A deep learning-based road transport vehicle risk level assessment method

Abstract

The present application relates to the field of road transport safety management and artificial intelligence technology, and particularly relates to a road transport vehicle risk level evaluation method based on deep learning; the method comprises the following steps; S1, driving behavior feature recognition; S2, vehicle tire deformation judgment; S3, vehicle running stability analysis; S4, risk contribution judgment; the present application extracts the behavior feature vector in real time, judges the dangerous transport operation behavior category and risk level to which it belongs based on clustering analysis; obtains the deformation category of the tire, combines the tire deformation judgment result with the vehicle running stability judgment objective transport hidden danger risk, and performs risk contribution analysis based on the correlation of dangerous behavior and hidden danger risk events within the time window, realizing dynamic adjustment of the risk level; the present application dynamically adjusts the alarm level through multi-channel data recognition, avoiding single behavior misjudgment; at the same time, it can also adapt to different slope sections, loads and road conditions, and has good scene versatility and engineering practicability.

Description

Technical Field

[0001] This invention relates to the fields of road transport safety management and artificial intelligence technology, and in particular to a method for assessing the risk level of road transport vehicles based on deep learning. Background Technology

[0002] With the rapid development of the road transport industry, the vehicle operating environment is becoming increasingly complex, and road transport safety issues are becoming more prominent. Especially in scenarios such as long-distance freight, high-frequency transportation, and complex road conditions, improper driver operation and abnormal vehicle operation can easily lead to traffic accidents. Currently, existing risk assessment methods are mostly based on static rule settings and simplified behavioral characteristic judgments, failing to effectively integrate driver behavior, vehicle status, and environmental factors. This results in low risk identification accuracy and delayed level assessment results, making it difficult to meet the actual needs of transport safety management.

[0003] Furthermore, some methods based on traditional machine learning models only involve single-dimensional data features, lacking a deep understanding and classification of driving behavior patterns, and failing to achieve dynamic updates of risk levels and multi-dimensional attribution, especially lacking the ability to comprehensively analyze dangerous operating behaviors and objective hidden danger events.

[0004] Therefore, there is an urgent need for a risk level assessment method based on deep learning, which can extract and cluster driver operating behavior in real time, and combine vehicle operating conditions and objective hidden danger information such as tire deformation to achieve highly robust and accurate risk level identification and dynamic adjustment, so as to improve the overall safety management level in the road transportation process.

[0005] Chinese Patent Publication No. CN114022000A discloses a method and system for assessing vehicle road transport risks based on fuzzy comprehensive evaluation. This invention constructs a vehicle road transport safety evaluation index system from multiple perspectives, including vehicle factors, cargo factors, real-time road environment, natural environment, human factors, and transport administration factors. It uses fuzzy comprehensive evaluation to transform qualitative evaluation into quantitative evaluation based on the membership degree of fuzzy data, systematically addressing indeterminate influencing factors. However, this invention is a multi-factor static weighted scoring model, which cannot fully reflect the causal relationships between different behaviors and makes it difficult to predict the likelihood of risks in advance. Furthermore, many key parameters, such as factor weights and membership function design, rely on manual or expert scoring methods for determination. Therefore, the method and system for assessing vehicle road transport risks based on fuzzy comprehensive evaluation suffer from the following problems: lack of dynamic modeling capability, static evaluation results, poor self-learning and adaptive capabilities, and a lack of interpretability and traceability of the evaluation results. Summary of the Invention

[0006] To address this, the present invention provides a method for assessing the risk level of road transport vehicles based on deep learning, which overcomes the problem in the prior art of lacking risk warnings for transport vehicles by combining tire deformation characteristics and transport section characteristics, resulting in weak targeted warning response capabilities and inaccurate risk level classification.

[0007] To achieve the above objectives, this invention provides a method for assessing the risk level of road transport vehicles based on deep learning, characterized by comprising: Real-time collection of driver's transportation operation behavior data and corresponding actual behavior feature vectors; cluster analysis to determine whether there are dangerous transportation operation behaviors in the actual behavior feature vectors; and when they exist, obtaining the dangerous transportation operation behavior category and risk level corresponding to the dangerous transportation operation behavior, as well as collecting the target vehicle's operation data. The tire deformation type of the target vehicle within the transportation section is determined based on the operational data. Based on the tire deformation category judgment result and the vehicle operation stability analysis mode, determine whether there is an objective transportation hazard risk, and if so, judge the contribution of the objective transportation hazard risk to the risk of the dangerous transportation operation behavior. The steps for determining risk contribution include storing transportation operation events and objective transportation potential risk events by timestamp. Using the first time window as the time search window and the dangerous transportation operation as the starting point of the time search window, only the number of objective transportation hidden danger risk events within the time search window is obtained and compared with the standard occurrence number. When the first occurrence comparison result is obtained, the corresponding enhanced risk coefficient is obtained to adjust the risk level of the dangerous transportation operation.

[0008] Furthermore, the transportation operation behavior data includes driver posture information and driver eye attention information, and the behavior feature vector is a feature vector that represents the driver posture information and eye attention information after normalization processing. Cluster analysis is used to determine the category of hazardous transportation operation to which the behavioral feature vector belongs and the risk level of that hazardous transportation operation. Obtain the feature space corresponding to the database of dangerous transport operations and the feature vector of each dangerous transport operation. The actual similarity between the actual behavior feature vector and each dangerous operation behavior feature vector is determined to determine whether dangerous transportation operation behavior exists. Based on the judgment results, the category of dangerous transportation operation behavior is determined, and combined with the risk classification criteria of each risk level within the dangerous transportation operation behavior, the types of dangerous transportation operation behaviors and their corresponding risk levels of the current transportation operation behavior are determined. Specifically, when the determination result is that the danger exists, the corresponding dangerous transportation operation behavior type is obtained, and the risk level is judged. The risk level determination process involves using a first time window as the time search window, with the detection of dangerous transportation operations as the starting point of the time search window, and only obtaining the risk level of the dangerous transportation operations at the starting point of the time search window.

[0009] Furthermore, determining the actual similarity between the actual behavior feature vector and each dangerous operation behavior feature vector includes: Calculate the similarity scores between the actual behavior feature vector and each dangerous behavior feature vector; Each similarity judgment value is compared with its corresponding similarity threshold, and the existence of dangerous transportation operations is determined based on the comparison results.

[0010] Furthermore, based on the operational data, the tire deformation category of the target vehicle within the transportation section is determined to include: Determine the tire deformation of the target vehicle based on tire deformation. Determine whether to perform the transportation section interference identification step based on the tire deformation judgment results; By combining the tire deformation and the interference identification results of the transportation section, the tire deformation category of the target vehicle under the transportation section is determined, and based on the tire deformation category determination results and the vehicle operation stability analysis mode, it is determined whether there are any objective transportation hidden risks.

[0011] Furthermore, based on the tire deformation category judgment results and the vehicle operation stability analysis mode, determining whether there are objective transportation potential risks includes: Based on the tire deformation category judgment results, a vehicle operation stability analysis mode is selected, including vehicle operation condition detection and vehicle operation center of gravity shift monitoring. The existence of objective transportation risks is determined based on vehicle operation stability analysis models.

[0012] Furthermore, the vehicle operating condition detection includes dynamic identification of vehicle operating angles and identification of tire wear during vehicle operation, wherein, The process of performing dynamic recognition of vehicle running angle is as follows: The vehicle speed of the target vehicle during the turning process of the dangerous transportation operation behavior at the starting point of the window is obtained, and the centripetal component speed of the vehicle pointing towards the center of the curve is decomposed based on the vehicle speed. The corresponding centripetal angular velocity of the vehicle is calculated based on the turning radius of the target vehicle and compared with the standard centripetal angular velocity threshold. If the vehicle's centripetal angular velocity exceeds the standard centripetal angular velocity threshold and there is no objective transportation hazard risk, a deceleration warning will be triggered. If the vehicle's centripetal angular velocity is less than or equal to the standard centripetal angular velocity threshold, the vehicle's tire wear identification operation will be performed to determine whether there are any objective transportation hazards based on the tire wear identification results.

[0013] Furthermore, the operation of identifying tire wear during vehicle operation includes, The tire tread depth of the target vehicle at the start of the window during dangerous transportation operations is obtained, along with the tire tread depth of the target vehicle to be detected, and the wear value of the tire tread depth during vehicle operation is calculated. Wherein, the wear value of the tire tread depth of the vehicle is the difference between the tire tread depth of the target vehicle being detected en route and the tire tread depth to be detected, and the tire tread depth being detected en route is the tire tread depth at the start of the window when dangerous transportation operation behavior occurs. Compare the tire tread depth wear value of the vehicle under operation with the standard tire tread wear value. If the tire tread wear value of a vehicle exceeds the standard tire tread wear value, there is an objective transportation hazard risk, triggering a deceleration warning and guiding the vehicle to the nearest service area for vehicle maintenance. If the tire tread depth wear value of the vehicle is less than or equal to the standard tire tread wear value, and there is no objective transportation hazard risk, the initial tire pressure adjustment standard reminder will be triggered.

[0014] Furthermore, the process of monitoring vehicle center of gravity shift includes, Obtain the load distribution information from the target vehicle's pre-departure detection information, and obtain the target vehicle's pre-departure center of gravity based on the load distribution information; The vehicle's center of gravity offset is calculated based on the target vehicle's center of gravity before departure and the vehicle's center of gravity during operation, and then compared with the standard center of gravity offset value. If the vehicle's center of gravity offset value is greater than the standard center of gravity offset threshold, there is an objective transportation hazard risk, triggering a deceleration reminder and guiding the vehicle to the nearest service area to adjust the cargo; If the vehicle's center of gravity offset is less than or equal to the standard center of gravity offset threshold, and there is no objective transportation hazard risk, a tire replacement reminder will be triggered.

[0015] Furthermore, the number of objective transportation hazard risk events obtained within the time lookup window includes, Get the preset total number of occurrences of objective transportation hidden risks corresponding to the risk level of each dangerous transportation operation behavior at the starting point of the time search window; The preset total number of occurrences is the sum of the preset number of occurrences of objective transportation hidden risks corresponding to the risk level of each dangerous transportation operation behavior; Collect the number of objective transportation potential risks in the time search window, excluding the number of times the dangerous transportation operation is carried out at the starting point, and record them as the number of times of subsequent objective transportation potential risks. The number of subsequent objective transportation potential risks will be compared with the preset total number of occurrences. If the number of subsequent objective transportation potential risks exceeds the preset total number of occurrences, the first occurrence comparison result is obtained, and the escalation risk coefficient is acquired. If the number of subsequent hazardous transport operations is less than or equal to the preset total number of occurrences, a second occurrence comparison result is obtained. The hazardous transport operation information at the starting point of the aforementioned time lookup window is entered and used as training data for the hazardous transport operation behavior database.

[0016] Furthermore, the upgrade risk coefficient is the ratio of the number of upgrades to the base upgrade amount; The increment number is the difference between the number of subsequent hazardous transport operations and the preset total number of occurrences; The enhancement base is a preset enhancement coefficient corresponding to the hazardous transport operation at the starting point of the time lookup window.

[0017] Compared with existing technologies, the beneficial effects of this invention are as follows: by collecting driver behavior feature vectors in real time and combining them with clustering algorithms to identify dangerous behaviors, the early identification rate of abnormal transportation operations is effectively improved; by combining tire deformation categories and terrain interference analysis, the accuracy of identifying potential objective hazards is further enhanced; and by using a risk contribution judgment method to jointly assess subjective behavior and objective risks, the accuracy of risk level adjustment and response timeliness of the transportation safety assessment system are significantly improved.

[0018] Furthermore, by normalizing the feature vectors and introducing similarity comparisons of dangerous behavior features from historical databases, high-confidence dangerous behavior identification can be achieved in the early stages. At the same time, by combining risk level classification standards, the computational burden can be effectively calculated simply by analyzing the risk level at the starting point of the search window, thereby improving the system's computational efficiency and response speed.

[0019] Furthermore, by jointly identifying tire deformation and road segment interference, the causes of tire anomalies can be accurately located, and the causal relationship between terrain interference and actual tire anomalies can be effectively distinguished; thus, accurate classification of tire deformation categories can be achieved.

[0020] Furthermore, by comparing the tire tread depth during dangerous operations with the initial detection depth, the system can quantitatively analyze tire wear. If the wear exceeds a preset threshold, a real-time warning is issued and the user is guided to a service station for maintenance. If the wear is within the threshold but approaching the threshold, an initial pressure adjustment reminder is issued in a timely manner, thus comprehensively enhancing the system's maintenance reminder strategy and safety control capabilities. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the structure of the deep learning-based road transport vehicle risk level assessment method described in this invention; Figure 2 This is a logic diagram for determining the tire deformation type of a target vehicle within a transportation section based on the operational data, according to an embodiment of the present invention. Figure 3 This is a logic diagram for vehicle operating condition detection in an embodiment of the present invention; Figure 4 This is a logic diagram for obtaining the number of objective transportation hazard risk events within the time lookup window in an embodiment of the present invention. Detailed Implementation

[0022] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0023] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0024] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0025] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0026] Please see Figure 1 The diagram shows a flowchart of a road transport vehicle risk level assessment method based on deep learning, as described in an embodiment of the present invention. The present invention provides a road transport vehicle risk level assessment method based on deep learning, comprising: Real-time collection of driver's transportation operation behavior data and corresponding actual behavior feature vectors; cluster analysis to determine whether there are dangerous transportation operation behaviors in the actual behavior feature vectors; and when they exist, obtaining the dangerous transportation operation behavior category and risk level corresponding to the dangerous transportation operation behavior, as well as collecting the target vehicle's operation data. The tire deformation type of the target vehicle within the transportation section is determined based on the operational data. Based on the tire deformation category judgment result and the vehicle operation stability analysis mode, determine whether there is an objective transportation hazard risk, and if so, judge the contribution of the objective transportation hazard risk to the risk of the dangerous transportation operation behavior. The steps for determining risk contribution include storing transportation operation events and objective transportation potential risk events by timestamp. Using the first time window as the time search window and the dangerous transportation operation behavior as the starting point of the time search window, only the number of objective transportation hidden danger risk events within the time search window is obtained and compared with the standard occurrence number. When the first occurrence comparison result is obtained, the corresponding enhanced risk coefficient is obtained to adjust the risk level of the dangerous transportation operation behavior. In this embodiment, the operational data includes target vehicle pre-departure detection information, target vehicle en route detection information, and slope information of the transportation section. The target vehicle's pre-inspection information includes the contact area of ​​the tires to be inspected, the tread depth of the tires to be inspected, and load distribution information; The target vehicle's on-the-road inspection information includes the tire contact area and tire tread depth detected on the way. The slope information of the transportation route includes the terrain marking information of the transportation route itself; By collecting driver behavior feature vectors in real time and combining them with clustering algorithms to identify dangerous behaviors, the early identification rate of abnormal transportation operations is effectively improved. By combining tire deformation category and terrain interference analysis, the accuracy of identifying potential objective hazards is further enhanced. By using a risk contribution judgment method to jointly assess subjective behavior and objective risks, the accuracy of risk level adjustment and response timeliness of the transportation safety assessment system are significantly improved.

[0027] Specifically, the transportation operation behavior data includes driver posture information and driver eye attention information, and the behavior feature vector is a feature vector that represents the driver posture information and eye attention information after normalization. Cluster analysis is used to determine the category of hazardous transportation operation to which the behavioral feature vector belongs and the risk level of that hazardous transportation operation. Obtain the feature space corresponding to the database of dangerous transport operations and the feature vector of each dangerous transport operation. The actual similarity between the actual behavior feature vector and each dangerous operation behavior feature vector is determined to determine whether dangerous transportation operation behavior exists. Based on the judgment results, the category of dangerous transportation operation behavior is determined, and combined with the risk classification criteria of each risk level within the dangerous transportation operation behavior, the types of dangerous transportation operation behaviors and their corresponding risk levels of the current transportation operation behavior are determined. Specifically, when the determination result is that the danger exists, the corresponding dangerous transportation operation behavior type is obtained, and the risk level is judged. The process of risk level determination is to use a first time window as the time search window, take the detection of dangerous transportation operation behavior as the starting point of the time search window, and only obtain the risk level of dangerous transportation operation behavior at the starting point of the time search window. In this embodiment, the driver's posture information includes head tilt angle, hand position offset distance, and torso offset angle; Driver eye attention information includes eye occlusion assessment, eye gaze direction, and blink duration; The behavioral feature vector is formed by normalizing the above 6 types of raw data; Each feature's original value is normalized to form a one-dimensional feature vector of length 6, for example: [0.87, 0.95, 0.15, 0.92, 0.32, 0.45]. The normalization formula is as follows:

[0028] These are the normalized eigenvalues, and the results are in the interval [0, 1]. These are the original eigenvalues; This is the minimum value of the feature in the sample data; This represents the maximum value of the feature in the sample data. In this embodiment, the head deflection angle range is [0, 90], in degrees; the hand position offset distance range is [0, 20], in centimeters; and the torso offset angle range is [0, 30], in degrees. The eye occlusion judgment is 0 when there is no occlusion and 1 when there is occlusion. The eye occlusion feature vector value is 0 or 1. The eye gaze direction range is [0, 30], in degrees; the blink duration range is [0.1, 0.4], in seconds; The first three items are posture features, and the last three items are eye features; If any feature exceeds its corresponding interval, it is uniformly treated as 1. Cluster analysis methods include database preparation, similarity calculation, and cluster affiliation determination. During the database preparation phase, a feature vector library containing various known dangerous transportation behaviors has been established, with each category of behavior having a set of categorized typical feature vectors. Similarity calculation compares the current behavior feature vector and uses cosine similarity to calculate similarity; the feature vector is normalized; cosine similarity is calculated to determine the most similar dangerous transport operation behavior; combined with the risk level rules of the dangerous transport operation behavior database, the risk level of the current behavior is determined. In this embodiment, the first time window is the time when the dangerous transportation operation is detected, extending 5 minutes forward as a time lookup window; this window is used to observe whether the behavior continues and to make a preliminary judgment on its risk impact on the overall transportation task. By normalizing the feature vectors and introducing similarity comparisons of dangerous behavior features from historical databases, high-confidence dangerous behavior identification can be achieved in the early stages. At the same time, by combining risk level classification standards, the computational burden can be effectively calculated simply by analyzing the risk level at the starting point of the search window, thereby improving the system's computational efficiency and response speed.

[0029] Specifically, determining the actual similarity between the actual behavior feature vector and each dangerous operation behavior feature vector includes, Calculate the similarity scores between the actual behavior feature vector and each dangerous behavior feature vector; Each similarity judgment value is compared with its corresponding similarity threshold, and the existence of dangerous transportation operations is determined based on the comparison results. In this embodiment, the similarity thresholds for head deflection angle, hand position offset distance, torso offset angle, eye occlusion judgment, eye gaze direction, and blink duration obtained from historical data are 0.7, 0.6, 0.9, 1, 0.7, and 0.5, respectively. By employing multi-feature similarity calculation and introducing a threshold determination mechanism, non-key behavioral feature matching is effectively eliminated, reducing the false recognition rate. This enhances the system's ability to understand complex behavioral patterns, helps to accurately extract the feature representation of high-risk behaviors, and improves the accuracy and scalability of behavior classification and judgment.

[0030] See Figure 2 As shown, it is a logic determination diagram of the tire deformation category of the target vehicle in the transportation section based on the operation data in an embodiment of the present invention; Specifically, determining the tire deformation category of the target vehicle within the transportation section based on the operational data includes: Determine the tire deformation of the target vehicle based on tire deformation. Determine whether to perform the transportation section interference identification step based on the tire deformation judgment results; The tire deformation category of the target vehicle under the transportation section is determined by combining the tire deformation and the interference identification and judgment results of the transportation section. Based on the tire deformation category judgment results and the vehicle operation stability analysis mode, it is determined whether there is an objective transportation hazard risk. In this embodiment, determining the tire deformation of the target vehicle based on tire deformation includes: Get the tire deformation variables of the target vehicle under the dangerous transportation operation behavior at the starting point of the time lookup window, and perform tire deformation variable judgment to obtain the tire deformation variable judgment result; Among them, a high-resolution industrial camera is set up to use an edge detection algorithm to extract the outer contour of the tire, so as to determine the shape of the contact area between the target vehicle and the ground during operation and judge the degree of compression deformation, that is, to detect the tire contact area in the road. Determine the difference between the contact area of ​​the tire being inspected en route and the contact area of ​​the tire to be inspected, and compare it with the standard tire contact area threshold; where the standard tire contact area range refers to the normal deformation range of the tire-ground contact area measured when the vehicle is running on a flat transportation section without significant terrain interference. If the difference between the contact area of ​​the tire being tested en route and the contact area of ​​the tire to be tested is within the standard tire contact area range, the first tire deformation category is obtained, and the tire deformation of the target vehicle is normal. If the difference between the contact area of ​​the tire being detected en route and the contact area of ​​the tire to be detected is greater than the maximum value of the standard tire contact area range, a second tire deformation category is obtained. If the tire deformation of the target vehicle is greater than the normal deformation range, the transportation section interference identification step is executed. If the difference between the contact area of ​​the tire being detected en route and the contact area of ​​the tire to be detected is less than the minimum value of the standard tire contact area range, a third tire deformation category is obtained. If the tire deformation of the target vehicle is less than the normal deformation range, the transportation section interference identification step is executed. Determining whether to perform the transportation section interference identification step based on the tire deformation judgment results includes: If the result is the second tire deformation or the third tire deformation, obtain the transportation segment where the dangerous transportation operation occurred at the starting point of the time lookup window, and determine whether the tire deformation is caused by the terrain of the transportation segment based on the terrain marking information of the transportation segment. In this embodiment, the transport segment refers to a continuous road segment area on the vehicle's travel path, which is used to associate the vehicle's operating status with road environment characteristics. The real-time location coordinates of the target vehicle are obtained through GPS or other positioning systems, and the terrain markings of the current transportation section are also obtained, i.e., whether the section is at the top of a slope, the bottom of a slope, or a flat area. If the vehicle is at the top or bottom of a slope, abnormal tire deformation may be due to normal deformation caused by changes in slope and does not need to be judged as abnormal. If the vehicle is in a flat area and the tire deformation is not within the standard tire contact area range, then the tire itself is abnormal, and the cause of the abnormality will be determined through further assessment. In this embodiment, a medium-sized truck is selected, with a total load of approximately 12 tons, equipped with a single 9.00R20 radial tire and a standard tire pressure of approximately 850 kPa. The standard tire contact area range is the total contact area range of all tires in the vehicle. Under the condition of constant speed straight driving of 70-80km / h, the unit contact area of ​​a single tire under this speed and load can be set to approximately 0.038-0.045m². The total contact area range of the 6 wheels of the vehicle is [0.228, 0.27], with the unit being square meters.

[0031] By combining tire deformation and road segment interference identification, the causes of tire abnormalities can be accurately located, and the causal relationship between terrain interference and actual tire abnormalities can be effectively distinguished, thereby achieving accurate classification of tire deformation categories.

[0032] Specifically, determining the existence of objective transportation risks based on the tire deformation category judgment results and vehicle operation stability analysis mode includes: Based on the tire deformation category judgment results, a vehicle operation stability analysis mode is selected, including vehicle operation condition detection and vehicle operation center of gravity shift monitoring. Determine whether there are any objective transportation risks based on vehicle operation stability analysis models; In this embodiment, selecting the vehicle operational stability analysis mode based on the tire deformation category judgment result includes, If the tire deformation category is the second deformation category, perform vehicle operating condition detection on the target vehicle; If the tire deformation category is the third deformation category, monitor the vehicle's center of gravity shift.

[0033] See Figure 3 As shown, it is a logic decision diagram for vehicle operating condition detection in an embodiment of the present invention; Specifically, the vehicle operating condition detection includes dynamic identification of vehicle operating angles and identification of tire wear during vehicle operation, wherein, The process of performing dynamic recognition of vehicle running angle is as follows: The vehicle speed of the target vehicle during the turning process of the dangerous transportation operation behavior at the starting point of the window is obtained, and the centripetal component speed of the vehicle pointing towards the center of the curve is decomposed based on the vehicle speed. The corresponding centripetal angular velocity of the vehicle is calculated based on the turning radius of the target vehicle and compared with the standard centripetal angular velocity threshold. If the vehicle's centripetal angular velocity exceeds the standard centripetal angular velocity threshold and there is no objective transportation hazard risk, a deceleration warning will be triggered. If the vehicle's centripetal angular velocity is less than or equal to the standard centripetal angular velocity threshold, the vehicle's tire wear identification operation will be performed to determine whether there are any objective transportation risks based on the vehicle's tire wear identification results. In this embodiment, the standard centripetal angular velocity threshold value is related to the vehicle model and cargo weight. Under the standard urban road transport conditions of a medium-sized truck, fully loaded with 8 tons of cargo, and a turning radius of 15 meters, the standard centripetal angular velocity threshold is set to 0.25 rad / s. By calculating the centripetal angular velocity of a vehicle when turning and comparing it with a benchmark threshold, it is possible to determine whether the vehicle has a potential risk of slippage due to insufficient turning speed or deterioration of tire performance. Compared with static judgment methods that rely solely on speed or angle, when the angular velocity is determined to be below the threshold, the system can further activate tire wear detection, improve the efficiency of identifying tire abnormalities, reduce missed detections, and enhance the targeting and practicality of the detection.

[0034] Specifically, the operations for identifying tire wear during vehicle operation include: The tire tread depth of the target vehicle at the start of the window during dangerous transportation operations is obtained, along with the tire tread depth of the target vehicle to be detected, and the wear value of the tire tread depth during vehicle operation is calculated. Wherein, the wear value of the tire tread depth of the vehicle is the difference between the tire tread depth of the target vehicle being detected en route and the tire tread depth to be detected, and the tire tread depth being detected en route is the tire tread depth at the start of the window when dangerous transportation operation behavior occurs. Compare the tire tread depth wear value of the vehicle under operation with the standard tire tread wear value. If the tire tread wear value of a vehicle exceeds the standard tire tread wear value, there is an objective transportation hazard risk, triggering a deceleration warning and guiding the vehicle to the nearest service area for vehicle maintenance. If the tire tread depth wear value of the vehicle is less than or equal to the standard tire tread wear value, there is no objective transportation risk, triggering the standard reminder to increase the initial air pressure. In this embodiment, an on-board vision sensor is used to monitor the tread depth of the tire surface in real time; the collected images are processed by image processing algorithms and depth measurement algorithms, and the tread depth is calculated after preprocessing the collected tire tread images. Standard tire tread wear value refers to the maximum allowable wear depth of the tire tread at a specified transport distance. Exceeding this value poses a risk of tire wear. Based on historical data from the database, the formula for calculating the standard tire tread wear value is as follows:

[0035] in, This refers to the standard tire tread wear value. This refers to the maximum permissible wear depth of the tire. This is the wear rate adjustment factor, which is determined based on tire material, road conditions, and road segment factors. For transportation distance; is the base of the natural logarithm; The standard tire tread wear value is related to the vehicle model and cargo weight; in this example, a medium-sized truck is selected with a full load weight of about 12 tons, including 4 tons of tare weight and 8 tons of cargo weight. The vehicle is equipped with radial tires of specification 9.00R20, with a factory tread depth of 15mm. The standard tread wear value for this vehicle under normal driving conditions, with a full load of 8 tons of cargo and an average driving speed of 70–80 km / h, is set at an average wear of 0.04 mm per thousand kilometers.

[0036] By comparing the tire tread depth during dangerous operations with the initial detection depth, the system achieves quantitative analysis of tire wear. If the wear exceeds a preset threshold, a real-time warning is issued and the user is guided to a service station for maintenance. If the wear is within the threshold but approaching the threshold, an initial pressure adjustment prompt is issued in a timely manner, thus comprehensively enhancing the system's maintenance prompt strategy and safety control capabilities.

[0037] Specifically, the process of monitoring vehicle center of gravity shift includes, Obtain the load distribution information from the target vehicle's pre-departure detection information, and obtain the target vehicle's pre-departure center of gravity based on the load distribution information; The vehicle's center of gravity offset is calculated based on the target vehicle's center of gravity before departure and the vehicle's center of gravity during operation, and then compared with the standard center of gravity offset value. If the vehicle's center of gravity offset value is greater than the standard center of gravity offset threshold, there is an objective transportation hazard risk, triggering a deceleration reminder and guiding the vehicle to the nearest service area to adjust the cargo; If the vehicle's center of gravity offset is less than or equal to the standard center of gravity offset threshold, there is no objective transportation hazard risk, and a tire replacement reminder is triggered. In this embodiment, the center of gravity of the target vehicle to be launched is obtained geometrically. The vehicle's center of gravity is obtained using an onboard gyroscope, and the offset value of the vehicle's center of gravity is calculated by combining the target vehicle's center of gravity before departure with the vehicle's current center of gravity. The calculation formula is as follows:

[0038] in, This represents the vehicle's center of gravity offset, expressed in centimeters. The position of the target vehicle's center of gravity in the longitudinal direction during operation; The position of the target vehicle's center of gravity in the longitudinal direction when the vehicle is ready to depart; The position of the target vehicle's center of gravity in the left-right direction during operation; The position of the target vehicle's center of gravity in the longitudinal direction when the vehicle is ready to depart; The standard center of gravity offset value is related to the vehicle model and cargo weight. In this embodiment, the standard center of gravity offset value is taken as 5 cm. By monitoring the center of gravity offset value in real time and comparing it with the standard offset threshold, the system can accurately identify operational instability caused by unreasonable loading structure. It can guide the vehicle into the service area for cargo adjustment in time before potential imbalance occurs, effectively preventing serious accidents such as rollover caused by center of gravity offset and ensuring the safety of the entire vehicle operation.

[0039] See Figure 4 As shown, it is a logic determination diagram for obtaining the number of objective transportation hidden danger risk events within the time lookup window in an embodiment of the present invention; Specifically, the number of objective transportation hazard risk events retrieved within the time lookup window includes: Get the preset total number of occurrences of objective transportation hidden risks corresponding to the risk level of each dangerous transportation operation behavior at the starting point of the time search window; The preset total number of occurrences is the sum of the preset number of occurrences of objective transportation hidden risks corresponding to the risk level of each dangerous transportation operation behavior; Collect the number of objective transportation potential risks in the time search window, excluding the number of times the dangerous transportation operation is carried out at the starting point, and record them as the number of times of subsequent objective transportation potential risks. The number of subsequent objective transportation potential risks will be compared with the preset total number of occurrences. If the number of subsequent objective transportation potential risks exceeds the preset total number of occurrences, the first occurrence comparison result is obtained, and the escalation risk coefficient is acquired. If the number of subsequent hazardous transport operations is less than or equal to the preset total number of occurrences, a second occurrence comparison result is obtained. The information on each hazardous transport operation at the starting point of the aforementioned time lookup window is entered and used as training data for the hazardous transport operation behavior database. In this embodiment, if the dangerous transportation operation behaviors are not unique and are of different categories, the preset occurrence number of each dangerous transportation operation behavior corresponding to the risk level is obtained and summed. If the dangerous transport operation is not unique and the category is the same, or if the dangerous transport operation is unique, obtain the risk level corresponding to each dangerous transport operation, and take the preset occurrence number corresponding to the highest risk level as the preset total occurrence number. If there is only one dangerous transport operation, the preset number of occurrences for the risk level corresponding to that dangerous transport operation is the preset total number of occurrences; Among them, the risk levels corresponding to various dangerous transportation operations are, from top to bottom, Level 1, Level 2, and Level 3 risk levels; The preset occurrence count for Level 1 risk is three times, for Level 2 risk is two times, and for Level 3 risk is one time. The actual total number of occurrences of objective transportation hazards and risks after the start of the time lookup window is obtained in real time and compared with the preset total number of occurrences; By calculating the number of similar behaviors occurring after dangerous operations and comparing them with preset standards, the subsequent evolution trend of risky behaviors can be effectively quantified; the escalation risk coefficient can be quickly obtained to adjust the risk level, reflecting the dynamic risk situation and improving the real-time nature of risk level updates.

[0040] Specifically, the upgrade risk coefficient is the ratio of the number of upgrades to the base upgrade amount; The increment number is the difference between the number of subsequent hazardous transport operations and the preset total number of occurrences; The increment base is the preset total number of occurrences of objective transportation hidden risks corresponding to dangerous transportation operations at the starting point of the time lookup window; In this embodiment, if the risk coefficient is greater than or equal to 0.5, the level of dangerous transportation operation behavior at the starting point of the time search window is adjusted up by one level; if it is the highest level of risk, it is not adjusted. By introducing an upgraded risk coefficient, the quantitative identification of the development trend of dangerous transportation operations over time can dynamically reflect the relative increase in the frequency of risky behaviors. Compared with static threshold judgment, this method has stronger adaptability and foresight, and can effectively identify short-term, high-frequency, and high-intensity dangerous operations, triggering early warning or intervention measures in advance, thereby improving the sensitivity and accuracy of road transport risk assessment.

[0041] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0042] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for assessing the risk level of road transport vehicles based on deep learning, characterized in that, include, Real-time collection of driver's transportation operation behavior data and corresponding actual behavior feature vectors; cluster analysis to determine whether there are dangerous transportation operation behaviors in the actual behavior feature vectors; and when they exist, obtaining the dangerous transportation operation behavior category and risk level corresponding to the dangerous transportation operation behavior, as well as collecting the target vehicle's operation data. The tire deformation type of the target vehicle within the transportation section is determined based on the operational data. Based on the tire deformation category judgment result and the vehicle operation stability analysis mode, determine whether there is an objective transportation hazard risk, and if so, judge the contribution of the objective transportation hazard risk to the risk of the dangerous transportation operation behavior. The steps for determining risk contribution include storing transportation operation events and objective transportation potential risk events by timestamp. Using the first time window as the time search window and the dangerous transportation operation as the starting point of the time search window, only the number of objective transportation hidden danger risk events within the time search window is obtained and compared with the standard occurrence number. When the first occurrence comparison result is obtained, the corresponding enhanced risk coefficient is obtained to adjust the risk level of the dangerous transportation operation.

2. The method for assessing the risk level of road transport vehicles based on deep learning according to claim 1, characterized in that, The transportation operation behavior data includes driver posture information and driver eye attention information, and the behavior feature vector is a normalized feature vector representing the driver posture information and eye attention information. Cluster analysis is used to determine the category of hazardous transportation operation to which the behavioral feature vector belongs and the risk level of that hazardous transportation operation. Obtain the feature space corresponding to the database of dangerous transport operations and the feature vector of each dangerous transport operation. The actual similarity between the actual behavior feature vector and each dangerous operation behavior feature vector is determined to determine whether dangerous transportation operation behavior exists. Based on the judgment results, the category of dangerous transportation operation behavior is determined, and combined with the risk classification criteria of each risk level within the dangerous transportation operation behavior, the types of dangerous transportation operation behaviors and their corresponding risk levels of the current transportation operation behavior are determined. Specifically, when the determination result is that the danger exists, the corresponding dangerous transportation operation behavior type is obtained, and the risk level is judged. The risk level determination process involves using a first time window as the time search window, with the detection of dangerous transportation operations as the starting point of the time search window, and only obtaining the risk level of the dangerous transportation operations at the starting point of the time search window.

3. The method for assessing the risk level of road transport vehicles based on deep learning according to claim 2, characterized in that, Determining the actual similarity between the actual behavior feature vector and the corresponding feature vectors of each dangerous operation behavior includes... Calculate the similarity scores between the actual behavior feature vector and each dangerous behavior feature vector; Each similarity judgment value is compared with its corresponding similarity threshold, and the existence of dangerous transportation operations is determined based on the comparison results.

4. The method for assessing the risk level of road transport vehicles based on deep learning according to claim 1, characterized in that, Based on the operational data, the tire deformation categories of the target vehicles within the transportation section are determined as follows: Determine the tire deformation of the target vehicle based on tire deformation. Determine whether to perform the transportation section interference identification step based on the tire deformation judgment results; By combining the tire deformation and the interference identification results of the transportation section, the tire deformation category of the target vehicle under the transportation section is determined, and based on the tire deformation category determination results and the vehicle operation stability analysis mode, it is determined whether there are any objective transportation hidden risks.

5. The method for assessing the risk level of road transport vehicles based on deep learning according to claim 4, characterized in that, Based on the tire deformation category judgment results and the vehicle operation stability analysis mode, it is determined whether there are objective transportation hidden risks, including: Based on the tire deformation category judgment results, a vehicle operation stability analysis mode is selected, including vehicle operation condition detection and vehicle operation center of gravity shift monitoring. The existence of objective transportation risks is determined based on vehicle operation stability analysis models.

6. The method for assessing the risk level of road transport vehicles based on deep learning according to claim 5, characterized in that, The vehicle operating condition detection includes dynamic identification of vehicle operating angles and identification of tire wear during vehicle operation. The process of performing dynamic recognition of vehicle running angle is as follows: The vehicle speed of the target vehicle during the turning process of the dangerous transportation operation behavior at the starting point of the window is obtained, and the centripetal component speed of the vehicle pointing towards the center of the curve is decomposed based on the vehicle speed. The corresponding centripetal angular velocity of the vehicle is calculated based on the turning radius of the target vehicle and compared with the standard centripetal angular velocity threshold. If the vehicle's centripetal angular velocity exceeds the standard centripetal angular velocity threshold and there is no objective transportation hazard risk, a deceleration warning will be triggered. If the vehicle's centripetal angular velocity is less than or equal to the standard centripetal angular velocity threshold, the vehicle's tire wear identification operation will be performed to determine whether there are any objective transportation hazards based on the tire wear identification results.

7. The method for assessing the risk level of road transport vehicles based on deep learning according to claim 6, characterized in that, The operations for identifying tire wear during vehicle operation include: The tire tread depth of the target vehicle at the start of the window during dangerous transportation operations is obtained, along with the tire tread depth of the target vehicle to be detected, and the wear value of the tire tread depth during vehicle operation is calculated. Wherein, the wear value of the tire tread depth of the vehicle is the difference between the tire tread depth of the target vehicle being detected en route and the tire tread depth to be detected, and the tire tread depth being detected en route is the tire tread depth at the start of the window when dangerous transportation operation behavior occurs. Compare the tire tread depth wear value of the vehicle under operation with the standard tire tread wear value. If the tire tread wear value of a vehicle exceeds the standard tire tread wear value, there is an objective transportation hazard risk, triggering a deceleration warning and guiding the vehicle to the nearest service area for vehicle maintenance. If the tire tread depth wear value of the vehicle is less than or equal to the standard tire tread wear value, and there is no objective transportation hazard risk, the initial tire pressure adjustment standard reminder will be triggered.

8. The method for assessing the risk level of road transport vehicles based on deep learning according to claim 5, characterized in that, The process of monitoring vehicle center of gravity shift includes, Obtain the load distribution information from the target vehicle's pre-departure detection information, and obtain the target vehicle's pre-departure center of gravity based on the load distribution information; The vehicle's center of gravity offset is calculated based on the target vehicle's center of gravity before departure and the vehicle's center of gravity during operation, and then compared with the standard center of gravity offset value. If the vehicle's center of gravity offset value is greater than the standard center of gravity offset threshold, there is an objective transportation hazard risk, triggering a deceleration reminder and guiding the vehicle to the nearest service area to adjust the cargo; If the vehicle's center of gravity offset is less than or equal to the standard center of gravity offset threshold, and there is no objective transportation hazard risk, a tire replacement reminder will be triggered.

9. The method for assessing the risk level of road transport vehicles based on deep learning according to claim 1, characterized in that, The number of objective transportation hazard risk events retrieved within the time lookup window includes: Get the preset total number of occurrences of objective transportation hidden risks corresponding to the risk level of each dangerous transportation operation behavior at the starting point of the time search window; The preset total number of occurrences is the sum of the preset number of occurrences of objective transportation hidden risks corresponding to the risk level of each dangerous transportation operation behavior; Collect the number of objective transportation potential risks in the time search window, excluding the number of times the dangerous transportation operation is carried out at the starting point, and record them as the number of times of subsequent objective transportation potential risks. The number of subsequent objective transportation potential risks will be compared with the preset total number of occurrences. If the number of subsequent objective transportation potential risks exceeds the preset total number of occurrences, the first occurrence comparison result is obtained, and the escalation risk coefficient is acquired. If the number of subsequent hazardous transport operations is less than or equal to the preset total number of occurrences, a second occurrence comparison result is obtained. The hazardous transport operation information at the starting point of the aforementioned time lookup window is entered and used as training data for the hazardous transport operation behavior database.

10. The method for assessing the risk level of road transport vehicles based on deep learning according to claim 9, characterized in that, The upgrade risk coefficient is the ratio of the number of upgrades to the base upgrade amount; The increment number is the difference between the number of subsequent hazardous transport operations and the preset total number of occurrences; The enhancement base is a preset enhancement coefficient corresponding to the hazardous transport operation at the starting point of the time lookup window.

Images