Green vehicle ai identification data processing and remote value-added inspection coordination system
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGXI LUTONG TECH CO LTD
- Filing Date
- 2026-03-05
- Publication Date
- 2026-06-02
Smart Images
Figure CN122135566A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent transportation and inspection management technology, and more specifically, to a collaborative system for AI-based identification data processing and remote on-duty inspection of green channel vehicles. Background Technology
[0002] In the field of intelligent transportation and inspection management, the inspection of green channel vehicles is a crucial link in ensuring the efficiency of road network traffic and compliance supervision. Its core requirement is to achieve rapid vehicle release while ensuring inspection accuracy, thereby avoiding road congestion. Currently, the green channel inspection process involves multi-dimensional data collection, including cargo identification, vehicle status, road network environment, and historical records. However, existing technologies generally suffer from fragmented data integration, inconsistent formats of multi-source data, and a lack of standardized preprocessing procedures, resulting in insufficient data reliability and difficulty in forming effective inspection decision support, which in turn affects the efficiency of subsequent links.
[0003] The existing green channel inspection system fails to establish a collaborative link covering the entire process. Key nodes such as AI identification, remote monitoring, and inspection execution are relatively independent, and the efficiency of each link lacks a comprehensive quantitative evaluation system. The selection of remote monitoring personnel relies heavily on single-condition judgments, failing to fully consider multi-dimensional suitability factors such as personnel workload and cargo handling experience. The allocation of inspection channels also lacks a systematic consideration of vehicle suitability and channel efficiency, resulting in unreasonable resource matching. Some high-quality resources are occupied by inefficient tasks, while high-demand tasks face resource shortages.
[0004] Traditional green channel inspections often employ a rigid "first-come, first-served" model, failing to consider the efficiency differences and risk levels of various inspection tasks, thus hindering the priority processing of high-efficiency tasks. Furthermore, existing systems lack differentiated inspection process design, applying a uniform process to both high-confidence compliance tasks and low-confidence high-risk tasks. This either reduces traffic efficiency due to process redundancy or increases the risk of misjudgment due to insufficient control, making it difficult to balance inspection efficiency and regulatory accuracy, and ultimately failing to meet the intelligent inspection needs in complex road network environments. Summary of the Invention
[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a collaborative system for AI-based identification data processing and remote monitoring and inspection of green channel vehicles.
[0006] To achieve the above objectives, the present invention provides the following technical solution:
[0007] A collaborative system for AI-based identification, data processing, and remote monitoring and inspection of green channel vehicles includes:
[0008] The inspection data set module determines the target inspection section. After each preset time window, it collects multi-dimensional inspection data of each green channel vehicle that passes through the target inspection section within that preset time window, and then generates a collaborative data set of each green channel vehicle.
[0009] The collaborative inspection and analysis module constructs the collaborative inspection links for each green channel vehicle within the preset time window, and then obtains the link query execution index of the collaborative inspection links.
[0010] The collaborative inspection and execution module marks the corresponding collaborative inspection links as high-efficiency links, medium-efficiency links, or low-efficiency indices based on the link query execution index. All green channel vehicles are sorted in descending order according to the value of the link query execution index of the corresponding collaborative inspection link, and the green channel vehicles are inspected in the order of sorting.
[0011] Furthermore, the multidimensional inspection data includes the following four dimensions: AI recognition data, vehicle sensor data, road network and scene data, and historical inspection data.
[0012] Furthermore, the steps for generating the collaborative data set of green channel vehicles are as follows: Select a green channel vehicle, acquire its AI recognition data, vehicle sensor data, road network and scene data, and historical inspection data. Preprocess the multi-dimensional inspection data, using the license plate and vehicle identification number as unique identifiers. Align the preprocessed multi-dimensional inspection data with the same time window, and divide the data into four subsets according to data categories: AI recognition data subset, vehicle sensor data subset, road network and scene data subset, and historical inspection data subset. Integrate the four subsets in a structured manner to form the collaborative data set of green channel vehicles.
[0013] Furthermore, the execution index of the collaborative verification link query. Where, k1, k2, and k3 are all weighting coefficients.
[0014] Furthermore, Score for remote monitoring effectiveness; WL represents the load factor, and MA represents the fit. The historical average response time is y1; the baseline response time is the system-preset threshold for the longest acceptable effective response time when remote personnel handle remote inspection tasks for green channel vehicles; y1, y2, and y3 are all weighting coefficients, and y1+y2+y3=1.
[0015] Furthermore, Score the AI recognition performance. ; Confidence level for cargo type identification, x1 and x2 are the predicted values for cargo compliance; x1 and x2 are both weighting coefficients.
[0016] Furthermore, To verify the performance score; D represents the distance between the channel and the vehicle; the baseline maximum distance is a system-preset threshold for the maximum reasonable travel distance from the current GPS location to the entrance of the inspection channel for green channel vehicles; T represents the historical average travel time; the baseline travel time is a system-preset threshold for the maximum reasonable travel time for green channel vehicles to complete the entire inspection process in the inspection channel; z1 and z2 are both weighting coefficients.
[0017] Furthermore, the construction process of the collaborative inspection link for green channel vehicles is as follows: determine the AI recognition node, remote monitoring node, and inspection execution node of the green channel vehicle, and form a collaborative inspection link according to the one-way inspection logic.
[0018] Furthermore, when the link query execution index of the collaborative verification link is higher than the high-efficiency index, the corresponding collaborative verification link is marked as a high-efficiency link; when the link query execution index of the collaborative verification link is lower than the low-efficiency index, the corresponding collaborative verification link is marked as a low-efficiency link; and when the link query execution index of the collaborative verification link is between the high-efficiency index and the low-efficiency index, the corresponding collaborative verification link is marked as a medium-efficiency link.
[0019] Compared with the prior art, the present invention has the following beneficial effects:
[0020] This system constructs a collaborative inspection link encompassing key nodes such as AI recognition, remote monitoring, and inspection execution. It scientifically calculates the link query execution index to comprehensively quantify the efficiency of each inspection stage, making the overall operational status of the link measurable and assessable. This effectively solves the problems of fragmented data, loose connections between stages, and difficulty in coordinating efficiency in traditional inspections. Based on the link query execution index, the collaborative inspection link is divided into different efficiency levels, achieving precise definition and differentiation of the inspection link efficiency. Green channel vehicles are arranged in descending order of execution index for inspection, breaking the traditional first-come, first-served model. This prioritizes higher-efficiency inspection tasks, ensuring that high-quality resources are allocated to high-efficiency links and avoiding resource waste in low-efficiency links. This intelligent inspection sorting method makes the inspection process more targeted and orderly, improving overall inspection efficiency and ensuring the rational allocation and efficient utilization of inspection resources.
[0021] An integrated green channel inspection system has been established, encompassing comprehensive data support, precise link quantification, and intelligent and orderly inspection. The construction of collaborative inspection links and execution indices enables a systematic quantitative assessment of the entire inspection process, making link efficiency clearly visible. Meanwhile, link grading and orderly inspection based on indices achieve intelligent scheduling and optimization of the inspection process, effectively addressing pain points in traditional green channel inspections such as the difficulty in balancing efficiency and accuracy and unreasonable resource allocation. This significantly improves the scientific nature, intelligence level, and overall operational efficiency of green channel inspection work. Attached Figure Description
[0022] Figure 1 This is a schematic diagram of the system of the present invention;
[0023] Figure 2 A flowchart for generating collaborative data sets for green channel vehicles;
[0024] Figure 3 A flowchart for constructing a collaborative inspection link for green channel vehicles. Detailed Implementation
[0025] Reference Figures 1 to 3 A collaborative system for AI-based identification, data processing, and remote monitoring and inspection of green channel vehicles, including:
[0026] Inspection Data Set Module: Determine the target inspection section. After each preset time window (e.g., 15-30 seconds / preset time window, dynamically adjusted according to the traffic density of the target inspection section), collect multi-dimensional inspection data (multi-dimensional inspection data includes the following four dimensions: AI recognition data, vehicle sensor data, road network and scene data, and historical inspection data) of each green channel vehicle passing through the target inspection section within the preset time window, and then generate a collaborative data set of each green channel vehicle.
[0027] The steps for generating the collaborative data set for green channel vehicles are as follows: Select a green channel vehicle and obtain its AI recognition data (capture images / videos of the cargo after the cargo box is opened using a 4K high-definition camera at the toll station and a vehicle-mounted front-view camera, covering the overall appearance of the cargo and key details); extract the license plate, vehicle identification number (VIN), and vehicle registration information (including approved load capacity, vehicle type, etc.) using an OCR recognition device; analyze the cargo images / videos using AI visual algorithms to extract cargo contour features and texture features (such as surface roughness and color distribution), and determine the cargo type (such as cruciferous vegetables, etc.) based on feature matching with the national green channel product catalog. Temperate fruits, live livestock and poultry, etc.), synchronously outputting cargo type identification confidence level (0-1 range)), vehicle-mounted sensor data (collecting real-time GPS positioning information of the vehicle through the OBU terminal; collecting the cargo storage environment temperature through the temperature sensor built into the cargo box; collecting the actual load data of the vehicle through roadside dynamic weighing equipment or vehicle-mounted load sensors; synchronously recording the data collection timestamp (accurate to the second)), road network and scene data (obtaining the real-time traffic density (vehicles / 10 minutes) of the target inspection section (such as the green channel area of the toll station) and the idle status (idle / occupied) of the corresponding inspection lane through roadside radar and traffic flow monitoring equipment; collecting data through meteorological sensors). The system collects current weather conditions (sunny / cloudy / rainy / snowy) and light intensity (lux); obtains the operating status (normal / faulty) of high-definition cameras and weighing equipment in the inspection lane through the toll station equipment management system; historical inspection data (retrieved from the provincial green channel inspection database, including past inspection records of the vehicle (using "license plate + VIN" as a unique index), including historical passage counts, compliance / violation judgment results of each inspection (such as whether it is a false declaration, whether it is overloaded), corresponding cargo type, and average inspection time); and performs multi-source data standardization preprocessing: uniformly organizes the four types of collected data to ensure consistent format and data reliability; format standardization: Data timestamps are uniformly converted to UTC+8 format (e.g., "2024-06-20 10:15:32"); GPS positioning information is converted to the WGS-84 coordinate system standard format (e.g., 113.4250°E, 23.3125°N); text data is aligned according to a unified standard—goods types are semantically normalized according to the national green channel product catalog (e.g., "broccoli" and "cauliflower" are uniformly classified as "cruciferous vegetables"), and text fields such as vehicle registration information and historical compliance / violation results uniformly adopt the "full Chinese name + code" format (e.g., "compliance-01", "violation (false declaration)-02").Data cleaning: Median filtering is used to remove noise interference from cargo images to ensure accurate feature extraction; outlier detection is performed on sensor data (e.g., load data exceeding the vehicle's rated load capacity by more than 30%, temperature data exceeding the suitable storage range of green channel cargo by ±5℃). After marking outliers, secondary data collection is triggered. Data that cannot be collected a second time is marked as "invalid data"; missing non-critical fields (e.g., light intensity, some historical inspection notes) are filled using interpolation. Missing critical fields (e.g., license plate, cargo type, actual load) are directly judged as data collection failure and need to be collected again. Numerical data is normalized according to standard range (mapped to the [0,1] interval) - load data is calculated as "actual load / rated load capacity", temperature data is calculated as "(actual temperature - suitable lower limit of green channel cargo) / (suitable upper limit of suitable lower limit of suitable temperature)", and traffic density is calculated as "current density / maximum carrying capacity of road section". Data association and integration generate a collaborative set: Using "license plate + vehicle identification number (VIN)" as the unique identifier, data is aligned within the same time window (e.g., 15 seconds / data, consistent with the data collection cycle), and divided into four subsets according to data category: AI recognition data subset (including license plate, VIN, vehicle type, cargo image keyframes, cargo feature vector, cargo type, and cargo type recognition confidence); vehicle-mounted sensor data subset (including GPS positioning, cargo box temperature, actual load, and normalized values); road network and scene data subset (including traffic density, lane availability, weather conditions, and equipment operating status); and historical inspection data subset (including historical passage counts, compliance / violation records, and average inspection time). These four subsets are then structurally integrated to form a collaborative data set for green channel vehicles.
[0028] The method for obtaining confidence in cargo type identification is as follows: Image / video preprocessing: Illumination equalization (CLAHE algorithm) and median filtering are performed on the acquired cargo images (4K resolution), and non-cargo areas such as vehicle compartments and tarpaulins are cropped; the video is extracted at 1 second / frame, retaining key frames such as the overall cargo view and close-up details, and uniformly scaled to 640×640 resolution to adapt to the algorithm input. Feature extraction: ① Contour features: Cargo edges are extracted using Canny edge detection, and shape complexity, aspect ratio, and other quantitative indicators are calculated using minimum bounding rectangle and contour moments; ② Texture features: Surface roughness features are extracted using the LBP (Local Binary Mode) algorithm, and texture contrast and correlation are calculated using GLCM (Gray-Level Co-occurrence Matrix), while simultaneously calculating a color distribution histogram (such as green proportion and brightness distribution) in the HSV color space; ③ A 128-dimensional cargo feature vector is generated through fusion. Green Channel Catalog Matching: A national green channel product catalog feature template library is constructed (containing standard feature vectors for various goods across 8 major categories, including cruciferous vegetables and temperate fruits, with freshness-related feature thresholds annotated). A cosine similarity algorithm is used to calculate the matching degree between the feature vector of the goods to be identified and the features of each category in the template library, selecting the candidate goods type with the highest matching degree. Confidence Calculation and Output: Confidence = Top 1 Matching Degree.
[0029] Collaborative Inspection and Analysis Module: Constructs collaborative inspection links for each green channel vehicle within the preset time window, and then obtains the link query execution index of the collaborative inspection links.
[0030] Link query execution index of collaborative verification link ;in, Score the AI recognition performance. ; Confidence level for cargo type identification, The compliance prediction value for goods is calculated by comparing the AI-identified and semantically normalized goods type (e.g., "cruciferous vegetables") with the National Green Channel Product Catalog. If the goods belong to the catalog, CC=1 (compliance prediction); otherwise, CC=0 (non-compliance prediction). This is a binary judgment value. x1 and x2 are both weighting coefficients. Since the core pain point of green channel inspection is "false declaration (goods type does not match)," CI directly targets this pain point to quantify the reliability of identification, which is a key indicator to avoid misjudgment; while CC is only a basic compliance threshold judgment, and its information content and decision priority are lower than CI. Therefore, the value of x1 can be 0.7 and the value of x2 can be 0.3. Score for remote monitoring effectiveness; ; The historical average response time is used; the baseline response time is the system-preset threshold for the longest acceptable effective response time for remote personnel handling remote inspection tasks of green channel vehicles; y1+y2+y3=1, because (1-WL) directly reflects the current workload and immediate response capability of the personnel on duty, which is the foundation of the core requirement for rapid response in green channel inspection. If personnel are busy (high WL), even with excellent adaptability and historical efficiency, tasks cannot be handled in time, so it is given a high weight. MA quantifies the personnel's experience in handling current goods, which directly determines the accuracy and speed of verification (e.g., personnel familiar with cruciferous vegetables can quickly identify features), and is a core indicator to avoid inspection misjudgments. Its importance is comparable to real-time idleness, so its weight is consistent with y1. TR is historical data, which only reflects past response performance. Its impact on the current task is weaker than the two immediate indicators of real-time idleness and real-time adaptability, so it is given a lower weight to avoid historical data from excessively affecting the performance evaluation of the current task. The value of y1 can be 0.4, the value of y2 can be 0.4, and the value of y3 can be 0.2; To verify the performance score; D represents the distance between the channel and the vehicle; the baseline maximum distance is a system-preset threshold for the maximum reasonable travel distance for green channel vehicles from their current GPS location to the entrance of the inspection channel; T represents the historical average travel time; the baseline travel time is a system-preset threshold for the maximum reasonable travel time for green channel vehicles to complete the entire inspection process of "on-site verification → data review → release / verification" in the inspection channel; z1 and z2 are both weighting coefficients, z1+z2=1, because the arrival time and the processing time within the channel have equally important impacts on the overall efficiency of the inspection execution, so the value of z1 can be 0.5 and the value of z2 can be 0.5; k1, k2, and k3 are all weighting coefficients, k1+k2+k3=1, because The corresponding inspection and execution phase is the "final landing point" of the collaborative link. The core objective of green channel inspection is to "efficiently complete on-site verification and release." The efficiency of the execution phase (channel) directly determines the actual passage time of vehicles and also directly affects the traffic congestion at toll stations. It is the "final manifestation" of the overall efficiency of the entire link and therefore is given the highest weight. (AI recognition) is the "data foundation" of the process, ensuring the accuracy of the starting point for decision-making; Remote monitoring acts as a "decision bridge" in the process, ensuring efficient data-to-execution transition. Both remote monitoring and execution are supporting components of the execution process, and their value must be realized through execution; therefore, their weight is lower than remote monitoring. Meanwhile, the accuracy of AI recognition and the adaptability of remote monitoring have a similar support for execution efficiency, so they are weighted equally. Therefore, the value of k1 can be 0.3, the value of k2 can be 0.3, and the value of k3 can be 0.4.
[0031] The construction process of the collaborative inspection link for green channel vehicles is as follows: determine the AI recognition node, remote monitoring node, and inspection execution node of the green channel vehicle, and form a collaborative inspection link according to the one-way inspection logic (the one-way inspection logic is AI recognition node → remote monitoring node → inspection execution node).
[0032] The steps for determining the AI recognition node of green channel vehicles are as follows: directly use a subset of AI recognition data of green channel vehicles as the basic data of AI recognition nodes, extract core attributes (license plate, VIN, vehicle type, cargo type, cargo type recognition confidence CI, cargo compliance prediction value CC, cargo feature vector, image key frame), and generate a unique node identifier by AIN-the last 6 digits of the license plate, thereby generating a dedicated AI recognition node for green channel vehicles.
[0033] The steps for determining the remote monitoring nodes for green channel vehicles are as follows: Obtain the monitoring data of each monitoring personnel in the online monitoring personnel resource pool (the online monitoring personnel resource pool is a set of monitoring personnel who are specifically configured for the target inspection section (such as a green channel of a highway toll station), are online in real time, and have the qualifications for remote inspection of green channel vehicles). The monitoring data of the monitoring personnel includes the real-time online status (idle / busy), the number of tasks to be processed, the maximum number of tasks that can be carried (system preset, such as 5 tasks / person), the list of cargo types processed in the past, and the historical average response time (the historical average response time refers to the average time from receiving the task instruction to completing the first effective response when the remote monitoring personnel have processed remote inspection tasks for green channel vehicles in the past). A three-tiered screening rule is set up based on the logic of "resource availability → high efficiency of adaptation → priority sorting" to verify whether the on-duty personnel meet the requirements in the following order: First tier (resource availability verification): Only personnel whose "online status is idle" or "load rate WL = current number of pending tasks / maximum number of tasks ≤ 0.7" are retained, while busy personnel who cannot respond immediately are removed; Second tier (adaptability verification): Among the on-duty personnel who pass the first tier, those whose "historical list of processed goods types includes the current green channel vehicle's goods type" (adaptability MA = 1) are selected; if there are no perfect matches, general personnel with "historical processed green channel categories ≥ 5" (adaptability MA = 0.6) are retained; Third tier (priority sorting): If there are ≥ 2 personnel who pass the first two tiers, they are sorted by "lowest load rate WL" and the personnel with the highest idleness are selected. Remote on-duty node determination and attribute extraction: After passing the three-tiered screening, one on-duty personnel is uniquely determined as the remote on-duty node for this green channel vehicle, and the core attributes of the node are extracted (node identifier = RVN - on-duty personnel ID, WL, MA, historical average response time).
[0034] The steps for determining the inspection execution node for green channel vehicles are as follows: Obtain channel data for each inspection channel in the green channel inspection channel resource pool (channel data includes channel number, real-time idle status (idle / occupied), key equipment operating status (high-definition camera, dynamic weighing equipment, temperature verification sensor, etc.), compatible vehicle type (light / medium / heavy truck), compatible cargo type (ordinary green channel / cold chain green channel), distance from the vehicle's current GPS location, and historical average passage time). Verify the channel sequentially according to availability → adaptability → priority logic, and select the optimal channel: First layer (availability verification): retain channels that are "idle" and "all key equipment are operating normally" (equipment integrity rate ER=1; key equipment refers to irreplaceable hardware that supports the core process of green channel vehicle inspection (identity verification, cargo compliance verification, data verification), including channel-level 4K high-definition cameras, channel-specific OCR recognition equipment, and temperature verification sensors). The first layer (equipment, auxiliary displays, voice announcers, etc.) removes channels that are occupied or have malfunctioning core equipment, ensuring that channels are available immediately. The second layer (compatibility verification) filters available channels that are "compatible with the target vehicle type" (e.g., light trucks corresponding to small green channel) and "compatible with the target vehicle's cargo type" (e.g., cold chain green channel corresponding to channel with temperature verification equipment), avoiding channels that are incompatible with vehicles / cargo. The third layer (priority sorting) sorts channels that meet the first two layers' rules by being closest to the vehicle's GPS and having the shortest historical average travel time (i.e., if multiple channels are at the same distance from the vehicle's GPS (e.g., channels A1 and A2 are both 30 meters from the vehicle), then the "historical average travel time" of each channel is retrieved), selecting the channel with the best overall efficiency. Inspection execution node determination and attribute extraction: After three layers of screening, one channel is uniquely determined as the inspection execution node for the green channel vehicle, and the core attributes are extracted (node identifier = EEN-channel number, channel idle rate CR=1, equipment integrity rate ER=1, suitable cargo type, channel and vehicle distance).
[0035] Collaborative Inspection Execution Module: When the link query execution index of a collaborative inspection link is higher than the high-efficiency index, the corresponding collaborative inspection link is marked as a high-efficiency link. When the link query execution index of a collaborative inspection link is lower than the low-efficiency index, the corresponding collaborative inspection link is marked as a low-efficiency link. When the link query execution index of a collaborative inspection link is between the high-efficiency index and the low-efficiency index, the corresponding collaborative inspection link is marked as a medium-efficiency link. (High-efficiency index: must correspond to the LEI threshold of "the link can stably achieve fast and accurate inspection", usually set to 0.8; Low-efficiency index: must correspond to the LEI threshold of the link having low efficiency or insufficient accuracy risk, can be set to 0.5; and dynamically adjusted according to the real-time traffic density of the target inspection section: peak traffic (≥50 vehicles / 10 minutes): high-efficiency index increased to 0.85 (receiving...) To tighten the threshold for rapid release and prioritize resources for the best links, the inefficiency index has been increased to 0.55 (expanding the scope of enhanced review and reducing the risk of misjudgment). During periods of low traffic flow (≤20 vehicles / 10 minutes): the high efficiency index has been decreased to 0.75 (expanding the scope of rapid release and improving traffic efficiency), and the inefficiency index has been decreased to 0.45 (reducing the scope of enhanced review and reducing resource redundancy). All green channel vehicles are sorted in descending order of the link query execution index of their corresponding collaborative inspection links, and inspected sequentially according to the sorted order. (If a green channel vehicle corresponds to a high efficiency link, manual review is skipped, and AI directly pushes the release instruction; if a green channel vehicle corresponds to a medium efficiency link, AI identification is retained → 1 manual review → release; if a green channel vehicle corresponds to an inefficiency link, AI secondary matching, cross-node cross-review, and comparison of channel data with historical data are added.)
[0036] Example: High-efficiency link (vehicle 1): LEI=0.88; Core parameters: AI recognition node: CI=0.95 (cargo type = temperate fruit - apple), CC=1 (compliant within the catalog); Remote monitoring node: MA=1 (historically processed apples 5 times), WL=0.2 (currently 1 pending processing, maximum capacity 5), TR=12 seconds; Inspection execution node: D=25 meters (distance from the vehicle's current GPS location), T=18 seconds (historical average passage time), Applicable vehicle type = light truck (consistent with vehicle type). Inspection method: After the AI recognition node generates a collaborative data set, it automatically determines that LEI=0.88≥High Efficiency Index (0.8); The system triggers a simplified process: skipping the manual review by the remote monitoring node, the AI recognition node directly extracts the cargo feature vector and compliance prediction result (CC=1) to generate a "Fast Release Instruction"; The AI recognition node pushes the instruction to the inspection execution node in real time (channel number E03, short-distance high efficiency channel); After receiving the instruction, the inspection execution node only verifies that the vehicle license plate + VIN is consistent with the instruction, without additional verification of the cargo, and completes the gate opening and release within 1 second.
[0037] Medium-efficiency link (vehicle 2): LEI=0.65; Core parameters: AI recognition node: CI=0.75 (cargo type = cruciferous vegetables - broccoli), CC=1 (compliant within the catalog); Remote monitoring node: MA=0.6 (historically processed 6 categories of green channel goods, broccoli has not been processed), WL=0.4 (currently 2 items to be processed), TR=18 seconds; Inspection execution node: D=40 meters, T=28 seconds, compatible vehicle type = medium-sized truck (consistent with vehicle type). Inspection Method: After the AI recognition node generates a collaborative data set, it determines LEI=0.65 (0.5≤LEI<0.8); the system triggers the standard process: the AI recognition node pushes the keyframes, feature vectors, and CI / CC results of the cargo image to the remote monitoring node (monitor ID: RV08); after receiving the task, the remote monitoring node completes the first effective response: quickly checks the cargo image and feature vectors, confirms that it matches the "cruciferous vegetables" template, and signs the "Review Approval Opinion"; the remote monitoring node pushes the review results to the inspection execution node (channel number E05, regular channel); the inspection execution node conducts on-site verification: the license plate + VIN is recognized by OCR, and the load is verified by dynamic weighing equipment (actual load / rated load capacity = 0.8, compliant), completing the verification.
[0038] Low-efficiency link (vehicle 3): LEI=0.42; Core parameters: AI recognition node: CI=0.4 (cargo type=live poultry-broiler chicken), CC=1 (compliant within the catalog); Remote monitoring node: MA=0.6 (historically processed 5 categories of green channel goods, never processed broiler chicken), WL=0.5 (currently 2 pending processing, maximum capacity 4), TR=22 seconds; Inspection execution node: D=75 meters, T=45 seconds, compatible vehicle type=heavy truck (consistent with vehicle type), equipment integrity rate=1. Verification method: After the AI identification node generates a collaborative data set, it determines that LEI=0.42 < Inefficiency Index (0.5); The system triggers an enhanced process: Step 1 (AI secondary matching): The AI identification node re-extracts the cargo feature vector and calculates the cosine similarity again with the 3 sets of standard feature vectors of "fresh livestock and poultry - broiler chickens" in the National Green Channel Catalogue. The matching degree is ≥0.85, and a "Secondary Matching Report" is generated; Step 2 (Cross-node cross-verification): Two remote monitoring nodes are assigned (e.g., main RV05 + auxiliary RV12, where the main remote monitoring node is the first-ranked remote monitoring node in the third layer (priority sorting) of the determination steps for the auxiliary remote monitoring node); In the middle, the second-ranked remote monitoring node in the third layer (priority sorting), the main RV05 checks the cargo image against the secondary matching report, and the auxiliary RV12 retrieves historical inspection cases of similar cargo (key characteristics of broiler inspection: feather integrity, activity level), and the two people sign the "Review Approval Opinion" respectively; the third step (comparison of channel data with historical data): the inspection execution node (channel number E01, reinforced channel) collects cargo temperature (18℃, suitable range 15-20℃) and load data (actual load / approved load capacity = 0.9) on-site and compares them with the vehicle's historical inspection data (the temperature of the last 3 broiler transports was 16-19℃, and the load was compliant) to ensure consistency; after all the reinforced steps are completed, the inspection execution node raises the barrier to allow passage.
[0039] The above formulas are all dimensionless calculations, and the preset parameters in the formulas should be set by those skilled in the art according to the actual situation.
[0040] 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, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.
[0041] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.
[0042] Those skilled in the art will recognize that the units 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.
[0043] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0044] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
[0045] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0046] 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.
Claims
1. A collaborative system for AI-based identification data processing and remote monitoring and inspection of green channel vehicles, characterized in that: include: The inspection data set module determines the target inspection section. After each preset time window, it collects multi-dimensional inspection data of each green channel vehicle that passes through the target inspection section within that preset time window, and then generates a collaborative data set of each green channel vehicle. The collaborative inspection and analysis module constructs the collaborative inspection links for each green channel vehicle within the preset time window, and then obtains the link query execution index of the collaborative inspection links. The collaborative inspection and execution module marks the corresponding collaborative inspection links as high-efficiency links, medium-efficiency links, or low-efficiency indices based on the link query execution index. All green channel vehicles are sorted in descending order according to the value of the link query execution index of the corresponding collaborative inspection link, and the green channel vehicles are inspected in the order of sorting.
2. The collaborative system for AI-based identification data processing and remote monitoring and inspection of green channel vehicles according to claim 1, characterized in that, The multidimensional inspection data includes the following four dimensions: AI recognition data, vehicle sensor data, road network and scene data, and historical inspection data.
3. The collaborative system for AI recognition data processing and remote monitoring and inspection of green channel vehicles according to claim 1, characterized in that, The steps for generating the collaborative data set for green channel vehicles are as follows: Select a green channel vehicle, acquire its AI recognition data, vehicle sensor data, road network and scene data, and historical inspection data. Preprocess the multi-dimensional inspection data, using the license plate and vehicle identification number as unique identifiers. Align the preprocessed multi-dimensional inspection data with the same time window, and divide the data into four subsets according to data categories: AI recognition data subset, vehicle sensor data subset, road network and scene data subset, and historical inspection data subset. Integrate the four subsets in a structured manner to form the collaborative data set for green channel vehicles.
4. The collaborative system for AI recognition data processing and remote monitoring and inspection of green channel vehicles according to claim 1, characterized in that, Link query execution index of collaborative verification link Where, k1, k2, and k3 are all weighting coefficients.
5. The collaborative system for AI recognition data processing and remote monitoring and inspection of green channel vehicles according to claim 4, characterized in that, Score for remote monitoring effectiveness; WL represents the load factor, and MA represents the fit. The historical average response time is y1; the baseline response time is the system-preset threshold for the longest acceptable effective response time when remote personnel handle remote inspection tasks for green channel vehicles; y1, y2, and y3 are all weighting coefficients, and y1+y2+y3=1.
6. The collaborative system for AI recognition data processing and remote monitoring and inspection of green channel vehicles according to claim 4, characterized in that, Score the AI recognition performance. ; Confidence level for cargo type identification, x1 and x2 are the predicted values for cargo compliance; x1 and x2 are both weighting coefficients.
7. The collaborative system for AI recognition data processing and remote monitoring and inspection of green channel vehicles according to claim 4, characterized in that, To verify the performance score; D represents the distance between the channel and the vehicle; the baseline maximum distance is a system-preset threshold for the reasonable maximum travel distance of a green channel vehicle from its current GPS location to the entrance of the inspection channel. T represents the historical average passage time; the benchmark passage time is the system-preset threshold for the maximum reasonable time for green channel vehicles to complete the entire inspection process in the inspection lane; z1 and z2 are both weighting coefficients.
8. The collaborative system for AI recognition data processing and remote monitoring and inspection of green channel vehicles according to claim 1, characterized in that, The construction process of the collaborative inspection link for green channel vehicles is as follows: determine the AI recognition node, remote monitoring node and inspection execution node of green channel vehicles, and form a collaborative inspection link according to the one-way inspection logic.
9. The collaborative system for AI recognition data processing and remote monitoring and inspection of green channel vehicles according to claim 1, characterized in that, When the link query execution index of a collaborative verification link is higher than the high-efficiency index, the corresponding collaborative verification link is marked as a high-efficiency link. When the link query execution index of a collaborative verification link is lower than the low-efficiency index, the corresponding collaborative verification link is marked as a low-efficiency link. When the link query execution index of a collaborative verification link is between the high-efficiency index and the low-efficiency index, the corresponding collaborative verification link is marked as a medium-efficiency link.