An intelligent inspection system based on machine vision

Abstract

This invention relates to the field of image analysis technology, and particularly to an intelligent inspection system based on machine vision, comprising: an inspection area vision module, a visual information filtering module, a visual analysis module, an inspection judgment module, and a visual inspection early warning module. The inspection area vision module determines the location of each detection target and its corresponding inspection sub-region within the inspection area and acquires visual information. The visual information filtering module determines anomaly analysis images in each inspection sub-region. The visual analysis module further determines abnormal behavior information and records the characteristics of abnormal individuals, determining the inspection anomaly representation parameters of the current inspection sub-region. The inspection judgment module determines the inspection category of each detection target based on the inspection representation parameters of each inspection sub-region, thereby determining the target management method for the next inspection cycle. The visual inspection early warning module performs visual locking and issues an early warning. This invention optimizes the allocation of inspection resources and effectively improves the intelligence level and inspection efficiency of the inspection system.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, and in particular to an intelligent inspection system based on machine vision. Background Technology

[0002] In the exhibition field, cameras and sensors are deployed within the exhibition area to continuously collect video and image data. Utilizing deep learning algorithms, the system can identify and analyze visitor behavior, the environment surrounding exhibits, and potential safety hazards in real time. For example, this method can detect whether visitors are getting too close to exhibits, whether there is abnormal crowding behavior, or identify safety risks around exhibits. The system can automatically generate inspection reports and issue timely alerts when anomalies are detected. This intelligent inspection method is widely used in various exhibition venues, improving inspection efficiency and accuracy, reducing the cost and risk of manual inspections, and enhancing the security of exhibits.

[0003] Chinese Patent Publication No. CN118840334A discloses a distributed drone inspection management system based on machine vision, relating to the field of image processing. The system includes a management center, which is communicatively connected to a planning module, a video data acquisition module, a machine vision module, and a feedback module. The planning module divides inspection areas and plans the drone's inspection route according to these areas. The video data acquisition module controls the drone to perform inspection and filming according to the planned inspection route, obtaining color video images of the target facilities. The machine vision module analyzes the obtained color video images of the target facilities and determines whether the target facilities have defects based on the analysis results. The feedback module provides feedback based on the determination results, thus performing comprehensive inspection and filming of the target facilities.

[0004] However, the existing technology still has the following problems: the inspection method is not adjusted by analyzing the detection data of different inspection targets, resulting in a waste of inspection or management resources. Summary of the Invention

[0005] To address this issue, the present invention provides an intelligent inspection system based on machine vision, which overcomes the problem of wasted inspection or management resources caused by failing to analyze and adjust the inspection methods based on the detection data of different inspection targets.

[0006] To achieve the above objectives, the present invention provides an intelligent inspection system based on machine vision, comprising:

[0007] Compared with the prior art, the beneficial effects of the present invention are that the intelligent inspection system based on machine vision proposed in the present invention includes:

[0008] The inspection area vision module is used to determine the location of each detection target and the corresponding inspection sub-area in the inspection area, and continuously acquire the visual information of each inspection sub-area, wherein the visual information includes heat distribution information, flow information and behavior information.

[0009] A visual information filtering module, which is connected to the inspection area visual module, is used to determine the abnormal analysis images in the inspection sub-regions of each detection target based on a preset abnormal behavior image database and the behavior information.

[0010] The visual analysis module is connected to the visual information filtering module and the inspection area visual module respectively. It is used to magnify and refine all the anomaly analysis images, determine the abnormal behavior information in the anomaly analysis images and record the characteristics of the abnormal individuals, and combine the traffic information in each inspection time period to determine the inspection anomaly characterization parameters of the current inspection sub-area.

[0011] The inspection determination module, connected to both the inspection area vision module and the vision analysis module, determines the inspection category of each detection target based on the inspection characterization parameters of each inspection sub-region, thereby determining the target management method for each detection target in the next inspection cycle, including:

[0012] The positions of each detection target remain unchanged, and the corresponding inspection sub-areas are adjusted according to the abnormal behavior information;

[0013] Alternatively, based on the inspection anomaly characterization parameters, each of the detection targets can be classified into actual inspection safety categories, and the position of each detection target can be adjusted according to the determined actual inspection safety category level.

[0014] The visual inspection and early warning module is connected to the inspection area visual module and the visual analysis module, respectively, and is used to visually lock and issue an early warning when the inspection area visual module identifies abnormal heat distribution and / or early warning behavior information.

[0015] Furthermore, the inspection area vision module includes:

[0016] The camera unit is used to continuously acquire video information of each inspection sub-area in the inspection area;

[0017] Infrared unit, used to acquire infrared images of the inspection area;

[0018] The identification unit is connected to the camera unit and the infrared unit respectively, and is used to determine the corresponding traffic information and behavior information based on the video information and each inspection sub-region, and to determine the corresponding heat distribution information based on the infrared image and each inspection sub-region.

[0019] Furthermore, the visual information filtering module compares the abnormal behavior information with the preset abnormal behaviors already existing in the preset abnormal behavior image database, and determines whether there is an abnormal analysis image based on the comparison result.

[0020] If the overlap rate between the identified behavior information of a moving individual and the preset abnormal behavior exceeds the preset overlap rate, then an abnormal analysis image is determined to exist.

[0021] Furthermore, the visual analysis module determines the total number of abnormal behaviors corresponding to each inspection sub-area in different inspection time periods based on the abnormal behavior information determined in the second step and in combination with each inspection time period.

[0022] For a single inspection sub-area, the visual analysis module determines the inspection anomaly factor based on the total number of abnormal behaviors and the corresponding traffic information in a single inspection time period. The visual analysis module then determines the inspection anomaly representation parameter of the current inspection sub-area based on all abnormal inspection factors within the current inspection cycle.

[0023] Furthermore, the inspection judgment module determines the inspection anomaly tendency category of each inspection sub-region based on the inspection anomaly characterization parameters, including,

[0024] If the inspection anomaly characterization parameter is greater than or equal to the standard inspection characterization parameter, then the inspection determination module determines that the inspection category of the corresponding inspection sub-region is a high anomaly tendency inspection category.

[0025] If the abnormality characterization parameter of the inspection is less than the standard inspection characterization parameter, the inspection determination module determines that the inspection category of the corresponding inspection sub-region is a low-abnormality tendency inspection category.

[0026] Furthermore, the inspection determination module determines the target management method for the corresponding detection target in the next inspection cycle based on the inspection category of the inspection sub-region, including:

[0027] If the inspection category of the inspection sub-area is a high anomaly tendency inspection category, the detection target management method determined by the inspection judgment module is to divide each detection target into actual inspection safety categories according to the inspection anomaly characterization parameters, and adjust the position of each detection target according to the determined actual inspection safety category level.

[0028] If the inspection category of the inspection sub-area is a low-abnormality-prone inspection category, the inspection judgment module determines the detection target management method to keep the position of each detection target unchanged, and adjust the corresponding inspection sub-area according to the abnormal behavior information.

[0029] Furthermore, the inspection judgment module determines the total number of safety category classification levels based on the location distribution of the detection targets in the inspection area;

[0030] The inspection judgment module determines the range of values ​​for the inspection anomaly characterization parameters corresponding to each safety category level based on the total number of safety category classification levels and the maximum difference of the inspection anomaly characterization parameters.

[0031] Furthermore, the inspection judgment module places inspection targets of the same level adjacent to each other based on the actual inspection safety category level of each inspection target.

[0032] Furthermore, the inspection judgment module adjusts the corresponding inspection sub-area based on the abnormal behavior information:

[0033] If there is abnormal behavior information that belongs to the warning behavior information, the inspection judgment module determines that the corresponding inspection sub-area needs to be expanded;

[0034] If there is no abnormal behavior information that falls under the category of early warning behavior information, the inspection judgment module determines that the corresponding inspection sub-area range remains unchanged.

[0035] Furthermore, the visual inspection and early warning module visually locks onto the individual exhibiting the early warning behavior based on the characteristics of the abnormal individual.

[0036] Compared with existing technologies, the advantages of this invention are as follows: By accurately identifying abnormal heat and behavior, the system achieves comprehensive and precise monitoring of the inspection area, effectively preventing potential fire sources from damaging exhibits and promptly detecting various abnormal situations; based on overlap rate thresholds set according to different risk levels, it accurately determines abnormal behavior, adapts to inspection needs under different traffic conditions, and avoids misjudgments or omissions that may be caused by fixed thresholds; by calculating inspection anomaly factors and characterization parameters, it accurately reflects the abnormal fluctuations of the inspection sub-area, improving the inspection system's adaptability to different scenarios; it dynamically divides the abnormal tendency categories of the inspection sub-area according to the inspection anomaly characterization parameters, rationally adjusts the allocation of inspection resources, and improves inspection efficiency; it adjusts the location of detection targets according to safety category levels, facilitating centralized management and monitoring of detection targets with similar risk levels, and optimizing resource allocation; it helps to implement targeted safety strategies, strengthen protection for high-risk targets, and simplify processes for low-risk targets to save resources, thereby improving inspection effectiveness; and it facilitates rapid response and handling of abnormal situations, ensuring efficient and accurate inspection work and effectively protecting the safety of the inspection area.

[0037] In particular, in this invention, the camera unit and infrared unit under the visual module of the inspection area capture video information and infrared images of the inspection area respectively. Combined with machine vision-related algorithms, accurate real-time traffic statistics and behavior recognition are achieved, including the identification of abnormal heat sources, effectively preventing potential fire sources from damaging exhibits. The visual information filtering module compares abnormal behavior information with a preset abnormal behavior image database. Based on the overlap rate threshold set for different risk levels, it accurately determines abnormal behavior, realizing comprehensive and accurate monitoring of the inspection area, timely detection and early warning of various abnormal situations, effectively ensuring the safety of the inspection area, especially the exhibition area, rationally allocating inspection resources, and optimizing the inspection process.

[0038] In particular, this invention uses a visual analysis module to determine the total number of abnormal behaviors in each inspection sub-area based on the inspection time period. It then calculates inspection anomaly factors based on abnormal behaviors and traffic flow information, further determining the anomaly characterization parameters to accurately reflect the abnormal fluctuations in the inspection sub-area. The difference in traffic flow information between weekdays and rest days in the exhibition area is addressed by calculating anomaly factors to adapt to inspection needs under different traffic conditions, avoiding misjudgments or omissions that might occur with fixed thresholds, and improving the inspection system's adaptability to different scenarios. Simultaneously, the inspection judgment module divides the inspection sub-areas into high-anomaly and low-anomaly tendency categories based on the inspection anomaly characterization parameters. For high-anomaly tendency areas, inspection efforts can be strengthened specifically; for low-anomaly tendency areas, the investment of inspection resources can be appropriately reduced, optimizing resource allocation and effectively improving the intelligence level and efficiency of the inspection system, providing strong support for the safety management of the inspection area.

[0039] In particular, this invention flexibly determines the management method of the detection target in the next inspection cycle based on the inspection category of the inspection sub-region. For detection targets in the high-anomaly-prone inspection category, the actual inspection safety category is accurately classified by analyzing the inspection anomaly characterization parameters. The position of the detection target is adjusted according to the classification results, thereby effectively enhancing the pertinence and effectiveness of the inspection. It can quickly focus on areas with high anomaly risk, improve the accuracy and efficiency of the inspection, and promptly discover and deal with potential problems. At the same time, it can avoid the excessive dispersion of high-risk detection targets, which would waste unnecessary safety protection resources. For detection targets in the low-anomaly-prone inspection category, their positions remain unchanged, and the inspection sub-region is adjusted only according to the inspection anomaly characterization parameters. This ensures the comprehensiveness of the inspection while avoiding over-inspection of low-risk areas, rationally allocating inspection resources, optimizing the inspection process, reducing inspection costs, improving the overall rationality of the inspection, and ensuring that the inspection system can operate efficiently and intelligently to better meet the needs of different inspection scenarios.

[0040] In particular, in this invention, the inspection judgment module places targets of the same level adjacently based on their actual inspection safety category level. This facilitates centralized management and monitoring of targets with similar risk levels, improving inspection efficiency. Furthermore, it rationally allocates inspection resources according to risk level, strengthening inspections in high-risk areas and optimizing resource allocation. Simultaneously, it helps implement targeted safety strategies, such as strengthening protection for high-risk targets and simplifying processes for low-risk targets to save resources, thereby improving inspection effectiveness. In addition, it facilitates rapid response and handling of abnormal situations, ensuring efficient and accurate inspection work and effectively protecting the safety of the inspection area. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the structure of an intelligent inspection system based on machine vision according to an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of the structure of the inspection area vision module according to an embodiment of the present invention;

[0043] Figure 3 A logic diagram for determining the inspection anomaly tendency category of each inspection sub-region in an embodiment of the present invention;

[0044] Figure 4 This is a logic diagram illustrating how the corresponding inspection sub-regions are adjusted based on abnormal behavior information according to an embodiment of the present invention. Detailed Implementation

[0045] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0046] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0047] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0048] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0049] Please see Figure 1 As shown, this is a structural schematic diagram of an intelligent inspection system based on machine vision according to an embodiment of the present invention. The invention provides an intelligent inspection system based on machine vision, comprising:

[0050] The inspection area vision module is used to determine the location of each detection target and the corresponding inspection sub-area in the inspection area, and continuously acquire the visual information of each inspection sub-area, wherein the visual information includes heat distribution information, flow information and behavior information.

[0051] A visual information filtering module, which is connected to the inspection area visual module, is used to determine the abnormal analysis images in the inspection sub-regions of each detection target based on a preset abnormal behavior image database and the behavior information.

[0052] The visual analysis module is connected to the visual information filtering module and the inspection area visual module respectively. It is used to magnify and refine all the anomaly analysis images, determine the abnormal behavior information in the anomaly analysis images and record the characteristics of the abnormal individuals, and combine the traffic information in each inspection time period to determine the inspection anomaly characterization parameters of the current inspection sub-area.

[0053] The inspection determination module, connected to both the inspection area vision module and the vision analysis module, determines the inspection category of each detection target based on the inspection characterization parameters of each inspection sub-region, thereby determining the target management method for each detection target in the next inspection cycle, including:

[0054] The positions of each detection target remain unchanged, and the corresponding inspection sub-areas are adjusted according to the abnormal behavior information;

[0055] Alternatively, based on the inspection anomaly characterization parameters, each of the detection targets can be classified into actual inspection safety categories, and the position of each detection target can be adjusted according to the determined actual inspection safety category level.

[0056] The visual inspection and early warning module is connected to the inspection area visual module and the visual analysis module, respectively, and is used to visually lock and issue an early warning when the inspection area visual module identifies abnormal heat distribution and / or early warning behavior information.

[0057] In implementation, the center of the inspection sub-area is the detection target. It's understandable that the location of the detection target, i.e., the exhibit, is a key factor in determining the inspection sub-area. Inspection sub-areas are typically delineated with the detection target as the center, but the presence or absence of a warning line will affect the definition. When a warning line is included, the inspection sub-area's scope includes the warning line as a boundary, indicating that personnel must not cross it. In this case, the inspection sub-area covers not only the space around the exhibit but also the area enclosed by the warning line, used to monitor personnel behavior. When a warning line is not included, the delineation of the inspection sub-area is more precise, focusing primarily on a core area within a certain range around the exhibit. This area is centered on the exhibit and, based on inspection needs and the exhibition layout, is a tightly enclosed area. The size of this area is usually determined by the exhibit's size and shape, as well as the required coverage area for inspection, to ensure effective monitoring of the exhibit's condition and its surrounding environment.

[0058] For detection targets that do not include warning lines, the preferred method for determining the radius of the inspection sub-region is to refer to the theoretical safety distance determination method used in museums.

[0059] In this invention, by accurately identifying abnormal heat and behavior, the system achieves comprehensive and precise monitoring of the inspection area, effectively preventing potential fire sources from damaging exhibits and promptly detecting various anomalies. Based on overlap rate thresholds set for different risk levels, it accurately determines abnormal behavior, adapting to inspection needs under varying traffic conditions and avoiding misjudgments or omissions that may occur with fixed thresholds. By calculating inspection anomaly factors and characterization parameters, it accurately reflects the abnormal fluctuations in the inspection sub-areas, improving the system's adaptability to different scenarios. Based on the inspection anomaly characterization parameters, it dynamically classifies the anomaly tendency categories of the inspection sub-areas, rationally adjusting the allocation of inspection resources and improving inspection efficiency. It adjusts the location of detection targets according to safety category levels, facilitating centralized management and monitoring of detection targets with similar risk levels, optimizing resource allocation. This helps implement targeted safety strategies, strengthening protection for high-risk targets and simplifying processes for low-risk targets to save resources, thereby improving inspection effectiveness. It also facilitates rapid response and handling of abnormal situations, ensuring efficient and accurate inspection work and effectively protecting the safety of the inspection area.

[0060] Please see Figure 2 As shown, this is a structural schematic diagram of the inspection area vision module according to an embodiment of the present invention. The inspection area vision module includes:

[0061] The camera unit is used to continuously acquire video information of each inspection sub-area in the inspection area;

[0062] Infrared unit, used to acquire infrared images of the inspection area;

[0063] The identification unit is connected to the camera unit and the infrared unit respectively, and is used to determine the corresponding traffic information and behavior information based on the video information and each inspection sub-region, and to determine the corresponding heat distribution information based on the infrared image and each inspection sub-region.

[0064] In practice, the identification unit can obtain the heat distribution information of each inspection sub-region by combining the infrared image of the inspection area acquired by the infrared unit with the distribution of the inspection sub-regions, thereby identifying whether there is abnormal heat near the detection target, such as open flames, cigarette butts or other sources of fire that may damage the detection target. The acquisition of video information by the camera unit and the acquisition of infrared images by the infrared unit are either existing technologies or not, and no specific limitation is made.

[0065] The recognition unit adopts an improved target detection algorithm based on YOLOv7, combined with optical flow and region convolution feature extraction technology, to achieve the following within a pre-divided inspection sub-region grid: real-time traffic statistics based on moving target trajectory tracking; and detection of abnormal behavior patterns through a behavior recognition model (e.g., 3D-CNN + spatiotemporal attention mechanism). The recognition method can be any of the existing technologies, which will not be elaborated here.

[0066] Specifically, the visual information filtering module compares the abnormal behavior information with the preset abnormal behaviors already existing in the preset abnormal behavior image database, and determines whether there is an abnormal analysis image based on the comparison result.

[0067] If the overlap rate between the identified behavior information of a moving individual and the preset abnormal behavior exceeds the preset overlap rate, then an abnormal analysis image is determined to exist.

[0068] In practice, the initial abnormal behavior data in the abnormal behavior database is determined based on the behavior data of valid abnormal individuals recorded in different inspection areas.

[0069] The pre-defined abnormal behavior image database contains several behavioral levels. In the abnormal behavior recognition system for exhibition areas, different pre-defined overlap rate thresholds are set according to the behavioral risk level. For Level 1 high-risk behaviors, such as violent destruction of exhibits, arson, and fighting, the pre-defined overlap rate is greater than 85%. The matching method includes multi-dimensional strong matching of action features, trajectory features, and temporal features to ensure high-precision recognition. For Level 2 risk behaviors, such as climbing exhibition stands and touching cultural relics beyond the designated area, the pre-defined overlap rate ranges from 75% to 80%. The matching method is feature matching combined with environmental verification. For Level 3 alert behaviors, such as prolonged stay in no-parking areas and abnormal gatherings, the pre-defined overlap rate ranges from 65% to 70%. The system uses basic behavioral pattern matching and short-term continuous triggering logic judgment. In addition, a child missing warning scenario is also set up. When a child is detected wandering alone in a dangerous area, if the overlap rate between height feature recognition and the guardian's separation detection is ≥60% and lasts for 5 seconds, an alert is triggered.

[0070] In this invention, the camera unit and infrared unit under the visual module of the inspection area capture video information and infrared images of the inspection area, respectively. Combined with machine vision-related algorithms, accurate real-time traffic statistics and behavior recognition are achieved, including the identification of abnormal heat sources, effectively preventing potential fire sources from damaging exhibits. The visual information filtering module compares abnormal behavior information with a preset abnormal behavior image database. Based on the overlap rate threshold set for different risk levels, it accurately determines abnormal behavior, realizing comprehensive and accurate monitoring of the inspection area, timely detection and early warning of various abnormal situations, effectively ensuring the safety of the inspection area, especially the exhibition area, rationally allocating inspection resources, and optimizing the inspection process.

[0071] Please see Figure 3 As shown, it is a logic diagram for determining the inspection anomaly tendency category of each inspection sub-area in an embodiment of the present invention. The visual analysis module determines the total number of abnormal behaviors corresponding to each inspection sub-area in different inspection time periods based on the abnormal behavior information determined in the second determination and each inspection time period.

[0072] For a single inspection sub-area, the visual analysis module determines the inspection anomaly factor based on the total number of abnormal behaviors and the corresponding traffic information in a single inspection time period. The visual analysis module then determines the inspection anomaly representation parameter of the current inspection sub-area based on all abnormal inspection factors within the current inspection cycle.

[0073] In practice, the visual analysis module counts abnormal behaviors based on the characteristics of abnormal individuals.

[0074] The inspection time period is selected from 0.5 hours to 2 hours, preferably 1 hour. The inspection cycle is 0.5 days or 1 day, preferably 1 day. If there are many sub-areas to be inspected, 0.5 days can be selected according to the actual situation to ensure the integrity of the inspected sub-areas.

[0075] The inspection anomaly characterization parameters are determined based on the average deviation of all inspection anomaly factors and the average value of all inspection anomaly factors within a single inspection cycle.

[0076] It is understandable that for a single inspection area, the traffic flow information differs in different inspection cycles. The nature of the exhibition area affects the traffic flow information on weekdays and rest days. Therefore, it is necessary to determine the abnormal factors of each inspection sub-area to characterize the abnormal tendency of the corresponding detection target in the exhibition area. This provides data support for classifying the abnormal tendency categories of inspection sub-areas (i.e. detection targets), facilitating subsequent adaptive adjustments to the management methods of detection targets. Compared with directly setting a threshold for the number of abnormal behaviors, this method is more accurate, better characterizing the likelihood of exhibit abnormalities when there are too many personnel (more abnormal behaviors) or too few personnel (fewer abnormal behaviors). Under different traffic conditions, the abnormal tendency of the detection target is determined based on the dispersion of the abnormality ratio. If the dispersion is significant, it is determined that the current detection target is more likely to be damaged and needs to be adjusted in a timely manner.

[0077] Specifically, the inspection judgment module determines the inspection anomaly tendency category of each inspection sub-region based on the inspection anomaly characterization parameters, including,

[0078] If the inspection anomaly characterization parameter is greater than or equal to the standard inspection characterization parameter, then the inspection determination module determines that the inspection category of the corresponding inspection sub-region is a high anomaly tendency inspection category.

[0079] If the abnormality characterization parameter of the inspection is less than the standard inspection characterization parameter, the inspection determination module determines that the inspection category of the corresponding inspection sub-region is a low-abnormality tendency inspection category.

[0080] In practice, the standard inspection characterization parameters are selected from the interval consisting of the average value and twice the standard deviation of the inspection anomaly characterization parameters in the historical records of unadjusted detection target locations.

[0081] In this invention, a visual analysis module, combined with the inspection time period, determines the total number of abnormal behaviors in each inspection sub-area. Based on the abnormal behaviors and traffic flow information, anomaly factors are calculated to further determine anomaly characterization parameters, thereby accurately reflecting the abnormal fluctuations in the inspection sub-area. The difference in traffic flow information between weekdays and rest days in the exhibition area is addressed by calculating anomaly factors to adapt to inspection needs under different traffic conditions, avoiding misjudgments or omissions that might occur with fixed thresholds, and improving the inspection system's adaptability to different scenarios. Simultaneously, the inspection judgment module divides the inspection sub-areas into high-anomaly and low-anomaly tendency categories based on the anomaly characterization parameters. For high-anomaly tendency areas, inspection efforts can be strengthened specifically; for low-anomaly tendency areas, inspection resource investment can be appropriately reduced, optimizing resource allocation and effectively improving the intelligence level and efficiency of the inspection system, providing strong support for the safety management of the inspection area.

[0082] Specifically, the inspection judgment module determines the target management method for the corresponding detection target in the next inspection cycle based on the inspection category of the inspection sub-region, including:

[0083] If the inspection category of the inspection sub-area is a high anomaly tendency inspection category, the detection target management method determined by the inspection judgment module is to divide each detection target into actual inspection safety categories according to the inspection anomaly characterization parameters, and adjust the position of each detection target according to the determined actual inspection safety category level.

[0084] If the inspection category of the inspection sub-area is a low-abnormality-prone inspection category, the inspection judgment module determines the detection target management method to keep the position of each detection target unchanged, and adjust the corresponding inspection sub-area according to the abnormal behavior information.

[0085] In this invention, the management method for detection targets in the next inspection cycle is flexibly determined based on the inspection category of the inspection sub-region. For detection targets in the high-anomaly-prone inspection category, the actual inspection safety category is accurately classified by analyzing the inspection anomaly characterization parameters. The position of the detection target is then adjusted based on the classification results, thereby effectively enhancing the targeting and effectiveness of the inspection. This allows for rapid focus on areas with high anomaly risk, improving the accuracy and efficiency of the inspection, timely detection and handling of potential problems, and avoiding the overly dispersed distribution of high-risk detection targets, thus preventing unnecessary waste of safety protection resources. For detection targets in the low-anomaly-prone inspection category, their positions remain unchanged, and the inspection sub-region is adjusted only based on the inspection anomaly characterization parameters. This ensures comprehensive inspection while avoiding over-inspection of low-risk areas, rationally allocating inspection resources, optimizing the inspection process, reducing inspection costs, improving the overall rationality of the inspection, and ensuring that the inspection system can operate efficiently and intelligently to better meet the needs of different inspection scenarios.

[0086] Specifically, the inspection judgment module determines the total number of safety category classification levels based on the location distribution of the detection targets in the inspection area;

[0087] The inspection judgment module determines the range of values ​​for the inspection anomaly characterization parameters corresponding to each safety category level based on the total number of safety category classification levels and the maximum difference of the inspection anomaly characterization parameters.

[0088] In practice, the total number of safety category levels is determined based on the location distribution of the targets within the inspection area. If the inspection area is divided into several sub-areas, the total number can be determined based on the sub-areas and their corresponding areas. For the exhibition area in the main hall, the number of interconnected target placement blocks can be determined based on the placement of the targets, thus determining the total number of safety category levels. The method for determining the total number of levels is based on the actual situation of the inspection area.

[0089] Based on the total number of safety category levels and the maximum difference in the anomaly characteristic parameters, the inspection judgment module divides the value range of the anomaly characteristic parameters into multiple intervals, each interval corresponding to a safety category level. Specifically, the maximum difference in the anomaly characteristic parameters is equally divided and allocated to each safety category level, so that each level has a clear value range for the anomaly characteristic parameters. The preferred method is equal division.

[0090] For example, if the total number of safety category levels is 3, the maximum difference in the inspection anomaly characterization parameter is A, the minimum value of the inspection anomaly characterization parameter is a1, and the maximum value is a2, then the range of values ​​for the inspection anomaly characterization parameter is divided into three intervals: [a1, A / 3), [a1+A / 3, a1+2A / 3), and [a1+2A / 3, a2], corresponding to low, medium, and high safety category levels, respectively. In this way, when the inspection anomaly characterization parameter falls into different intervals, its corresponding safety category level can be quickly determined, thus providing a basis for subsequent inspection strategy adjustments.

[0091] Specifically, the inspection judgment module places inspection targets of the same level adjacent to each other based on the actual inspection safety category level of each inspection target.

[0092] In this invention, the inspection judgment module places targets of the same level adjacently based on their actual inspection safety category level. This facilitates centralized management and monitoring of targets with similar risk levels, improving inspection efficiency. It also rationally allocates inspection resources according to risk level, strengthening inspections in high-risk areas and optimizing resource allocation. Furthermore, it helps implement targeted safety strategies, such as strengthening protection for high-risk targets and simplifying processes for low-risk targets to save resources, thereby improving inspection effectiveness. In addition, it facilitates rapid response and handling of abnormal situations, ensuring efficient and accurate inspection work and effectively protecting the safety of the inspected area.

[0093] Please see Figure 4 As shown, this is a logic diagram of how the inspection sub-area is adjusted based on abnormal behavior information according to an embodiment of the present invention. The inspection determination module adjusts the corresponding inspection sub-area based on the abnormal behavior information.

[0094] If there is abnormal behavior information that belongs to the warning behavior information, the inspection judgment module determines that the corresponding inspection sub-area needs to be expanded;

[0095] If there is no abnormal behavior information that falls under the category of early warning behavior information, the inspection judgment module determines that the corresponding inspection sub-area range remains unchanged.

[0096] In practice, the warning behavior information includes abnormal behavior information of Level 1 high-risk behavior and Level 2 risk behavior.

[0097] Specifically, the visual inspection and early warning module visually locks onto individuals exhibiting warning behavior based on the characteristics of the abnormal individuals.

[0098] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0099] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A machine vision-based intelligent inspection system, characterized in that, include: The inspection area vision module is used to determine the location of each detection target and the corresponding inspection sub-area in the inspection area, and continuously acquire the visual information of each inspection sub-area, wherein the visual information includes heat distribution information, flow information and behavior information. A visual information filtering module, which is connected to the inspection area visual module, is used to determine the abnormal analysis images in the inspection sub-regions of each detection target based on a preset abnormal behavior image database and the behavior information. The visual analysis module is connected to the visual information filtering module and the inspection area visual module respectively. It is used to magnify and refine all the anomaly analysis images, determine the abnormal behavior information in the anomaly analysis images and record the characteristics of the abnormal individuals, and combine the traffic information in each inspection time period to determine the inspection anomaly characterization parameters of the current inspection sub-area. Among them, based on the abnormal behavior information determined in the second step and combined with each inspection time period, the total number of abnormal behaviors corresponding to each inspection sub-area in different inspection time periods is determined. For a single inspection sub-area, the visual analysis module determines the inspection anomaly factor based on the total number of abnormal behaviors and the corresponding traffic information in a single inspection time period. The visual analysis module determines the inspection anomaly characterization parameter of the current inspection sub-area based on all abnormal inspection factors in the current inspection cycle. The inspection anomaly characterization parameter is determined based on the average deviation of all inspection anomaly factors and the average value of all inspection anomaly factors within a single inspection cycle. The inspection determination module, connected to both the inspection area vision module and the vision analysis module, determines the inspection category of each detected target based on the inspection characterization parameters of each inspection sub-region. The inspection categories include high anomaly tendency and low anomaly tendency categories, to determine the target management method for each detected target in the next inspection cycle, including: If the inspection category of the inspection sub-area is a low-abnormality-prone inspection category, the determined detection target management method is to keep the position of each detection target unchanged, and adjust the corresponding inspection sub-area according to the abnormal behavior information; If the inspection category of the inspection sub-area is a high anomaly tendency inspection category, the determined detection target management method is to divide each detection target into actual inspection safety categories according to the inspection anomaly characterization parameters, and place detection targets of the same level adjacent to each other according to the determined actual inspection safety category level. The visual inspection and early warning module is connected to the visual module of the inspection area and the visual analysis module, respectively, and is used to visually lock and issue an early warning when the visual module of the inspection area identifies abnormal heat distribution and / or early warning behavior information.

2. The intelligent inspection system based on machine vision according to claim 1, characterized in that, The inspection area vision module includes: The camera unit is used to continuously acquire video information of each inspection sub-area in the inspection area; Infrared unit, used to acquire infrared images of the inspection area; The identification unit is connected to the camera unit and the infrared unit respectively, and is used to determine the corresponding traffic information and behavior information based on the video information and each inspection sub-region, and to determine the corresponding heat distribution information based on the infrared image and each inspection sub-region.

3. The intelligent inspection system based on machine vision according to claim 2, characterized in that, The visual information filtering module compares the abnormal behavior information with the preset abnormal behaviors already existing in the preset abnormal behavior image database, and determines whether there are any abnormal analysis images based on the comparison results. If the overlap rate between the identified behavior information of a moving individual and the preset abnormal behavior exceeds the preset overlap rate, then an abnormal analysis image is determined to exist.

4. The intelligent inspection system based on machine vision according to claim 3, characterized in that, The inspection judgment module determines the inspection anomaly tendency category of each inspection sub-region based on the inspection anomaly characterization parameters, including: If the inspection anomaly characterization parameter is greater than or equal to the standard inspection characterization parameter, then the inspection determination module determines that the inspection category of the corresponding inspection sub-region is a high anomaly tendency inspection category. If the abnormality characterization parameter of the inspection is less than the standard inspection characterization parameter, the inspection determination module determines that the inspection category of the corresponding inspection sub-region is a low-abnormality tendency inspection category.

5. The intelligent inspection system based on machine vision according to claim 4, characterized in that, The inspection judgment module determines the total number of safety category classification levels based on the location distribution of the detection targets in the inspection area; The inspection judgment module determines the range of values ​​for the inspection anomaly characterization parameters corresponding to each safety category level based on the total number of safety category classification levels and the maximum difference of the inspection anomaly characterization parameters.

6. The intelligent inspection system based on machine vision according to claim 5, characterized in that, The inspection judgment module adjusts the corresponding inspection sub-area based on the abnormal behavior information: If there is abnormal behavior information that belongs to the warning behavior information, the inspection judgment module determines that the corresponding inspection sub-area needs to be expanded; If there is no abnormal behavior information that falls under the category of early warning behavior information, the inspection judgment module determines that the corresponding inspection sub-area range remains unchanged.

7. The intelligent inspection system based on machine vision according to claim 6, characterized in that, The visual inspection and early warning module visually locks onto individuals exhibiting abnormal behavior based on their characteristics.

Images