attribute evaluation result is related to image semantic description information of the corresponding key part.
[0039] Specifically, the multi-task multi-label fine-grained classification and recognition technology based on deep learning can be used to perform attribute evaluation on the key parts of the evaluation target. The specific process refers to the existing technology, and the technical principles of the classification and recognition technology will not be elaborated here.
[0040] In a specific implementation scenario, please refer to Figure 3 , Figure 3 yes Figure 1 The above step S102 may include the following steps:
[0041] S301: Obtain all key parts to be evaluated based on the type of the evaluation target, and obtain corresponding image semantic description information for each key part.
[0042] Optionally, when the monitoring point is applied to different scenarios, the type of evaluation target can be adjusted according to the change of the application scenario. Among them, the type of evaluation target can be but not limited to any one or more of human targets, face targets, motor vehicle targets, and license plate targets. For example, for vehicle monitoring points set up in motor vehicle lanes, it is necessary to pay attention to the quality of motor vehicle and license plate targets at the same time, and the types of evaluation targets at this time include motor vehicle targets and license plate targets. For another example, for portrait monitoring points set up at the entrance and exit of shopping malls, it is mainly necessary to pay attention to the quality of face and human targets, and the types of evaluation targets at this time include human targets and face targets. It should be noted that although the face target itself belongs to the human target, since the target images with the highest overall quality scores corresponding to the face target and the human target in the above-mentioned acquired video frame may be different, and there are obvious differences between the two in the quality evaluation dimension, the attributes of the face target and the human target are evaluated separately; similarly, for motor vehicle targets and license plate targets, the two have different focuses in the quality evaluation dimension, so the attributes need to be evaluated separately.
[0043] For different types of evaluation targets, the key parts to be evaluated and the image semantic description information corresponding to each type are different. In a specific implementation scenario, corresponding to different types of evaluation targets, the above step S301 includes one or any combination of the following:
[0044] In response to the type of the evaluation target being a human target, obtaining the key parts to be evaluated including at least one of a torso, a head, and limbs, and obtaining corresponding image semantic description information for at least one of the torso, the head, and the limbs, wherein the image semantic description information may include but is not limited to at least one of the following information: human image size, human image completeness, human image clarity, human image detection confidence, and a human image ratio with an associated face;
[0045] In response to the type of the evaluation target being a face target, obtaining key parts to be evaluated including facial features, and obtaining corresponding image semantic description information for the facial features, wherein the image semantic description information may include but is not limited to at least one of facial image size, facial image completeness, facial image clarity, facial angle, facial image detection confidence, and whether a mask is worn;
[0046] In response to the type of the evaluation target being a motor vehicle target, obtaining key parts to be evaluated including at least one of a vehicle logo, a vehicle light, a vehicle window, and a vehicle license plate, and obtaining corresponding image semantic description information for at least one of the vehicle logo, the vehicle light, the vehicle window, and the vehicle license plate, wherein the image semantic description information may include but is not limited to at least one of the following information: a vehicle image size, a vehicle image integrity, a vehicle image clarity, a vehicle image detection confidence, and whether there is an associated license plate;
[0047] In response to the type of the evaluation target being a license plate target, key parts to be evaluated are obtained including printed characters, and corresponding image semantic description information is obtained for the printed characters, wherein the image semantic description information may include but is not limited to at least one of license plate image size, license plate image completeness, license plate image clarity, and license plate image detection confidence.
[0048] Of course, in other implementations, the key parts to be evaluated corresponding to each evaluation target type may also include other key parts, and the image semantic description information obtained for each key part may also include other information, which is not specifically limited here.
[0049] It should be noted that the above-mentioned image semantic description information including image size, image integrity, clarity and detection confidence is obtained for each key part, among which the image size specifically refers to the width and height of the image. Since the image recognition algorithm has the ability limit, a small image resolution has a greater impact on the effect of the algorithm; in addition, the integrity and clarity of the target image determine the quality of the information of the key parts of the image, and will also affect the recognition accuracy; the detection confidence reflects the quality of the evaluation target from another perspective. A low confidence level indicates that there is less effective information about the target contour.
[0050] S302: Determine whether all image semantic description information is greater than or equal to a corresponding preset threshold.
[0051] Optionally, for different results of image semantic description information, the corresponding preset thresholds should also be set accordingly. For example, for the target image size in the image semantic description information, the preset threshold is set to the size of the image; for the detection confidence, the threshold is the preset confidence size; for the scoring results of the integrity and clarity of the target image, the preset threshold is also set to the score threshold; for the binary results of whether a mask is worn or whether there is an associated license plate, the preset threshold is also two, such as 1 for yes and 0 for no.
[0052] S303: If yes, determine that the attribute evaluation result of the evaluation target is the first quality target.
[0053] Optionally, if the image semantic description information results of all dimensions in the above-mentioned image semantic description information are greater than or equal to a preset threshold, the evaluation target is considered to be a first quality target, where the first refers to the attribute evaluation level of the evaluation target, indicating that the attribute evaluation level of the evaluation target is high, and the attribute evaluation result of the evaluation target is a high-quality target.
[0054] S304: Otherwise, determine that the attribute evaluation result of the evaluation target is the second quality target.
[0055] Optionally, in this embodiment, if the information result of any dimension in the above-mentioned image semantic description information is lower than a preset threshold, the evaluation target is considered to be a second quality target, wherein the second refers to the attribute evaluation level of the evaluation target, indicating that the attribute evaluation level of the evaluation target is low, and the attribute evaluation result of the evaluation target is a low-quality target.
[0056] Through the above implementation mode, targeted attribute evaluation dimensions are designed for different evaluation target types, and the content included in the acquired image semantic description information is adaptively adjusted according to the needs of the monitoring points, so that it can more accurately describe the quality of a specific type, effectively improving the accuracy of the attribute evaluation results.
[0057] S103: Determine the quality level of the evaluation target using the attribute evaluation result.
[0058] See also Figure 4 , Figure 4 yes Figure 1 The above step S103 includes the following steps:
[0059] S401: Obtain the total number of target images of the evaluation target within a preset time period.
[0060] Optionally, in this embodiment, taking the face target type as an example, the preset time is set to 24 hours, which means that after completing step S203, the number of target images containing face targets is counted from the 24-hour video frame sequence.
[0061] S402: Obtain the number of evaluation targets whose attribute evaluation results are the second quality target within a preset time period.
[0062] S403: Determine the quality level of the evaluation target based on the ratio of the number of second quality targets to the total number of target images.
[0063] Specifically, in this embodiment, the ratio is compared with the first threshold and the second threshold; if the ratio is less than the first threshold, the quality level of the evaluation target is determined to be a high quality level; if the ratio is greater than or equal to the first threshold and less than the second threshold, the quality level of the evaluation target is determined to be a medium quality level; if the ratio is greater than or equal to the second threshold, the quality level of the evaluation target is determined to be a low quality level, wherein the first threshold is less than the second threshold.
[0064] In another specific embodiment, the first threshold is preset to 20%, and the second threshold is preset to 40%. If the proportion of low-quality targets in all target images is less than 20%, the current evaluation target type is considered to be of high quality level; if the low-quality proportion is between 20%-40%, the current evaluation target type is considered to be of medium quality level; if the low-quality proportion is higher than 40%, the current evaluation target type is considered to be of low quality level.
[0065] Through the above implementation method, the quality level of the evaluation target is judged based on the attribute evaluation result, which effectively improves the credibility of the grade division and provides technical support for the judgment of whether manual verification is needed in the future.
[0066] S104: Determine whether manual verification of the monitoring point is required based on the quality levels of the evaluation targets within a plurality of preset time periods.
[0067] See also Figure 5 , Figure 5 yes Figure 1 The above step S104 includes:
[0068] S501: Determine whether the quality level of the evaluation target within a plurality of preset time periods is reduced.
[0069] In this embodiment, the single preset time is 24 hours, and it is necessary to compare two 24 hours, that is, within two days to evaluate whether the quality level of the target type is reduced. The reduction here means changing from the original high quality level to the medium quality level or the low quality level, or from the original medium quality level to the low quality level.
[0070] In a specific implementation scenario, for a specific monitoring point checkpoint, such as a portrait checkpoint or a motor vehicle checkpoint, it is necessary to pay attention to multiple types of evaluation targets at the same time, that is, the evaluation target types include at least two of the human target type, the face target type, the motor vehicle target type, and the license plate target type. Then, the rule for determining whether manual verification is required needs to be configured to determine whether the quality level of any evaluation target decreases within multiple preset time periods. Through the above implementation, the evaluation of professional portrait checkpoints or vehicle checkpoints is more targeted.
[0071] S502: If yes, trigger an early warning and perform manual verification on the monitoring points.
[0072] In this embodiment, the manual verification includes confirming whether the monitoring points have unreasonable installation and deployment such as height, angle, image parameters, etc., and confirming whether the monitoring points have been abnormally adjusted or moved, and making targeted adjustments in a timely manner. Through the above implementation, it is possible to promptly alarm points with significant abnormal changes in quality, with high timeliness, and minimize the labor cost of point verification.
[0073] S503: Otherwise, obtain a video frame sequence of the monitoring point within the next preset time period.
[0074] Through the above implementation mode, the image semantic description information of key parts in the evaluation target is fully utilized to evaluate the attributes of the target image, so as to determine the quality level of the evaluation target type, and judge whether it is necessary to manually check the current monitoring point based on the quality level in different time periods; among them, the image semantic description information can not only evaluate the quality-related dimensions of the target image, but also conduct a comprehensive evaluation of the installation and deployment dimensions of the monitoring point, such as the installation height, angle, and scene rationality of the monitoring point, comprehensively and accurately feedback the quality of the monitoring point and warn of abnormal situations, thereby achieving more accurate monitoring point management.
[0075] See also Figure 6 , Figure 6 It is a framework diagram of an implementation method of the monitoring point management device of the present application. The identification device 100 includes an acquisition module 10, an attribute evaluation module 12, a quality evaluation module 14 and a judgment module 16. Specifically, the acquisition module 10 is used to obtain a video frame sequence collected by the monitoring point within a preset time period, and filter out a target image whose image quality meets the set image quality requirements from the video frame sequence; wherein the target image contains an evaluation target. The attribute evaluation module 12 is used to perform attribute evaluation on at least one key part of the evaluation target in the target image to obtain a corresponding attribute evaluation result, wherein the attribute evaluation result is related to the image semantic description information of the corresponding key part. The quality evaluation module 14 is used to determine the quality level of the evaluation target using the attribute evaluation result. The judgment module 16 is used to determine whether it is necessary to manually check the monitoring point based on the quality level of the evaluation target within multiple preset time periods. Through the above method, the image semantic description information of key parts of the evaluation target is fully utilized to evaluate the attributes of the target image, so as to determine the quality level of the evaluation target type, and judge whether it is necessary to manually check the current monitoring point based on the quality level in different time periods; among them, the image semantic description information can not only evaluate the quality-related dimensions of the target image, but also conduct a comprehensive evaluation of the installation and deployment dimensions of the monitoring point, such as the installation height, angle, and scene rationality of the monitoring point, comprehensively and accurately feedback the quality of the monitoring point and warn of abnormal situations, thereby achieving more accurate monitoring point management.
[0076] See also Figure 7 , Figure 7 The device 1000 includes a memory 101 and a processor 102 coupled to each other, the memory 101 stores program instructions, and the processor 102 is used to execute the program instructions to implement the monitoring point management method mentioned in any of the above embodiments.
[0077] Specifically, the processor 102 may also be referred to as a CPU (Central Processing Unit). The processor 102 may be an integrated circuit chip having the ability to process signals. The processor 102 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. In addition, the processor 102 may be implemented by a plurality of integrated circuit chips.
[0078] See also Figure 8 , Figure 8 It is a schematic diagram of a framework of an embodiment of a computer-readable storage medium of the present application. The computer-readable storage medium 20 stores a computer program 200, which can be read by a computer, and the computer program 200 can be executed by a processor to implement the monitoring point management method mentioned in any of the above embodiments. Among them, the computer program 200 can be stored in the above-mentioned computer-readable storage medium 20 in the form of a software product, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (processor) to execute all or part of the steps of the method described in each embodiment of the present application. The computer-readable storage medium 20 with a storage function can be a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, RandomAccess Memory), a disk or an optical disk, etc., which can store program codes, or a terminal device such as a computer, a server, a mobile phone, and a tablet.
[0079] The above description is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.
Claims
1. A monitoring point management method, characterized in that: include: Obtaining a video frame sequence collected by the monitoring point within a preset time period, and screening out a target image whose image quality meets the set image quality requirements from the video frame sequence; wherein the target image includes an evaluation target; Performing attribute evaluation on at least one key part of the evaluation target in the target image to obtain a corresponding attribute evaluation result, wherein the attribute evaluation result is related to the image semantic description information of the corresponding key part; Determining the quality level of the evaluation target using the attribute evaluation result; Determining whether it is necessary to manually check the monitoring point based on the quality level of the evaluation target within the plurality of preset time periods; Among them, the attribute evaluation of at least one key part of the evaluation target in the target image is performed to obtain a corresponding attribute evaluation result, including: obtaining all the key parts to be evaluated based on the type of the evaluation target, and obtaining the corresponding image semantic description information for each key part; in response to all the image semantic description information being greater than or equal to the corresponding preset threshold, determining that the attribute evaluation result of the evaluation target is a first quality target; in response to any one of the image semantic description information being less than the corresponding preset threshold, determining that the attribute evaluation result of the evaluation target is a second quality target.
2. The monitoring point management method according to claim 1 is characterized in that: The step of obtaining all the key parts to be evaluated based on the type of the evaluation target, and obtaining the corresponding image semantic description information for each key part includes: In response to the type of the evaluation target being a human target, obtaining the key parts to be evaluated including at least one of a torso, a head, and limbs, and obtaining the corresponding image semantic description information for at least one of the torso, the head, and the limbs, wherein the image semantic description information includes a human image size, a human image completeness, a human image clarity, a human image detection confidence, and a human image ratio with an associated face; and / or, In response to the type of the evaluation target being a face target, the key parts to be evaluated include facial features, and for the facial features, the corresponding image semantic description information is obtained, wherein the image semantic description information includes face image size, face image completeness, face image clarity, face angle, face image detection confidence, and whether a mask is worn; and / or, In response to the type of the evaluation target being a motor vehicle target, obtaining the key parts to be evaluated including at least one of a vehicle logo, a headlight, a vehicle window, and a license plate, and obtaining the corresponding image semantic description information for at least one of the vehicle logo, the headlight, the vehicle window, and the license plate, wherein the image semantic description information includes the size of the motor vehicle image, the completeness of the motor vehicle image, the clarity of the motor vehicle image, the confidence of the motor vehicle image detection, and whether there is an associated license plate; and / or, In response to the type of the evaluation target being a license plate target, the key parts to be evaluated include printed characters, and for the printed characters, the corresponding image semantic description information is obtained, wherein the image semantic description information includes license plate image size, license plate image completeness, license plate image clarity and license plate image detection confidence.
3. The monitoring point management method according to claim 1 is characterized in that: The step of determining the quality level of the evaluation target by using the attribute evaluation result comprises: Acquire the total number of target images of the evaluation target within a preset time period; Obtaining the number of the attribute evaluation results of the evaluation targets within a preset time period that are the second quality target; The quality level of the evaluation target is determined based on the ratio of the number of the second quality targets to the total number of the target images.
4. The monitoring point management method according to claim 3 is characterized in that: The step of determining the quality level of the evaluation target based on the ratio of the number of the second quality targets to the total number of the target images comprises: Compare the ratio with the first threshold and the second threshold; if the ratio is less than the first threshold, determine that the quality level of the evaluation target is a high quality level; if the ratio is greater than or equal to the first threshold and less than the second threshold, determine that the quality level of the evaluation target is a medium quality level; if the ratio is greater than or equal to the second threshold, determine that the quality level of the evaluation target is a low quality level, wherein the first threshold is less than the second threshold.
5. The monitoring point management method according to claim 1 is characterized in that: The step of judging whether it is necessary to manually check the monitoring point based on the quality level of the evaluation target within the plurality of preset time periods comprises: Determining whether the quality level of the evaluation target within the plurality of preset time periods is reduced; If so, an early warning is triggered, and the monitoring point is manually checked; Otherwise, the video frame sequence of the monitoring point in the next preset time period is obtained.
6. The monitoring point management method according to claim 5 is characterized in that: The types of the evaluation targets include at least two of a human target type, a face target type, a motor vehicle target type, and a license plate target type; The step of judging whether the quality level of the evaluation target within the preset time is reduced comprises: Determine whether the quality level of any one of the evaluation targets decreases within the plurality of preset time periods.
7. The monitoring point management method according to claim 1 is characterized in that: The step of obtaining a video frame sequence collected by a monitoring point within a preset time period, and screening out a target image whose image quality meets the set image quality requirement from the video frame sequence, comprises: Performing target tracking on each video frame in the video frame sequence to obtain the video frame where the evaluation target is located and its coordinate information in the current video frame; Using the coordinate information of all evaluation targets of the evaluation target in the current video frame, a corresponding image quality score is obtained; The video frame with the highest image quality score is used as the target image.
8. A monitoring point management device, characterized in that: It comprises a memory and a processor coupled to each other, wherein the memory stores program instructions, and the program instructions are used to be executed by the processor to implement the monitoring point management method described in claims 1-7.
9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is used to implement the monitoring point management method described in claims 1-7.
