A method and system for identifying and detecting illegally parked hazardous chemical vehicles in service areas

Abstract

This invention belongs to the field of vehicle safety monitoring technology and relates to a method and system for identifying and detecting illegally parked hazardous chemical vehicles in service areas. It uses a mobile close-up camera unit to perform close-up photography on suspected targets that meet the basic characteristics of hazardous chemical vehicles or areas to be analyzed where the image quality confidence level is insufficient, to obtain local close-up image data. It simultaneously identifies hazardous chemical markings and analyzes the three-dimensional structure of the tank body, jointly confirming the vehicle's hazardous chemical attributes. A tracking file is established for confirmed vehicles, and the location trajectory is updated frequently, recording the cumulative parking time. It is correlated with a pre-constructed spatial risk zoning map to quantify dynamic illegal parking risks and triggers an early warning when the risk exceeds a threshold. This avoids the limitations of single monitoring perspectives and single feature accuracy. Secondary confirmation through local close-up images improves the accuracy and real-time performance of illegal parking identification, enhances the comprehensiveness and timeliness of supervision, and significantly improves the accuracy and reliability of hazardous chemical vehicle attribute identification.

Description

Technical Field

[0001] This invention belongs to the field of vehicle safety monitoring technology, and specifically relates to a method and system for identifying and detecting illegally parked hazardous chemical vehicles in service areas. Background Technology

[0002] With the rapid development of the chemical industry and the continuous growth in energy demand, the scale of hazardous chemical transportation has been increasing year by year. Hazardous chemicals are flammable, explosive, and corrosive. Once an accident occurs during their transportation, it can easily cause serious personal injury, environmental pollution, and property damage. Therefore, transportation safety has become a key area of ​​public safety management.

[0003] Highway service areas, as crucial stops for hazardous chemical transport vehicles and rest stops for passengers, not only provide basic services but also fulfill safety management responsibilities. However, some hazardous chemical transport vehicle drivers, driven by convenience or a sense of impunity, frequently park illegally in non-designated areas within service areas, creating significant safety hazards. Traditional regulatory methods are no longer sufficient to effectively address such dangerous behavior.

[0004] Currently, monitoring of illegally parked hazardous chemical vehicles in service areas mainly relies on fixed-layout surveillance systems. This involves manually analyzing real-time footage or video playback based on experience at pre-set monitoring points in a limited area, resulting in low efficiency and a high risk of missed detections. While technical solutions using image recognition algorithms for assisted detection exist, primarily analyzing static features such as vehicle outlines and color for preliminary identification, achieving a degree of automation, these methods still have significant limitations:

[0005] First, the service area's panoramic surveillance cameras have a wide coverage area and a long shooting distance, resulting in limited image resolution. This makes it difficult to clearly capture key details such as the markings of hazardous chemical vehicles and the shape of the tanks, leading to low recognition confidence.

[0006] Secondly, existing methods are mostly limited to single-frame image or single-viewpoint analysis, lacking continuous tracking and dynamic risk assessment of vehicle behavior in the time and space dimensions. They cannot effectively identify illegal parking status and its risk evolution trend, resulting in a lag in regulatory response and significant room for improvement in overall prevention and control capabilities. Summary of the Invention

[0007] In view of this, in order to solve the problems mentioned in the background technology, a method and system for identifying and detecting illegally parked hazardous chemical vehicles in service areas is proposed.

[0008] The technical solution adopted by the present invention to solve its technical problem is as follows: Firstly, the present invention provides a method for identifying and detecting illegally parked hazardous chemical vehicles in service areas, comprising: continuously scanning the monitoring range of the service area through a fixed panoramic camera unit to acquire a global scene image sequence in real time.

[0009] The global scene image sequence is divided into several regions to be analyzed. If any region to be analyzed initially identifies a suspected target with basic characteristics of a hazardous chemical vehicle or the image quality judgment is not confident enough, a close-range acquisition command containing the initial three-dimensional position information of the region to be analyzed is generated and issued.

[0010] The camera unit is deployed in the area corresponding to the initial three-dimensional position information, and its gimbal is controlled to adjust its attitude and focus to align with the target, so as to take close-up pictures and obtain local close-up image data.

[0011] Multimodal parallel processing was performed on local close-up image data, and the hazardous chemical attributes of the vehicle were confirmed based on the joint results of hazardous chemical mark recognition and tank body three-dimensional structure analysis.

[0012] Establish tracking files for confirmed vehicles, frequently update their location trajectories and record cumulative parking time, quantify dynamic illegal parking risks by associating them with pre-built spatial risk zoning maps, and trigger warnings when risks exceed thresholds.

[0013] Secondly, the present invention provides a service area hazardous chemical vehicle illegal parking identification and detection system, including: an image acquisition module, an instruction generation module, a close-up shooting module, an attribute confirmation module, and a risk quantification module.

[0014] The image acquisition module is connected to the instruction generation module, the instruction generation module is connected to the close-up shooting module, the close-up shooting module is connected to the attribute confirmation module, and the attribute confirmation module is connected to the risk quantification module.

[0015] The image acquisition module continuously scans the service area monitoring range using a fixed panoramic camera unit to acquire global scene image sequences in real time.

[0016] The instruction generation module divides the global scene image sequence into several regions to be analyzed. If any region to be analyzed is initially identified as a suspected target with basic characteristics of a hazardous chemical vehicle or the image quality judgment confidence is insufficient, it generates and issues a close-range acquisition instruction containing the initial three-dimensional position information of the region to be analyzed.

[0017] The close-up shooting module schedules and deploys close-up camera units in the area corresponding to the initial three-dimensional position information, controls their gimbals to adjust their attitude and focus to align with the target, and performs close-up shooting to obtain local close-up image data.

[0018] The attribute confirmation module performs multimodal parallel processing on local close-up image data and confirms the hazardous chemical attributes of the vehicle based on the joint results of hazardous chemical mark recognition and tank body three-dimensional structure analysis.

[0019] The risk quantification module establishes a tracking file for confirmed vehicles, updates their location trajectory frequently and records the cumulative parking time, associates it with a pre-built spatial risk zoning map to quantify dynamic illegal parking risks, and triggers an alert when the risk exceeds a threshold.

[0020] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects:

[0021] (1) This invention uses coordinated scheduling of close-up camera units to perform close-up shooting on suspected targets that meet the basic characteristics of hazardous chemical vehicles or areas to be analyzed where the image quality judgment confidence is insufficient. By utilizing pre-deployed high-definition camera resources, it avoids the problems of a single fixed monitoring perspective and limited accuracy. Through secondary confirmation of local close-up images, it improves the accuracy and reliability of illegal parking identification, thereby enhancing the comprehensiveness and timeliness of supervision.

[0022] (2) The present invention confirms the hazardous chemical attributes of vehicles by combining the results of hazardous chemical mark identification and tank body three-dimensional structure analysis, realizes multi-dimensional cross-verification, avoids misjudgment of single features, and significantly improves the accuracy and reliability of hazardous chemical vehicle attribute identification.

[0023] (3) This invention combines spatiotemporal information to output the risk assessment results of illegal parking, providing comprehensive data support for managers, helping to accurately predict potential risks, and thus quickly improving emergency response efficiency. Attached Figure Description

[0024] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0025] Figure 1 This is a flowchart illustrating the implementation of the method provided in the first embodiment of the present invention.

[0026] Figure 2 This is a logical diagram illustrating the process of identifying and issuing early warnings for dynamic illegal parking risks of vehicles in the first embodiment of the present invention.

[0027] Figure 3 This is a module connection block diagram provided for the second embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Example 1

[0030] Please see Figure 1 As shown, the first embodiment of the present invention provides a method for identifying and detecting illegally parked hazardous chemical vehicles in service areas, including: S11. Performing continuous wide-area scanning of the service area monitoring range using a fixed panoramic camera unit to acquire a global scene image sequence in real time.

[0031] In a preferred embodiment of the present invention, the real-time acquisition process of the global scene image sequence is as follows:

[0032] The fixed panoramic camera unit performs image acquisition at a preset frame rate.

[0033] The original images synchronously acquired by multiple wide-angle lenses in a fixed panoramic camera unit are stitched and fused in real time. The stitching and fusion process is performed in a set order based on the installation position and field of view configuration of each wide-angle lens. The requirement is that the field of view of a single frame of global scene image completely covers the vehicle activity area within the service area, including at least driving lanes, parking spaces and emergency lanes.

[0034] It should be noted that the fixed panoramic camera unit is installed on the roof or pillar of the service area building and is rigidly connected to the building structure through brackets. The specific process of stitching and blending its multiple wide-angle lenses is as follows: based on the spatial distribution of the wide-angle lens installation positions and the degree of overlap of the field of view, the wide-angle lenses are classified into primary lenses, secondary lenses, or edge lenses according to the following classification process:

[0035] i. Wide-angle lenses that cover a vehicle activity area in the acquired image in proportion exceeding a preset main field of view coverage threshold are identified as the main lens type.

[0036] ii. For the remaining lenses, if the overlap of their field of view with that of the adjacent lens is higher than the preset minimum effective overlap, they are further classified according to their relative distance from the main lens: if the installation distance between a lens and the main lens does not exceed the preset maximum adjacent distance, it is classified as a secondary lens; otherwise, it is classified as an edge lens. The priority order of lens arrangement follows the spatial logic from the center of the service area to the periphery or from the entrance to the exit.

[0037] The main lens is set as the highest priority, followed by the secondary lens, and the edge lens is set as the lowest priority. All wide-angle lens images are sorted according to this priority order.

[0038] For images acquired by adjacent priority lenses, pre-placed fixed scene feature points in their overlapping areas are extracted as matching references. These feature points include, but are not limited to, lane line turning points, parking space corner points, and pillar reference points. The spatial transformation matrix is ​​calculated in combination with the lens installation position parameters, and the secondary and edge lens images are mapped to the main lens coordinate system to achieve geometric alignment between multi-source images.

[0039] The image is stitched frame by frame according to image priority. A fusion strategy based on linear weight gradient is applied to overlapping areas: pixels closer to the stitching seam have lower weights, while pixels further away from the seam have higher weights. Non-overlapping areas directly adopt the original image data to initially synthesize a global scene image.

[0040] The integrity of the field of view of the initially stitched image is verified by using an edge detection algorithm to determine whether it completely covers all driving lanes, parking spaces, and emergency lanes. If any defects are found, edge pixel interpolation is performed based on the lens's field of view parameters to complete the image, ultimately outputting a seamless and complete single-frame global scene image.

[0041] S12. Divide the global scene image sequence into several regions to be analyzed. If any region to be analyzed initially identifies a suspected target with basic characteristics of a hazardous chemical vehicle or the image quality judgment confidence is insufficient, generate and issue a close-range acquisition command containing the initial three-dimensional position information of the region to be analyzed.

[0042] In a preferred embodiment of the present invention, the process of dividing the plurality of regions to be analyzed includes:

[0043] The global scene image sequence is subjected to frame-by-frame vehicle entity detection, and the minimum bounding rectangle of all vehicle entities in each frame is located.

[0044] It should be noted that in the process of vehicle entity detection frame by frame in the above global scene image sequence, deep learning-based target detection algorithms, such as YOLO, SSD or Faster R-CNN, can be used. These methods are mature existing technologies and will not be elaborated here.

[0045] Spatiotemporal correlation analysis is performed on the minimum bounding rectangle of all frames in the sequence. A group of rectangles with spatial overlap higher than a preset overlap threshold are associated as the same region. Multiple vehicle bounding rectangles belonging to the same region are merged to generate a corresponding region to be analyzed.

[0046] It should be noted that the aforementioned spatial overlap is specifically obtained by calculating the ratio of the intersection area of ​​the two rectangles to the average area of ​​all the smallest bounding rectangles.

[0047] For an independent rectangle that is not associated with any other rectangle, it is treated as the region to be analyzed, thereby dividing the global scene image sequence into several regions to be analyzed.

[0048] In a preferred embodiment of the present invention, the preliminary identification process of the suspected target with the basic characteristics of a hazardous chemical vehicle includes: extracting the overall contour features and surface texture features of the vehicle entity in the area to be analyzed.

[0049] The overall contour features are compared with a preset typical hazardous chemical vehicle contour library. The similarity index of the vehicle entity contour comparison in the area to be analyzed is quantified to determine whether it conforms to the basic contour characteristics. The overall contour features include at least the elliptical cylindrical contour of the tank body, a specific number of saddle support structures, and the tail geometry for pipeline connection.

[0050] It should be added that the specific quantification process of the vehicle entity contour comparison similarity index in the above-mentioned area to be analyzed is as follows: parameter detection is performed on the components involved in the overall contour features, including but not limited to the major and minor axis ratio of the fitted ellipse, the radius of curvature distribution, and the proportion of tank length to the total length of the vehicle body.

[0051] The saddle support structure parameters include, but are not limited to, the corresponding number of single and double saddles, the distance between adjacent saddles, and the coordinate distance between the connection points of the saddle and the bottom of the tank.

[0052] The tail structure parameters include, but are not limited to, the vertex coordinates of the outline polygon of the tail pipe interface, the height of the protruding structure, and its relative position vector with respect to the end of the tank.

[0053] The system retrieves standard parameter ranges for different vehicle models and standard tank body contours stored in a pre-defined library of typical hazardous chemical vehicle contours. It then calculates the geometric deviation between the physical contour of the vehicle in the analysis area and the standard tank body contours of different vehicle models in the library using Hausdorff distance, converting the result into a value range. The shape matching score is calculated, and the maximum value is retrieved as the basic shape matching index. The deviation rate between the measured values ​​of each parameter and the standard parameter range of different vehicle models is measured, and the deviation rate is substituted into the standard exponential decay function to output the degree of conformity between the measured values ​​of each parameter and the standard parameter range of different vehicle models. Different influence weights are assigned to the tank body, saddle support structure and rear structure according to the structural importance of hazardous chemical vehicles. The comprehensive conformity between the vehicle entity structure parameters of the area to be analyzed and the standard parameter range of different vehicle models in the database is output through linear weighted fusion, and the maximum value is retrieved as the basic structural matching index.

[0054] The sum of the basic morphological matching index and the basic structural matching index is used as the similarity index for the vehicle entity outline comparison in the area to be analyzed. If it is greater than the preset similarity threshold, it is judged to meet the basic outline characteristics.

[0055] Based on surface texture feature analysis, it is determined whether the vehicle entity in the area to be analyzed conforms to the high gloss reflection characteristics of the metal tank or the texture pattern of the anti-corrosion coating.

[0056] It should be noted that the process for determining whether the vehicle entity in the area to be analyzed conforms to the high gloss reflection characteristics of a metal can is as follows: the image is converted to HSV or HSL color space, the luminance component image is extracted, and an adaptive threshold or gradient detection method is used to identify continuous or discrete candidate highlight regions in the image. The candidate highlight regions are filtered based on the area threshold, and only candidate highlight regions with an internal pixel count not less than a preset percentage of the total number of pixels in the image are retained. The distribution concentration, shape regularity, and luminance gradient mean of the retained highlight regions are quantified in parallel and compared with the reasonable range of parameters obtained through statistical learning of a large number of good-quality metal can samples. If all parameters are within their respective preset reasonable ranges, it is determined that the vehicle conforms to the high gloss reflection characteristics of a metal can.

[0057] The process of quantifying the distribution of highlights is as follows: the coordinates of each pixel within the highlight area are obtained and the centroid of the area is calculated by mean value. The distance from each pixel to the centroid of the area is determined by Euclidean distance calculation formula. After normalization, the reciprocal of the standard deviation of the distance is used as the quantified value of the distribution of highlights. The reason for using this as a criterion is that the surface of a good metal can is smooth and the curvature is continuous. Under a directional light source, it will produce specular reflection, forming a concentration. High concentration is a direct spatial distribution characterization of the specular reflection characteristics of the metal can.

[0058] The shape regularity quantification process is as follows: Contour search is performed on the binarized highlight region to obtain its outermost pixel-level contour. Contour fitting is then performed based on the expected shape of the can. The shortest distance from each pixel on the outer contour to the expected fitted contour is obtained. The inverse of the root mean square error of all shortest distances is used as the shape regularity of the highlight region. The reason for using this as a criterion is that high-quality metal cans are precision-stamped, resulting in smooth, regular, and continuous geometric surfaces. The resulting highlight contours will inevitably resemble ideal geometric shapes. High shape regularity is a direct reflection of a metal can with a smooth surface and standard geometric shape.

[0059] The brightness gradient mean quantization process is as follows: the Sobel operator is used to calculate the gradient magnitude of each pixel in the highlight area, and the arithmetic mean of the gradient magnitude of all pixels in the highlight area is used as the brightness gradient mean. The metal can has specular reflection characteristics, and the edge of the highlight area is a region where the intensity of reflected light changes drastically, so a very steep brightness gradient will be generated in the image.

[0060] The process for determining whether the vehicle entity in the area to be analyzed conforms to the texture pattern of the anti-corrosion coating is as follows: the gray-level co-occurrence matrix is ​​used to extract the surface texture features of the vehicle entity, including contrast, energy, entropy value and correlation value.

[0061] The system divides the area into several overlapping intervals with a fixed numerical step size and defines fuzzy categories. Using a membership function, it calculates the membership degree of each parameter in the texture features of the vehicle entity surface in the area to be analyzed, and retrieves all the fuzzy category sets involved. Based on the fuzzy rule library trained by expert knowledge or historical data, it inputs the membership degree of each parameter into the matching fuzzy rule for calculation, and obtains the confidence degree of each triggered rule. The maximum confidence degree is compared with the preset texture compliance threshold. If the comparison relationship is greater than the threshold, it is determined that the texture pattern conforms to the anti-corrosion coating.

[0062] If at least two of the conditions are met in the feature analysis results, the vehicle entity is preliminarily identified as a suspected target with the basic characteristics of a hazardous chemical vehicle.

[0063] It should also be added that the determination process corresponding to insufficient confidence in the image quality of the region to be analyzed is as follows: the image of the region to be analyzed is converted into a grayscale image, and the grayscale image is convolved using the Laplacian operator to obtain the second derivative of each pixel, which reflects the sharpness of the image edge. The second derivatives of all pixels are integrated to obtain the pixel variance of the obtained Laplacian image. If the variance is lower than the preset sharpness threshold, it indicates that the image is blurry and lacks edge information, and can be determined as insufficient sharpness.

[0064] The smooth regions of the image are segmented based on texture complexity, and the gray-level variance of the pixels in the smooth regions is used as the noise intensity. If the noise intensity exceeds the preset permissible threshold for image recognition, it is determined to be strong noise interference.

[0065] If the judgment results show insufficient clarity or strong noise interference, it can be directly determined that the image quality confidence of the area to be analyzed is insufficient.

[0066] S13. Schedule and deploy the close-up camera unit in the area corresponding to the initial three-dimensional position information, control its gimbal to adjust its attitude and focus to align with the target, and perform close-up shooting to obtain local close-up image data.

[0067] It should be noted that the close-up camera units are high-definition camera clusters pre-deployed in key areas within the service area, and their combined field of view provides complete coverage of the vehicle activity area within the service area. Each close-up camera unit is equipped with a high-precision pan-tilt unit and an optical zoom lens, and its position, field of view, and status information are registered in the system database.

[0068] In a preferred embodiment of the present invention, the close-up camera unit scheduling process includes: at the timestamp of receiving the close-up acquisition instruction, matching and selecting a close-up camera unit whose field of view covers the location and is currently idle from a preset close-up camera unit deployment information database according to the initial three-dimensional position information of the area to be analyzed.

[0069] The system controls the proximity camera unit to quickly scan the vehicle entity, drives the gimbal to adjust the pitch and yaw angles, and performs optical zoom. When the proportion of the imaging area of ​​the key feature region on the sensor target surface reaches a preset qualified threshold, multi-frame exposure acquisition is performed.

[0070] Key feature areas include the permanent location area for hazardous chemical labels and the panoramic structural area of ​​the tank body.

[0071] This invention employs a mobile close-up camera unit to perform close-up photography on suspected targets that meet the basic characteristics of hazardous chemical vehicles or areas to be analyzed where the image quality assessment confidence is insufficient. This avoids the problems of a single monitoring perspective and limited accuracy. By using local close-up images for secondary confirmation, the accuracy and real-time performance of illegal parking identification are improved, thereby enhancing the comprehensiveness and timeliness of supervision.

[0072] S14. Perform multimodal parallel processing on local close-up image data, and confirm the hazardous chemical attributes of the vehicle based on the joint results of hazardous chemical mark recognition and tank body three-dimensional structure analysis.

[0073] In a preferred embodiment of the present invention, the hazardous chemical label identification process includes: performing image enhancement and geometric correction preprocessing on local close-up image data.

[0074] Identify and locate hazardous chemical marking elements in preprocessed images, extract visual display information, and structured coded information including UN numbers and hazard class codes.

[0075] Based on a pre-built hazardous materials knowledge graph, the extracted coded information is logically consistent. The verification process includes logical conflict verification between UN number and hazard category code, and between coded information and visual display information.

[0076] It should be added that the aforementioned pre-constructed hazardous materials knowledge graph is based on a structured knowledge system built upon standards for hazardous materials transportation, storage, and labeling. Its core consists of entities, relationships, and constraint rules, as detailed below:

[0077] Entities are the basic units of a knowledge graph, representing key concepts in the field of hazardous materials. They mainly include UN numbers, hazard class codes, and visual display elements, among which visual display elements include at least graphic symbols, color codes, and text symbols.

[0078] Relationships are used to connect entities and express the logical connections between entities. Core relationships include the mapping relationship between UN number and hazard class code, and between hazard class code or UN number and visual display elements. Examples are given below, such as UN1203 → Class 3 flammable liquids, Class 3 flammable liquids → flame graphic + red background + flammable liquid text.

[0079] The constraint rules are logical constraints defined based on the domain specification, meaning that a UN number must correspond to at least one hazard category code and cannot correspond to an irrelevant category.

[0080] Visual display elements must strictly match the corresponding hazard class code or UN number. For example, the graphic of a Class 1 explosive cannot be replaced by a flame graphic.

[0081] Therefore, the steps for verifying the logical conflict between the UN number and the hazard class code are as follows:

[0082] Search the knowledge graph for the set of standard hazard category codes corresponding to the UN number. If the extracted hazard category code is outside the set of standard hazard category codes, or if the UN number has no corresponding hazard category in the knowledge graph, then verify that there is a logical conflict between the UN number and the hazard category code.

[0083] The logical conflict verification process between encoded information and visual display information is as follows:

[0084] Based on the UN number and hazard category code in the encoded information, the corresponding standard visual display element sets are retrieved and integrated. If any visual display element in the visual display information is inconsistent with the standard visual display element set, it is verified that there is a logical conflict between the encoded information and the visual display information.

[0085] Prepare a set of hazardous chemical label samples with completely correct identification information. Select a predetermined proportion of samples and manually modify their UN numbers or hazard class codes to make their mapping relationship conflict with the knowledge graph standard, forming the first type of conflict sample subset. Select the same proportion of samples and manually modify their visual display information to make it conflict with the standard visual mapping of the encoded information, forming the second type of conflict sample subset. Statistically compare the recognition error rate caused by the conflict in the two types of sample sets to clarify the contribution of the logical conflict between the UN number and hazard class code, and between the encoded information and the visual display information, to the recognition accuracy of hazardous chemical labels. Assign reverse confidence values ​​to the two types of conflicts, and the sum of the two is 1. Therefore, if any type of logical conflict exists in the subsequent actual logical consistency verification results, subtract the reverse confidence value from 1 to obtain the first confidence level. If both types of logical conflicts exist, the first confidence level is 0.

[0086] Output the first confidence level of hazardous chemical label identification based on the logical consistency verification results.

[0087] In a preferred embodiment of the present invention, the three-dimensional structure analysis process of the tank body includes: identifying the structural attribute type based on the geometric structure information of the tank body.

[0088] It should be added that the above-mentioned tank body geometric structure information includes: main body shape, shape and size ratio, distribution pattern of outer surface reinforcing ribs, cross-sectional shape of reinforcing ribs and the ratio of their height to the tank body wall thickness, and tank body wall thickness distribution characteristics, including local thickening, uniform wall thickness or gradual wall thickness characteristics.

[0089] The system locates multiple key components on the tank body, analyzes the relative spatial positions, angles, and dimensional ratios between the key components, matches them with the corresponding structural attribute type adaptation rules, and generates a second confidence level for the three-dimensional structural analysis of the tank body.

[0090] It should be noted that the second confidence level mentioned above refers to the ratio of the number of matching rules to the number of rules involved in the structural attribute type adaptation rules.

[0091] It should be added that the process of confirming the vehicle's hazardous chemical attributes based on the combined results of hazardous chemical label recognition and tank body three-dimensional structure analysis is as follows: the sum of the first confidence level of hazardous chemical label recognition and the second confidence level of tank body three-dimensional structure analysis is compared with the preset hazardous chemical attribute recognition confidence threshold. If the comparison relationship is greater than or equal to the threshold, the vehicle's hazardous chemical attributes are confirmed.

[0092] This invention confirms the hazardous chemical attributes of vehicles by combining the results of hazardous chemical marking identification and three-dimensional structural analysis of the tank body, achieving multi-dimensional cross-verification, avoiding misjudgment based on a single feature, and significantly improving the accuracy and reliability of hazardous chemical vehicle attribute identification.

[0093] like Figure 2 As shown in S15, a tracking file is established for the confirmed vehicle, the location trajectory is updated frequently and the cumulative parking time is recorded, the dynamic illegal parking risk is quantified by associating it with the pre-constructed spatial risk zoning map, and an early warning is triggered when the risk exceeds the threshold.

[0094] In a preferred embodiment of the present invention, the pre-construction process of the spatial risk zoning map includes: importing the service area design map and converting it into a digital grid format.

[0095] Based on predefined safety rules, multiple functional zones are divided and marked on the grid map, and each zone is independently configured with a risk weight coefficient and a multi-level time threshold. The multi-level time threshold is used to define the time it takes for a hazardous chemical vehicle to transition from a compliant parking state to a state of warning for illegal parking at various levels. The risk weight coefficient is used to quantify the inherent risk level of each functional zone for hazardous chemical vehicle parking events.

[0096] It should be noted that the aforementioned predefined safety rules refer to a set of normative content formulated based on relevant national and industry safety standards, the actual operational and management needs of service areas, and the unique risk characteristics of hazardous chemical vehicles. This set is used to quantitatively assess the risk levels of different areas, and its specific content covers the following core principles and classification criteria:

[0097] Based on the division of safe distance and protection distance, the minimum safe distance between each functional area and surrounding sensitive facilities such as gas stations, charging stations, substations, densely populated buildings, and fire-fighting facilities is clearly defined. The closer the distance, the higher the risk weight.

[0098] Based on the division of area functions and vehicle behavior characteristics, the areas are arranged in order of risk level from low to high as follows: dedicated parking area for hazardous chemicals, general parking area, driving lane, refueling / charging operation area, emergency fire lane, evacuation exit, and area around fire-fighting facilities.

[0099] In a preferred embodiment of the present invention, the dynamic illegal parking risk quantification process includes: for the current location And the cumulative dwell time is The confirmed vehicle's dynamic illegal parking risk index is determined by the formula. Quantification, among which , These are the risk weight coefficient for determining the functional zone where the vehicle is currently located, and the multi-level time threshold vector, respectively. Here is the stationing risk evolution function, which is used to characterize the nonlinear evolution relationship between cumulative stationing duration and stationing risk.

[0100] It should be noted that the above For example, it can be a step or piecewise nonlinear function, when the cumulative dwell time... Exceed When the time threshold is different, the function output value will increase non-linearly.

[0101] It should be added that the above-mentioned early warning triggering process is as follows: multiple risk index thresholds are preset to divide different early warning levels, including attention level, warning level and emergency level.

[0102] When it is confirmed that the risk index of a vehicle's dynamic illegal parking exceeds the threshold of a certain level for the first time, a warning event corresponding to that level will be generated immediately.

[0103] The warning event is formatted as a JSON or XML data packet, which includes: the vehicle's identification ID, high-resolution evidence photos taken by a mobile proximity camera unit, GPS coordinates when the warning was triggered, complete historical location trajectory data, current dynamic illegal parking risk indicators, warning level, and the timestamp of the event generation.

[0104] The data packets are sent to a specific interface of the management terminal via a secure HTTPS or MQTT protocol, and are then visualized on the terminal interface in the form of map markers, sound alarms, and detailed information pop-ups.

[0105] This invention combines spatiotemporal information to output illegal parking risk assessment results, providing managers with comprehensive data support to help accurately predict potential risks and thus rapidly improve emergency response efficiency.

[0106] Example 2

[0107] like Figure 3 As shown, in the second embodiment of the present invention, a service area hazardous chemical vehicle illegal parking identification and detection system is provided, including: an image acquisition module, an instruction generation module, a close-up shooting module, an attribute confirmation module, and a risk quantification module.

[0108] The image acquisition module is connected to the instruction generation module, the instruction generation module is connected to the close-up shooting module, the close-up shooting module is connected to the attribute confirmation module, and the attribute confirmation module is connected to the risk quantification module.

[0109] The image acquisition module continuously scans the service area monitoring range using a fixed panoramic camera unit to acquire global scene image sequences in real time.

[0110] The instruction generation module divides the global scene image sequence into several regions to be analyzed. If any region to be analyzed is initially identified as a suspected target with basic characteristics of a hazardous chemical vehicle or the image quality judgment confidence is insufficient, it generates and issues a close-range acquisition instruction containing the initial three-dimensional position information of the region to be analyzed.

[0111] The close-up shooting module schedules and deploys close-up camera units in the area corresponding to the initial three-dimensional position information, controls their gimbals to adjust their attitude and focus to align with the target, and performs close-up shooting to obtain local close-up image data.

[0112] The attribute confirmation module performs multimodal parallel processing on local close-up image data and confirms the hazardous chemical attributes of the vehicle based on the joint results of hazardous chemical mark recognition and tank body three-dimensional structure analysis.

[0113] The risk quantification module establishes a tracking file for confirmed vehicles, updates their location trajectory frequently and records the cumulative parking time, associates it with a pre-built spatial risk zoning map to quantify dynamic illegal parking risks, and triggers an alert when the risk exceeds a threshold.

[0114] The service area hazardous chemical vehicle illegal parking identification and detection system provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the system embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0115] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0116] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0117] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0118] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0119] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0120] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A service area hazardous chemical vehicle parking violation identification detection method, characterized in that, Comprise: A fixed panoramic camera unit is used to continuously scan the monitoring range of the service area, and a global scene image sequence is obtained in real time; A plurality of to-be-analyzed regions are divided according to the global scene image sequence, and if any to-be-analyzed region preliminarily identifies a suspected target with basic features of a hazardous chemical vehicle or the image quality determination confidence is insufficient, a close-up collection instruction including the initial three-dimensional position information of the to-be-analyzed region is generated and sent out; A close-up camera unit deployed in the region corresponding to the initial three-dimensional position information is dispatched, and its gimbal is controlled to adjust the attitude and focal length to align with the target, and a close-up shooting is performed to obtain local close-up image data; Multi-modal parallel processing is performed on the local close-up image data, and the vehicle hazardous chemical attribute is confirmed based on the joint results of hazardous chemical sign identification and tank body three-dimensional structure analysis; A tracking file is established for the confirmed vehicle, the position trajectory is updated frequently, and the cumulative parking duration is recorded, the spatial risk zoning map is correlated to quantify the dynamic illegal parking risk, and the risk is triggered when the risk exceeds the threshold value. The pre-construction process of the spatial risk zoning map comprises: Import the service area design map and convert it into a digital grid form; Based on the predefined safety rules, a plurality of functional partitions are divided and labeled on the grid map, and a risk weight coefficient and a multi-level time threshold value are independently configured for each partition, the multi-level time threshold value is used to define the time length of the transition of the hazardous chemical vehicle from the compliant parking state to each level of illegal parking warning state, and the risk weight coefficient is used to quantify the inherent risk level of each functional partition for the hazardous chemical vehicle parking event; The dynamic illegal parking risk quantification process comprises: For the identified vehicle currently at location and accumulated parking duration , its dynamic risk indicator of illegal parking is quantified by the formula , where , are the risk weight coefficient of the functional sub-area where the identified vehicle is currently located and the multi-level time threshold vector, respectively, is the parking risk evolution function, which is used to represent the nonlinear evolution relationship between the accumulated parking duration and the parking risk.

2. The method according to claim 1, wherein, The real-time acquisition process of the global scene image sequence: The fixed panoramic camera unit performs image acquisition at a preset acquisition frame rate; The original images synchronously collected by the multiple wide-angle lenses in the fixed panoramic camera unit are subjected to real-time stitching and fusion processing, wherein the stitching and fusion processing is performed in a set order based on the installation position and field angle of each wide-angle lens, and the single-frame global scene image field of view is required to completely cover the vehicle activity area in the service area including at least a driving lane, a parking space and an emergency passage.

3. The method of claim 1, wherein the method further comprises: determining whether the vehicle is a dangerous chemical vehicle based on the vehicle type information; and determining whether the vehicle is parked in the service area based on the vehicle location information and the service area information. The plurality of to-be-analyzed region division process comprises: Vehicle entity detection is performed on each frame of the global scene image sequence to locate the minimum bounding rectangle of all vehicle entities in each frame; The minimum bounding rectangles of all frames in the sequence are subjected to spatio-temporal correlation analysis, a group of rectangles with a spatial overlap degree higher than a preset overlap threshold value are correlated as the same region, and multiple vehicle bounding rectangles belonging to the same region are merged to generate a corresponding to-be-analyzed region; For an independent rectangle that is not associated with any other rectangle, it is taken as a to-be-analyzed region, and thus a plurality of to-be-analyzed regions in the global scene image sequence are divided.

4. The method of claim 1, wherein the method further comprises: The preliminary identification process of the suspected target with basic features of a hazardous chemical vehicle comprises: The overall contour features and surface texture features of the vehicle entity in the to-be-analyzed region are extracted; The overall profile features are compared with a preset typical dangerous chemical vehicle profile library, and the similarity index of the vehicle entity profile in the area to be analyzed is quantified to determine whether it meets the basic profile characteristics, wherein the profile features at least include the elliptical cylindrical profile of the tank body, a certain number of saddle support structures, and the tail geometry for pipeline connection; Based on the surface texture feature analysis, it is judged whether the vehicle entity in the area to be analyzed meets the high light reflection characteristics of the metal tank body or the texture pattern of the corrosion-resistant coating; If at least two conditions are met in the feature analysis result, the vehicle entity is preliminarily identified as a suspected target with the basic features of a dangerous chemical vehicle.

5. The method of claim 1, wherein the method further comprises: determining whether the vehicle is a hazardous chemical vehicle based on the vehicle type information; and determining whether the vehicle is parked in the service area based on the vehicle location information and the service area information. The close-up camera unit scheduling process includes: At the close-up acquisition instruction receiving timestamp, according to the initial three-dimensional position information of the area to be analyzed, match and select the close-up camera unit with a current idle field of view covering the position from the preset close-up camera unit deployment information library; Control the close-up camera unit to quickly scan the vehicle entity, drive the gimbal to adjust the pitch and yaw angles, and perform optical zooming, when the key feature area occupies a proportion of the imaging area on the sensor target surface reaches a preset qualified threshold, perform multi-frame exposure acquisition; Wherein the key feature area includes the permanent position area of the dangerous chemical sign and the panoramic structure area of the tank body.

6. The method of claim 1, wherein the method further comprises: The dangerous chemical sign identification process includes: Image enhancement and geometric correction preprocessing is performed on the local close-up image data; Identify and locate the dangerous chemical sign elements in the preprocessed image, extract the visual display information, and the structured coding information including the UN number and the hazard category code; Based on the pre-constructed dangerous goods knowledge graph, the extracted coding information is logically consistent, and the verification process includes the logical conflict verification of the UN number and the hazard category code, the coding information and the visual display information; According to the logical consistency verification result, output the first confidence of the dangerous chemical sign identification.

7. The method of claim 1, wherein the method further comprises: determining whether the vehicle is a hazardous chemical vehicle based on the vehicle type information. The tank body three-dimensional structure analysis process includes: Identify the structure attribute type based on the tank body geometric structure information; Locate multiple key components on the tank body, analyze the relative spatial position, angle and size ratio relationship between the key components, match the corresponding structure attribute type adaptation rule, and generate the second confidence of the tank body three-dimensional structure analysis.

8. A service area dangerous chemical vehicle illegal parking identification detection system for performing the steps of a service area dangerous chemical vehicle illegal parking identification detection method according to any one of claims 1-7, characterized in that, It includes: An image acquisition module continuously scans the service area monitoring range through a fixed panoramic camera unit, and real-time acquires a global scene image sequence; An instruction generation module divides a plurality of areas to be analyzed for the global scene image sequence, and generates and issues a close-up acquisition instruction including the initial three-dimensional position information of the area to be analyzed if any area to be analyzed preliminarily identifies a suspected target with the basic features of a dangerous chemical vehicle or the image quality judgment confidence is insufficient; A close-up shooting module schedules the close-up camera unit deployed in the area corresponding to the initial three-dimensional position information, controls the gimbal to adjust the posture and focal length to aim at the target, and performs close-up shooting to acquire local close-up image data; An attribute confirmation module performs multi-modal parallel processing on the local close-up image data, and confirms the vehicle dangerous chemical attribute based on the joint results of the dangerous chemical sign identification and the tank body three-dimensional structure analysis. The risk quantification module confirms the vehicle to establish a tracking file, frequently updates the location trajectory, records the cumulative parking duration, associates the pre-constructed spatial risk partition map to quantify the dynamic illegal parking risk, and triggers a warning when the risk exceeds the threshold.

Images