Data processing-based scope target recognition method and system

Abstract

The application provides a data processing-based scope target recognition method and system, relates to the technical field of scopes, and performs abnormality detection by synchronizing multiple regional traffic video streams to a traffic anomaly recognition model, obtains multiple abnormality detection results; according to traffic flow direction information and multiple regional spatial identifiers, the multiple abnormality detection results are associated and verified to obtain a traffic anomaly tracking object; a scope camera locates and tracks the traffic anomaly tracking object in a traffic monitoring scene according to a recognition feature label, and obtains real-time tracking data to perform traffic temporary control. The technical problems that the existing technology has weak traffic violation anomaly recognition intervention, leading to insufficient interruption of violation behavior, and causing safety hazards to road traffic environment are solved. The technical effects of comprehensive monitoring and management of urban traffic violation behavior, effectively improving the intelligent level of traffic monitoring, and ensuring the safety and efficiency of road traffic are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sighting scopes, and particularly relates to a sighting scope target recognition method and system based on data processing. BACKGROUND

[0002] At present, traffic monitoring systems have obvious deficiencies in identifying and intervening in violations. For example, at busy intersections during peak hours, although cameras are installed, the system often can only perform basic video recording and lacks the ability for real-time analysis and rapid response.

[0003] When a violation is detected, such as a vehicle running a red light or changing lanes irregularly, immediate identification and measures cannot be taken, which may allow the violation to continue and increase the risk of traffic accidents.

[0004] In addition, due to the lack of intelligent tracking and real-time risk assessment, personnel cannot be dispatched or signal lights cannot be adjusted to interrupt these violations in a timely manner, so that the road traffic order and safety cannot be effectively maintained.

[0005] For example, a driver is speeding on an urban expressway, and the existing system may only record after an accident occurs, and cannot identify and take measures in a timely manner to prevent potential collisions before the accident occurs. The lack of timely intervention not only poses a threat to the safety of the driver and passengers, but also poses a safety hazard to other road users and the overall traffic environment.

[0006] In summary, the prior art has the technical problem of weak traffic violation abnormality identification intervention, which leads to insufficient interruption timeliness of violations and causes safety hazards to the road traffic environment. SUMMARY

[0007] The present application provides a sighting scope target recognition method and system based on data processing, which is used to solve the technical problem of weak traffic violation abnormality identification intervention in the prior art, which leads to insufficient interruption timeliness of violations and causes safety hazards to the road traffic environment.

[0008] In view of the above problems, the present application provides a sighting scope target recognition method and system based on data processing.

[0009] In a first aspect, the application provides a data processing-based scope target recognition method, which comprises: interacting with a high-position video camera of a traffic monitoring scene to obtain a real-time traffic video stream; presetting an intra-frame partition ratio and partitioning the real-time traffic video stream according to the intra-frame partition ratio to obtain a plurality of regional traffic video streams, wherein the plurality of regional traffic video streams are specifically a plurality of regional spatial identifiers; pre-constructing a traffic anomaly recognition model, synchronizing the plurality of regional traffic video streams to the traffic anomaly recognition model to perform anomaly detection, and obtaining a plurality of anomaly detection results; interacting to obtain traffic flow direction information of the traffic monitoring scene; correlating and checking the plurality of anomaly detection results according to the traffic flow direction information and the plurality of regional spatial identifiers to obtain a traffic anomaly tracking object, wherein the traffic anomaly tracking object has an identification feature label; activating and dispatching a scope camera, the scope camera locating and tracking the traffic anomaly tracking object in the traffic monitoring scene according to the identification feature label to obtain real-time tracking data; and locating a traffic regulation space according to the real-time tracking data and performing temporary regulation and control of traffic in the traffic regulation space.

[0010] In a second aspect, the application provides a data processing-based scope target recognition system, which comprises: a traffic video acquisition unit configured to interact with a high-position video camera of a traffic monitoring scene to obtain a real-time traffic video stream; a data partition execution unit configured to preset an intra-frame partition ratio and partition the real-time traffic video stream according to the intra-frame partition ratio to obtain a plurality of regional traffic video streams, wherein the plurality of regional traffic video streams are specifically a plurality of regional spatial identifiers; an anomaly detection execution unit configured to pre-construct a traffic anomaly recognition model, synchronize the plurality of regional traffic video streams to the traffic anomaly recognition model to perform anomaly detection, and obtain a plurality of anomaly detection results; a traffic flow direction acquisition unit configured to interact to obtain traffic flow direction information of the traffic monitoring scene; a correlation and checking execution unit configured to correlate and check the plurality of anomaly detection results according to the traffic flow direction information and the plurality of regional spatial identifiers to obtain a traffic anomaly tracking object, wherein the traffic anomaly tracking object has an identification feature label; a real-time tracking execution unit configured to activate and dispatch a scope camera, the scope camera locating and tracking the traffic anomaly tracking object in the traffic monitoring scene according to the identification feature label to obtain real-time tracking data; and a regulation space determination unit configured to locate a traffic regulation space according to the real-time tracking data and perform temporary regulation and control of traffic in the traffic regulation space.

[0011] The one or more technical solutions provided in the application have at least the following technical effects or advantages:

[0012] The method provided by the embodiment of the application obtains real-time traffic video streams through an interactive high-position video camera of a traffic monitoring scene; a preset intra-frame partition ratio is used to divide the real-time traffic video streams to obtain a plurality of regional traffic video streams, wherein the plurality of regional traffic video streams are specifically a plurality of regional spatial identifiers; a traffic anomaly recognition model is pre-constructed, the plurality of regional traffic video streams are synchronized to the traffic anomaly recognition model to perform anomaly detection to obtain a plurality of anomaly detection results; traffic flow direction information of the traffic monitoring scene is obtained interactively; the plurality of anomaly detection results are associated and checked according to the traffic flow direction information and the plurality of regional spatial identifiers to obtain a traffic anomaly tracking object, wherein the traffic anomaly tracking object has an identification feature label; a scope camera is activated and dispatched, the scope camera is positioned and tracks the traffic anomaly tracking object in the traffic monitoring scene according to the identification feature label to obtain real-time tracking data; and a traffic regulation space is positioned according to the real-time tracking data, and temporary traffic control is performed in the traffic regulation space. The technical effect of comprehensive monitoring and management of urban traffic violations, effectively improving the intelligent level of traffic monitoring, and guaranteeing the safety and efficiency of road traffic is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 A flowchart of a scope target identification method based on data processing provided by the application is shown.

[0014] Figure 2 A flowchart of constructing a traffic anomaly recognition model in a scope target identification method based on data processing provided by the application is shown.

[0015] Figure 3 A structural diagram of a scope target identification system based on data processing provided by the application is shown.

[0016] BRIEF DESCRIPTION OF DRAWINGS Traffic video acquisition unit 11, data segmentation execution unit 12, anomaly detection execution unit 13, traffic flow direction obtaining unit 14, associated checking execution unit 15, real-time tracking execution unit 16, and regulation space determination unit 17. DETAILED DESCRIPTION

[0017] The scope target identification method and system based on data processing provided by the application are used to solve the technical problem that the intervention of traffic violation anomaly recognition in the prior art is weak, resulting in insufficient timeliness of interrupting the violation behavior and causing safety hazards to the road traffic environment. The technical effect of comprehensive monitoring and management of urban traffic violations, effectively improving the intelligent level of traffic monitoring, and guaranteeing the safety and efficiency of road traffic is achieved.

[0018] Below, the technical solutions in the present application will be described clearly and completely with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application. In addition, it should be noted that, for the convenience of description, only the parts related to the present application are shown in the drawings, not all.

[0019] Embodiment one

[0020] As Figure 1 shown, the present application provides a data processing-based scope identification method for a scope, which comprises:

[0021] A100: a high-position video camera for an interactive traffic monitoring scene, obtaining a real-time traffic video stream.

[0022] A200: presetting an intra-frame partition ratio and partitioning the real-time traffic video stream according to the intra-frame partition ratio to obtain a plurality of regional traffic video streams, wherein the plurality of regional traffic video streams are specifically a plurality of regional spatial identifiers.

[0023] Specifically, it should be understood that the high-position video camera is a monitoring camera installed at a certain height, which can cover a wider traffic area and is usually used to capture a more comprehensive traffic view.

[0024] In this embodiment, the real-time traffic video stream is a continuous video frame captured by the high-position video camera, which is used to monitor the traffic situation in real time. The data is extracted by using the high-position video camera for the interactive traffic monitoring scene in real time to obtain the real-time traffic video stream.

[0025] The intra-frame partition ratio is a parameter for defining how to partition a single video frame into a plurality of small blocks or regions. For example, “5:4 horizontal-vertical partitioning” means that the video frame is partitioned into 4 equal parts in the vertical direction and 5 equal parts in the horizontal direction. This partitioning method will generate a total of 5x4=20 regions.

[0026] The captured real-time traffic video stream is partitioned according to the intra-frame partition ratio, and a plurality of regional traffic video streams of a plurality of fixed video frame regions are obtained after partitioning. Each fixed video frame region corresponds to an independent regional traffic video stream, and each regional traffic video stream is a subset of the original real-time traffic video stream.

[0027] Each partitioned fixed video frame region has its specific regional spatial identifier, which is the fixed position information (such as row number and column number) of the fixed video frame region in the video frame of the real-time traffic video stream.

[0028] The multiple region traffic video stream identifiers corresponding to the multiple fixed video frames are identified by using multiple region spatial identifiers of the multiple fixed video frames.

[0029] The embodiment realizes the separate processing and analysis of each region video stream by performing video segmentation, thereby realizing more accurate traffic flow monitoring, vehicle behavior analysis and abnormal event detection.

[0030] A300: Pre-construct a traffic anomaly recognition model, synchronize the multiple region traffic video streams to the traffic anomaly recognition model to perform abnormal detection, and obtain multiple abnormal detection results.

[0031] In one embodiment, as shown in Figure 2 The method step A300 provided by the application further includes:

[0032] A preset sample coverage condition is set, and data collection is performed according to the sample coverage condition to obtain multiple sample traffic video streams; the multiple sample traffic video streams are converted into multiple sample video image frame sequences; object-level labeling is performed on the multiple sample video image frame sequences to obtain multiple sample violation identifier sequences; a standard CNN model is called from an operator library, and the multiple sample video image frame sequences and multiple sample violation identifier sequences are used as training data to perform real-time performance optimization of the standard CNN model to obtain a standard traffic anomaly recognition branch; K standard traffic anomaly recognition branches are copied and connected in parallel to generate the traffic anomaly recognition model, wherein K is a positive integer greater than 30.

[0033] In one embodiment, the multiple region traffic video streams are synchronized to the traffic anomaly recognition model to perform abnormal detection to obtain multiple abnormal detection results, and the method step A300 provided by the application further includes:

[0034] According to the multiple region traffic video streams, multiple standard traffic anomaly recognition branches are randomly activated in the K standard traffic anomaly recognition branches of the traffic anomaly recognition model; the multiple region traffic video streams are converted into multiple region image frame sequences; the multiple region image frame sequences are used as input data to synchronize to the multiple standard traffic anomaly recognition branches to perform abnormal detection to obtain multiple traffic anomaly recognition results; a set of verification feature indicators is preset, and feature extraction is performed on the multiple traffic anomaly recognition results by using the set of verification feature indicators to obtain multiple sets of abnormal verification features; the multiple traffic anomaly recognition results are identified by using the multiple sets of abnormal verification features to obtain the multiple abnormal detection results.

[0035] Specifically, in the present embodiment, the sample coverage condition is used to ensure the diversity of the collected data and cover different traffic conditions, such as requiring different time periods (morning peak, evening peak, night, etc.), weather conditions (sunny, rainy, foggy, snowy, etc.), lighting conditions (daytime, dusk, night lighting, etc.), and other factors that may affect traffic flow (such as special events, holidays, etc.).

[0036] According to the sample coverage condition, a plurality of sample traffic video streams are obtained by collecting video streams from different locations, different time periods, different weather and lighting conditions, to ensure the diversity and representativeness of the plurality of sample traffic video streams.

[0037] The plurality of sample traffic video streams are parsed into discrete image frames to obtain a plurality of sample video image frame sequences, laying the foundation for subsequent image processing and analysis.

[0038] The plurality of sample video image frame sequences are subjected to object-level labeling to obtain a plurality of sample violation identification sequences, wherein the object-level labeling is a detailed labeling of key objects (such as vehicles, pedestrians, traffic signs, etc.) in each image frame, especially identifying and marking violations (such as running red lights, reverse driving, illegal parking, etc.).

[0039] An appropriate CNN (Convolutional Neural Network) model architecture is selected from the operator library, such as ResNet, Inception, etc., as the standard CNN model, and the plurality of sample video image frame sequences and the plurality of sample violation identification sequences are used as training data to adjust and optimize the standard CNN model to meet the real-time performance requirements, for example, the real-time performance requirements require the model to have a stable accuracy of more than 98% in identifying violations, and a standard traffic anomaly identification branch is obtained.

[0040] Through the training and verification process, the parameters and structure of the standard CNN model are optimized to ensure that the obtained standard traffic anomaly identification branch can quickly and accurately identify traffic anomalies in real-time environments. It should be noted that the training and verification process of the standard CNN model using training data in the present embodiment is a conventional CNN model training and optimization method.

[0041] K standard traffic anomaly identification branches are replicated, where K is a positive integer greater than 30. It should be understood that the intra-frame partition ratio in the present embodiment divides the image frame into 30 regions in general, so at least 30 standard traffic anomaly identification branches should be replicated to ensure that each region of the traffic video stream can be assigned a standard traffic anomaly identification branch for independent traffic anomaly identification.

[0042] Parallelly connecting the K standard traffic anomaly identification branches generates the traffic anomaly identification model, and the traffic anomaly identification model has the advantage that it can process traffic anomaly identification tasks of multiple regional traffic video streams in parallel to improve overall processing efficiency.

[0043] In this embodiment, the traffic anomaly identification model includes K standard traffic anomaly identification branches, and the K standard traffic anomaly identification branches can be randomly activated according to the number of real-time traffic video streams.

[0044] Based on this, the embodiment randomly activates K standard traffic anomaly identification branches of the traffic anomaly identification model according to the specific number of the multiple regional traffic video streams, and the specific number of the multiple regional traffic video streams is equal to the multiple standard traffic anomaly identification branches.

[0045] The multiple regional traffic video streams are parsed into discrete image frames to obtain multiple regional image frame sequences, and the multiple regional image frame sequences are synchronized to the activated multiple standard traffic anomaly identification branches as input data to independently perform anomaly detection to obtain multiple traffic anomaly identification results. Each traffic anomaly detection result is a rule-violating object and a specific rule-violating behavior identified by the corresponding video stream.

[0046] A set of preset verification feature indicators is provided, and the set of verification feature indicators specifically includes the model, color, brand, license plate number, speed, and direction data of the vehicle.

[0047] The multiple traffic anomaly identification results are subjected to feature extraction using the set of verification feature indicators to obtain multiple anomaly verification feature sets, and the multiple anomaly verification feature sets are used to map and identify the multiple traffic anomaly identification results to obtain the multiple anomaly detection results.

[0048] The traffic anomaly identification model is constructed to identify traffic rule violations in real-time traffic video streams, which achieves more accurate and efficient identification of traffic anomalies and provides a reference for subsequent management of traffic anomalies.

[0049] A400: Interactively obtain traffic flow direction information of the traffic monitoring scene.

[0050] Specifically, in this embodiment, traffic flow direction information is a key data point in traffic monitoring and management, which provides detailed information about the driving direction of the lanes or road segments included in the traffic monitoring scene.

[0051] This embodiment directly interacts with the urban traffic management system to obtain the traffic flow direction information of the traffic monitoring scene.

[0052] A500: According to the traffic flow direction information and a plurality of regional space identifiers, the plurality of anomaly detection results are associated and verified to obtain a traffic anomaly tracking object, wherein the traffic anomaly tracking object has an identification characteristic label.

[0053] In one embodiment, according to the traffic flow direction information and a plurality of regional space identifiers, the plurality of anomaly detection results are associated and verified to obtain a traffic anomaly tracking object, wherein the traffic anomaly tracking object has an identification characteristic label, and the method provided in the present application further comprises:

[0054] According to the traffic flow direction information, the plurality of regional space identifiers are updated to obtain a plurality of regional update identifiers; according to the plurality of regional update identifiers, the plurality of regional traffic video streams are divided into M groups of verification traffic video streams; according to the M groups of verification traffic video streams, the plurality of anomaly detection results are divided into M groups of anomaly detection results; based on an anomaly verification feature set, the M groups of anomaly detection results are subjected to intra-group anomaly consistency verification to obtain the traffic anomaly tracking object; and the traffic anomaly tracking object is subjected to feature collection to obtain the identification characteristic label.

[0055] Specifically, in the present embodiment, according to the obtained traffic flow direction information, the plurality of regional space identifiers obtained through video frame segmentation are updated to include the plurality of regional update identifiers reflecting the current traffic flow direction of each regional space identifier corresponding traffic section area, such as a regional update identifier newly added "left-turn traffic flow area" in the corresponding regional space identifier.

[0056] Based on the plurality of regional update identifiers, the plurality of regional space identifiers are divided into different groups, and each group of regional space identifiers has the same traffic flow direction feature. Further, according to the group division of the plurality of regional update identifiers, the plurality of regional traffic video streams are mapped and divided into M groups of verification traffic video streams. Further, according to the M groups of verification traffic video streams, the plurality of anomaly detection results are divided into M groups of anomaly detection results.

[0057] Further, according to the row number and column number of a plurality of regional space identifiers in each group of regional space identifiers, it is determined which regions are adjacent in physical layout to perform adjacent splicing of each group of verification traffic video streams, and based on the adjacent splicing, it is determined whether the adjacent anomaly detection results of the M groups of anomaly detection results are consistent. If consistent, it is proved that the violation object passes through two adjacent regions at the same time in the driving process, and the violation behavior is in a continuous state. Such consistency indicates that the same violation object continuously exhibits the same violation behavior in different time periods or different monitoring regions, which helps to confirm the authenticity and continuity of the violation behavior.

[0058] The M groups of anomaly detection results are subjected to intra-group anomaly consistency verification based on the anomaly verification feature set. If multiple anomaly detection results in a group show high consistency on the anomaly verification feature set, it can be determined that these results point to the same traffic anomaly tracking object.

[0059] The M groups of anomaly detection results are subjected to intra-group anomaly consistency verification, which can obtain M traffic anomaly tracking objects. However, since each traffic anomaly tracking object has the same way of tracking and identifying irregular behavior by scheduling a sighting mirror camera, this embodiment assumes that there is only one traffic anomaly tracking object.

[0060] The feature of the traffic anomaly tracking object is collected, that is, the corresponding anomaly verification feature set collected in the early stage is directly extracted as the identification feature label. The identification feature label includes the vehicle model, color, brand, license plate number, speed, and direction data of the traffic anomaly tracking object. The data for subsequent actual application includes the vehicle model, color, brand, and license plate number.

[0061] This embodiment realizes rapid identification and positioning of irregular objects based on intra-group anomaly consistency verification of the anomaly verification feature set, realizes more accurate identification and tracking of traffic anomaly behavior, and provides a tracking target to improve the intelligent and automated level of traffic management.

[0062] A600: activating and scheduling a sighting mirror camera, which locates and tracks the traffic anomaly tracking object in the traffic monitoring scene according to the identification feature label, and obtains real-time tracking data.

[0063] In one embodiment, the sighting mirror camera is activated and scheduled to locate and track the traffic anomaly tracking object in the traffic monitoring scene according to the identification feature label, and obtain real-time tracking data. The method provided in this application further includes: the sighting mirror camera locates the traffic anomaly tracking object in the traffic monitoring scene according to the identification feature label; the sighting mirror camera performs traffic data tracking and collection on the traffic anomaly tracking object to obtain the real-time tracking data; and the real-time tracking data includes a front vehicle interval distance sequence, a driving speed sequence, and a lane changing frequency parameter.

[0064] Specifically, in this embodiment, after the traffic anomaly tracking object with irregularities is located in the early stage, the sighting mirror camera is scheduled to perform targeted tracking and driving state data collection on the traffic anomaly tracking object. The sighting mirror camera is a monitoring camera equipped with an adjustable sighting mirror, which can accurately locate and track target vehicles.

[0065] Based on the foregoing, the identification feature label contains the identification information of the vehicle, such as the license plate number, vehicle type, color, etc., which is used to quickly locate and identify the target vehicle in a complex traffic monitoring scene.

[0066] The scope camera locates the traffic anomaly tracking object in the traffic monitoring scene according to the identification feature label using image matching, pattern recognition and other technologies. Once the target vehicle is located, the scope camera starts to perform traffic data tracking collection on the traffic anomaly tracking object to obtain the real-time tracking data, which includes the front vehicle interval distance sequence, the driving speed sequence, and the lane changing frequency parameter.

[0067] The front vehicle interval distance sequence is the continuously recorded distance change between the traffic anomaly tracking object and the vehicle in front, to monitor whether the traffic anomaly tracking object maintains a safe distance during driving. The driving speed sequence is the real-time recorded driving speed of the traffic anomaly tracking object, which is used to judge whether there is speeding or frequent acceleration behavior. The lane changing frequency parameter is the number of lane changes of the traffic anomaly tracking object within a certain time (3 minutes) obtained by tracking, to identify whether there is frequent lane changing or illegal lane changing behavior.

[0068] A700: According to the real-time tracking data, locate the traffic regulation space and perform temporary traffic control in the traffic regulation space.

[0069] In one embodiment, according to the real-time tracking data, the traffic regulation space is located, and temporary traffic control is performed in the traffic regulation space. The method step A700 provided by the present application further comprises:

[0070] According to the identification feature label, a historical violation information sequence is called. According to the real-time tracking data and the historical violation information sequence, a real-time risk coefficient is analyzed and obtained. According to the real-time risk coefficient, traffic flow information and traffic monitoring scene positioning, the traffic regulation space is obtained. Temporary traffic control is performed in the traffic regulation space.

[0071] In one embodiment, according to the real-time tracking data and the historical violation information sequence, a real-time risk coefficient is analyzed and obtained. The method step A700 provided by the present application further comprises:

[0072] A pre-constructed traffic risk calculation function is provided, which is as follows:

[0073] ;

[0074] Wherein, is the traffic risk coefficient, is the measured front vehicle interval distance in the front vehicle interval distance sequence, is the traffic risk coefficient, is the average of the front vehicle spacing distance, is the measured driving speed in the th of the driving speed sequence, is the average of the driving speed, is the front vehicle spacing distance measurement frequency, is the driving speed measurement frequency, is the lane changing frequency parameter; the real-time tracking data is substituted into the traffic risk calculation function to obtain a traffic risk coefficient; a violation monitoring window is preset, and the historical violation information sequence is divided based on the violation monitoring window to calculate a violation frequency coefficient; a first weight is assigned to the violation frequency, a second weight is assigned to the traffic risk, and the traffic risk coefficient and the violation frequency coefficient are weighted calculated based on the first weight and the second weight to obtain the real-time risk coefficient.

[0075] In one embodiment, the traffic regulation space is obtained according to the real-time risk coefficient, traffic flow direction information, and traffic monitoring scene positioning. The method provided in the present application further includes the following steps A700:

[0076] A risk level number table is preset, the real-time risk coefficient is used to traverse the risk level number table to obtain a real-time risk level; a standard regulation space number table is matched according to the real-time risk level to obtain a regulation space constraint; a regulation space vector is generated according to the traffic flow direction information; the traffic monitoring scene is taken as a regulation starting point, and the regulation space constraint is extended according to the regulation space vector to obtain the traffic regulation space.

[0077] Specifically, in the present embodiment, according to the license plate information in the identification feature label, the traffic police department interacts to obtain the historical violation information sequence of the multiple time information in which the traffic anomaly tracking object is identified and determined as a traffic violation multiple times in the past.

[0078] A real-time risk coefficient is obtained according to the real-time tracking data and the historical violation information sequence analysis, and the real-time risk coefficient quantifies the potential danger degree of the traffic anomaly tracking object to other vehicles / pedestrians in the current driving scene.

[0079] The method for quantitatively obtaining the real-time risk coefficient is as follows:

[0080] A traffic risk calculation function is pre-constructed, and the traffic risk calculation function is as follows:

[0081] ;

[0082] wherein, is a traffic risk coefficient, is the measured front vehicle spacing distance in the th of the front vehicle spacing distance sequence, is a front vehicle spacing distance mean value, is a driving speed sequence, is a driving speed measured in the is a driving speed mean value, is a front vehicle spacing distance measurement frequency, is a driving speed measurement frequency, is a lane changing frequency parameter; the real-time tracking data is substituted into the traffic risk calculation function to obtain a traffic risk coefficient, which is a numerical value quantifying the potential danger of the traffic anomaly tracking object to other vehicles / pedestrians in the current driving scene.

[0083] A predetermined violation monitoring window, for example, 3 months, is divided into the historical violation information sequence based on the violation monitoring window to obtain a plurality of violation times in three months, and then the average is calculated to obtain the violation frequency coefficient representing the average frequency of traffic violation behavior every three months.

[0084] The first weight is assigned to the violation frequency, and the second weight is assigned to the traffic risk, and the traffic risk coefficient and the violation frequency coefficient are weighted and calculated based on the first weight and the second weight to obtain the real-time risk coefficient.

[0085] A risk level number table is preset, which includes a risk coefficient interval corresponding to each risk level. The real-time risk coefficient is used to traverse the risk level number table to obtain the risk coefficient interval into which the real-time risk coefficient falls, and then the risk level of the risk coefficient interval is taken as the real-time risk level. It should be understood that the risk level-risk coefficient interval of the risk level number table in the embodiment is set based on actual traffic control needs, and the embodiment does not set the table value of the risk level number table.

[0086] The adjustment space is a certain length of road for authorized traffic management personnel (such as volunteers) to intervene to safely and effectively interrupt the violation behavior. The road length of the adjustment space is associated with the risk level of the violation behavior. The higher the risk level, the more serious the consequences that the violation behavior may cause, and therefore a shorter adjustment space is needed to respond quickly.

[0087] The embodiment pre-constructs a standard adjustment space number table for quick matching and determination of the adjustment space length under a certain risk level. The standard adjustment space number table records a plurality of sample adjustment spaces corresponding to a plurality of sample risk levels in the risk level number table. According to the real-time risk level, the standard adjustment space number table is matched to obtain the adjustment space constraint.

[0088] According to the traffic flow direction information, a regulation space vector is generated, which is used to direct the road direction of the regulation space positioned with the traffic monitoring scene as a regulation starting point, so as to ensure that the authorized traffic management personnel can effectively interrupt the violation behavior.

[0089] According to the regulation space vector, the regulation space constraint is extended with the traffic monitoring scene as a regulation starting point, and the traffic regulation space is obtained, in which the traffic temporary control measures are performed, such as setting up roadblocks, changing signal lights, guiding vehicles to detour, etc.

[0090] The embodiment achieves comprehensive monitoring and management of urban traffic violation behaviors, effectively improves the intelligent level of traffic monitoring, and guarantees the safety and efficiency of road traffic.

[0091] Embodiment two

[0092] Based on the same inventive concept as the data processing-based scope finder target recognition method in the foregoing embodiments, as shown in the following Figure 3 The application provides a data processing-based scope finder target recognition system, wherein the system comprises:

[0093] A traffic video acquisition unit 11 is used to interact with a high-position video camera of a traffic monitoring scene, and obtain a real-time traffic video stream.

[0094] A data segmentation execution unit 12 is used to preset an intra-frame partition ratio, and segment the real-time traffic video stream according to the intra-frame partition ratio, to obtain a plurality of regional traffic video streams, wherein the plurality of regional traffic video streams are specifically a plurality of regional space identifiers.

[0095] An anomaly detection execution unit 13 is used to pre-construct a traffic anomaly recognition model, synchronize the plurality of regional traffic video streams to the traffic anomaly recognition model to perform anomaly detection, and obtain a plurality of anomaly detection results.

[0096] A traffic flow direction obtaining unit 14 is used to interactively obtain traffic flow direction information of the traffic monitoring scene.

[0097] An association verification execution unit 15 is used to perform association verification on the plurality of anomaly detection results according to the traffic flow direction information and the plurality of regional space identifiers, to obtain a traffic anomaly tracking object, wherein the traffic anomaly tracking object has an identification feature label.

[0098] A real-time tracking execution unit 16 is used to activate and schedule a scope finder camera, and the scope finder camera is positioned and tracks the traffic anomaly tracking object in the traffic monitoring scene according to the identification feature label, to obtain real-time tracking data.

[0099] The adjusting space determining unit 17 is configured to locate a traffic adjusting space according to the real-time tracking data, and perform temporary control of the traffic in the traffic adjusting space.

[0100] In one embodiment, the anomaly detection executing unit 13 further comprises:

[0101] presetting a sample coverage condition, collecting data according to the sample coverage condition, and obtaining a plurality of sample traffic video streams; converting the plurality of sample traffic video streams into a plurality of sample video image frame sequences; performing object-level labeling on the plurality of sample video image frame sequences to obtain a plurality of sample violation identification sequences; calling a standard CNN model from an operator library, and executing real-time performance optimization of the standard CNN model by taking the plurality of sample video image frame sequences and the plurality of sample violation identification sequences as training data to obtain a standard traffic anomaly recognition branch; copying K standard traffic anomaly recognition branches obtained, and connecting the K standard traffic anomaly recognition branches in parallel to generate the traffic anomaly recognition model, wherein K is a positive integer greater than 30.

[0102] In one embodiment, the anomaly detection executing unit 13 further comprises:

[0103] randomly activating a plurality of standard traffic anomaly recognition branches in the K standard traffic anomaly recognition branches of the traffic anomaly recognition model according to the plurality of regional traffic video streams; converting the plurality of regional traffic video streams into a plurality of regional image frame sequences; synchronously inputting the plurality of regional image frame sequences as input data to the plurality of standard traffic anomaly recognition branches to execute anomaly detection, and obtaining a plurality of traffic anomaly recognition results; presetting a set of verification feature indicators, and performing feature extraction on the plurality of traffic anomaly recognition results by taking the set of verification feature indicators to obtain a plurality of anomaly verification feature sets; and mapping the plurality of anomaly verification feature sets to identify the plurality of traffic anomaly recognition results to obtain the plurality of anomaly detection results.

[0104] In one embodiment, the correlation verification executing unit 15 further comprises:

[0105] updating a traffic vector identifier according to the traffic flow direction information to obtain a plurality of regional update identifiers; dividing the plurality of regional traffic video streams into M groups of verification traffic video streams according to the plurality of regional update identifiers; dividing the plurality of anomaly detection results into M groups of anomaly detection results according to the M groups of verification traffic video streams; performing intra-group anomaly consistency verification on the M groups of anomaly detection results based on anomaly verification feature sets to obtain the traffic anomaly tracking object; and collecting features of the traffic anomaly tracking object to obtain the identification feature label.

[0106] In one embodiment, the real-time tracking executing unit 16 further comprises:

[0107] The aiming scope camera locates the traffic anomaly tracking object in the traffic monitoring scene based on the identification feature tag; the aiming scope camera performs traffic data tracking and collection on the traffic anomaly tracking object to obtain the real-time tracking data; wherein, the real-time tracking data includes the preceding vehicle interval distance sequence, driving speed sequence, and lane change frequency parameters.

[0108] In one embodiment, the adjustment space determination unit 17 further includes:

[0109] Historical violation information sequences are obtained by calling the identified feature tags; a real-time risk coefficient is obtained by analyzing the real-time tracking data and the historical violation information sequences; the traffic regulation space is obtained by analyzing the real-time risk coefficient, traffic flow information, and traffic monitoring scene location; and temporary traffic control is performed in the traffic regulation space.

[0110] In one embodiment, the adjustment space determination unit 17 further includes:

[0111] A pre-constructed traffic risk calculation function is provided, as follows:

[0112] ;

[0113] in, For traffic risk coefficient, The first vehicle in the sequence of preceding vehicle interval distances The measured distance between the vehicles in front. This is the average distance between vehicles in front. The first in the driving speed sequence The measured driving speed The average driving speed Frequency of measuring the distance between vehicles in front. Frequency of speed measurement The frequency of lane changes is a parameter; the real-time tracking data is substituted into the traffic risk calculation function to obtain the traffic risk coefficient; a violation monitoring window is preset, and the historical violation information sequence is divided based on the violation monitoring window to calculate the violation frequency coefficient; a first weight is assigned to the violation frequency, a second weight is assigned to the traffic risk, and the traffic risk coefficient and the violation frequency coefficient are weighted and calculated based on the first weight and the second weight to obtain the real-time risk coefficient.

[0114] In one embodiment, the adjustment space determination unit 17 further includes:

[0115] A preset risk level number table is adopted to traverse the risk level number table by using the real-time risk coefficient to obtain a real-time risk level; a standard adjustment space number table is obtained according to the real-time risk level; a standard adjustment space vector is generated according to the traffic flow direction information; and the traffic monitoring scene is taken as a standard adjustment starting point, the standard adjustment space constraint is extended according to the standard adjustment space vector to obtain the traffic adjustment space.

[0116] Any one of the above methods or steps can be stored as computer instructions or programs in various types of computer memories, and the computer instructions or programs are recognized by various types of computer processors to realize any one of the above methods or steps.

[0117] Based on the above specific embodiments of the present application, any improvement and modification of the present application made by those skilled in the art without departing from the principles of the present application shall fall within the scope of the patent protection of the present application.

Claims

1. A data processing based scope target identification method, characterized in that, The method comprises: interacting with a high-position video camera of a traffic monitoring scene to obtain a real-time traffic video stream; presetting an intra-frame partition ratio and partitioning the real-time traffic video stream according to the intra-frame partition ratio to obtain a plurality of regional traffic video streams, wherein the plurality of regional traffic video streams correspond to unique regional spatial identifiers respectively; preconstructing a traffic anomaly recognition model, synchronizing the plurality of regional traffic video streams to the traffic anomaly recognition model to perform anomaly detection, and obtaining a plurality of anomaly detection results; interacting to obtain traffic flow direction information of the traffic monitoring scene; associating and checking the plurality of anomaly detection results according to the traffic flow direction information and the plurality of regional spatial identifiers to obtain a traffic anomaly tracking object, wherein the traffic anomaly tracking object has an identification characteristic label; activating and dispatching a scope camera, the scope camera locating and tracking the traffic anomaly tracking object in the traffic monitoring scene according to the identification characteristic label to obtain real-time tracking data; locating a traffic regulation space according to the real-time tracking data, and performing temporary regulation and control of traffic in the traffic regulation space.

2. The data processing based scope target identification method of claim 1, wherein, The method further comprises: presetting a sample coverage condition and collecting data according to the sample coverage condition to obtain a plurality of sample traffic video streams; converting the plurality of sample traffic video streams into a plurality of sample video image frame sequences; performing object-level labeling on the plurality of sample video image frame sequences to obtain a plurality of sample violation identification sequences; calling a standard CNN model in an operator library, and taking the plurality of sample video image frame sequences and the plurality of sample violation identification sequences as training data to perform real-time performance optimization of the standard CNN model to obtain a standard traffic anomaly recognition branch; copying K standard traffic anomaly recognition branches to generate the traffic anomaly recognition model in parallel, wherein K is a positive integer greater than 30.

3. The data processing based scope target identification method of claim 2, wherein, Synchronizing the plurality of regional traffic video streams to the traffic anomaly recognition model to perform anomaly detection to obtain a plurality of anomaly detection results, the method further comprises: randomly activating a plurality of standard traffic anomaly recognition branches in the K standard traffic anomaly recognition branches of the traffic anomaly recognition model according to the plurality of regional traffic video streams; converting the plurality of regional traffic video streams into a plurality of regional image frame sequences; synchronizing the plurality of regional image frame sequences as input data to the plurality of standard traffic anomaly recognition branches to perform anomaly detection to obtain a plurality of traffic anomaly recognition results; presetting a set of verification feature indicators and performing feature extraction on the plurality of traffic anomaly recognition results according to the set of verification feature indicators to obtain a plurality of anomaly verification feature sets; mapping the plurality of anomaly verification feature sets to identify the plurality of traffic anomaly recognition results to obtain the plurality of anomaly detection results.

4. The data processing based scope target identification method of claim 3 wherein, According to the traffic flow direction information and the plurality of regional spatial identifiers, the plurality of anomaly detection results are associated and checked to obtain a traffic anomaly tracking object, wherein the traffic anomaly tracking object has an identification characteristic label, and the method further comprises: According to the traffic flow direction information, traffic vector identification is updated for the plurality of regional spatial identifiers, and a plurality of regional update identifiers are obtained; According to the plurality of regional update identifiers, the plurality of regional traffic video streams are divided into M groups of check traffic video streams; According to the M groups of check traffic video streams, the plurality of anomaly detection results are divided into M groups of anomaly detection results; Based on the anomaly check feature set, intra-group anomaly consistency check is performed on the M groups of anomaly detection results, and the traffic anomaly tracking object is obtained; Feature collection is performed on the traffic anomaly tracking object, and the identification feature label is obtained.

5. The data processing based scope target identification method of claim 1, wherein, The scope camera is activated and scheduled, and the scope camera locates and tracks the traffic anomaly tracking object in the traffic monitoring scene according to the identification feature label, and real-time tracking data is obtained. The method further comprises: The scope camera locates the traffic anomaly tracking object in the traffic monitoring scene according to the identification feature label; The scope camera performs traffic data tracking collection on the traffic anomaly tracking object, and obtains the real-time tracking data; The real-time tracking data includes a preceding vehicle interval distance sequence, a travel speed sequence, and a lane changing frequency parameter.

6. The data processing based scope target identification method of claim 5, wherein, According to the real-time tracking data, a traffic regulation space is located, and temporary traffic control is performed in the traffic regulation space. The method further comprises: According to the identification feature label, a historical violation information sequence is obtained; According to the real-time tracking data and the historical violation information sequence, a real-time risk coefficient is analyzed and obtained; According to the real-time risk coefficient, traffic flow direction information, and traffic monitoring scene positioning, the traffic regulation space is obtained; Temporary traffic control is performed in the traffic regulation space.

7. The data processing based scope target identification method of claim 6 wherein, According to the real-time tracking data and the historical violation information sequence, a real-time risk coefficient is analyzed and obtained. The method further comprises: A traffic risk calculation function is pre-constructed, and the traffic risk calculation function is as follows: ； wherein, is a traffic risk coefficient, is the measured preceding vehicle separation distance in the sequence of preceding vehicle separation distances, is the measured preceding vehicle separation distance in the sequence of preceding vehicle separation distances, is the mean preceding vehicle separation distance, is the measured driving speed in the sequence of driving speeds, is the measured driving speed in the sequence of driving speeds, is the mean driving speed, is the preceding vehicle separation distance measurement frequency, is the driving speed measurement frequency, is the lane change frequency parameter; The real-time tracking data is substituted into the traffic risk calculation function to obtain a traffic risk coefficient; A violation monitoring window is pre-set, and the historical violation information sequence is divided based on the violation monitoring window to calculate a violation frequency coefficient; A first weight is assigned to the violation frequency, a second weight is assigned to the traffic risk, and the traffic risk coefficient and the violation frequency coefficient are weighted and calculated based on the first weight and the second weight to obtain the real-time risk coefficient.

8. The data processing based scope target identification method of claim 6 wherein, According to the real-time risk coefficient, traffic flow direction information, and traffic monitoring scene positioning, the traffic regulation space is obtained. The method further comprises: A risk level number table is pre-set, and the real-time risk coefficient is used to traverse the risk level number table to obtain a real-time risk level; According to the real-time risk level, a standard adjustment space number table is matched to obtain a regulation space constraint; According to the traffic flow direction information, a regulation space vector is generated; Taking the traffic monitoring scene as an adjustment starting point, the regulation space constraint is extended according to the regulation space vector to obtain the traffic regulation space.

9. A data processing based scope target identification system characterized by, A system for implementing the data processing-based scope target identification method of any one of claims 1 to 8, the system comprising: A traffic video acquisition unit is configured to acquire a high-position video camera of an interactive traffic monitoring scene to obtain a real-time traffic video stream. A data segmentation execution unit is configured to preset an intra-frame partition ratio and segment the real-time traffic video stream according to the intra-frame partition ratio to obtain a plurality of regional traffic video streams, wherein the plurality of regional traffic video streams correspond to unique regional space identifiers respectively. An anomaly detection execution unit is configured to pre-construct a traffic anomaly recognition model, synchronize the plurality of regional traffic video streams to the traffic anomaly recognition model to perform anomaly detection, and obtain a plurality of anomaly detection results. A traffic flow direction acquisition unit is configured to interactively obtain traffic flow direction information of the traffic monitoring scene. An association verification execution unit is configured to perform association verification on the plurality of anomaly detection results according to the traffic flow direction information and the plurality of regional space identifiers to obtain a traffic anomaly tracking object, wherein the traffic anomaly tracking object has an identification feature label. A real-time tracking execution unit is configured to activate and dispatch a scope camera, and the scope camera is configured to locate and track the traffic anomaly tracking object in the traffic monitoring scene according to the identification feature label to obtain real-time tracking data. An adjustment space determination unit is configured to locate a traffic adjustment space according to the real-time tracking data, and perform temporary traffic control in the traffic adjustment space.

Images