Search

CN-122024213-A - Vehicle supervision method and system

CN122024213ACN 122024213 ACN122024213 ACN 122024213ACN-122024213-A

Abstract

The invention discloses a vehicle supervision method and a system, which relate to the technical field of vehicle supervision, wherein a synchronous acquisition step simultaneously acquires an external environment of a vehicle and a state image of a driver in a hardware synchronous mode, an intelligent routing mechanism based on dynamic image quality evaluation is introduced, the acquired external image is subjected to rapid quality evaluation, an enhancement flow is started only for an image which does not reach the standard, a loss function comprises a task-oriented perception loss item, the enhanced image is driven to be close to a clear license plate image in a feature space, the visual quality of the image is improved, and the feature representation of the image for subsequent license plate recognition is more directly optimized, so that the recognition accuracy of license plate characters is obviously improved in complex environments such as low illumination, the crossing of comprehensive situation judgment from single event detection is realized by simultaneously executing license plate recognition and driver behavior analysis, and the accuracy of early warning information is improved.

Inventors

  • WANG HUAIYU

Assignees

  • 无锡怀瑾握瑜科技有限公司

Dates

Publication Date
20260512
Application Date
20251225

Claims (9)

  1. 1. A vehicle supervision method is characterized by comprising the following steps of, S1, synchronously acquiring an external environment image of a vehicle and a driver state image of a driver; S2, positioning a candidate region and evaluating quality based on the characteristics of a license plate region in the external environment image, and routing the image of the candidate region to a direct recognition path or an enhancement processing path according to a quality evaluation result; s3, processing the candidate area images routed to the enhancement processing path by adopting an image enhancement model taking the improvement of the accuracy rate of the subsequent license plate recognition as an optimization target to obtain an enhanced image; s4, license plate recognition is carried out on the candidate area image or the enhanced image from the direct recognition path to obtain a first recognition result; s5, combining the first recognition result and the second recognition result, analyzing the risk correlation between the first recognition result and the second recognition result, and calculating to obtain a comprehensive risk value representing the risk degree of the current supervision scene through a risk assessment model; And S6, triggering external responses of different grades according to the preset range of the comprehensive risk value.
  2. 2. The method of vehicle supervision as set forth in claim 1, wherein the evaluating the quality of the candidate region includes analyzing a composite quality score of the candidate region, the composite quality score being determined by a combination of brightness uniformity, edge sharpness, and color saturation of the candidate region, and comparing the composite quality score to a dynamic threshold.
  3. 3. The method for supervising the vehicle according to claim 1, wherein in S3, the loss function used for training includes a task-oriented loss term for restricting the representation of the enhanced image in a predetermined license plate recognition network feature space with respect to the image enhancement model with the enhancement of the accuracy of the subsequent license plate recognition as the optimization target.
  4. 4. The method of vehicle supervision as set forth in claim 1, wherein in S4, the first recognition result includes obtaining license plate number information and a corresponding first confidence level, and the second recognition result includes obtaining a driver behavior category and a corresponding second confidence level.
  5. 5. The method of vehicle supervision as set forth in claim 4, wherein the step of analyzing the risk correlation between the two includes determining whether the risk event indicated by the first recognition result and the risk event indicated by the second recognition result have a correlation in time series, and generating a correlation indicator.
  6. 6. The vehicle supervision method according to claim 5, wherein the risk assessment model analyzes based on the first confidence level, the second confidence level, and the relevance indicator to generate the composite risk value having a value range between 0 and 1.
  7. 7. A vehicle supervision system based on the vehicle supervision method according to any one of claims 1 to 6, characterized by comprising, The synchronous acquisition module synchronously acquires an external environment image of the vehicle and a driver state image of a driver; the quality routing module is used for positioning candidate areas and evaluating quality based on the characteristics of license plate areas in the external environment images, and routing the images of the candidate areas to a direct recognition path or an enhancement processing path according to quality evaluation results; The task enhancement module is used for processing the candidate area images routed to the enhancement processing path by adopting an image enhancement model taking the improvement of the accuracy rate of the subsequent license plate recognition as an optimization target to obtain an enhanced image; the parallel analysis module is used for carrying out license plate recognition on the candidate area image or the enhanced image from the direct recognition path to obtain a first recognition result; the decision module is used for combining the first identification result and the second identification result, analyzing the risk correlation between the first identification result and the second identification result, and calculating to obtain a comprehensive risk value representing the risk degree of the current supervision scene through a risk assessment model; And the execution module triggers external responses of different grades according to the preset range of the comprehensive risk value.
  8. 8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the vehicle supervision method as claimed in any one of claims 1 to 6.
  9. 9. A computer-readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the vehicle supervision method as claimed in any one of the claims 1 to 6.

Description

Vehicle supervision method and system Technical Field The invention relates to the technical field of vehicle supervision, in particular to a vehicle supervision method and system. Background The vehicle supervision technology is taken as a core component of the intelligent traffic system, and is developed along with the breakthrough of computer vision and artificial intelligence technology. Modern vehicle supervision systems mainly rely on high-definition camera devices deployed at road key nodes or on vehicles themselves to continuously acquire vehicle external environment images and in-vehicle driver state images. The license plate recognition system and the driver behavior monitoring system of the existing vehicle supervision technology are usually designed and deployed as independent modules, effective information interaction is lacking among the systems, correlation analysis cannot be carried out on vehicle identity information and driver behavior states, the existing license plate recognition technology is severely dependent on imaging quality, and recognition accuracy is drastically reduced due to image degradation under complex environments such as low illumination, strong backlight, rainy and snowy weather and the like. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention provides a vehicle supervision method for solving the problems of unclear vehicle identification and inaccurate risk judgment. In order to solve the technical problems, the invention provides the following technical scheme: in a first aspect, the present invention provides a vehicle supervision method comprising, S1, synchronously acquiring an external environment image of a vehicle and a driver state image of a driver; S2, positioning a candidate region and evaluating quality based on the characteristics of a license plate region in the external environment image, and routing the image of the candidate region to a direct recognition path or an enhancement processing path according to a quality evaluation result; s3, processing the candidate area images routed to the enhancement processing path by adopting an image enhancement model taking the improvement of the accuracy rate of the subsequent license plate recognition as an optimization target to obtain an enhanced image; s4, license plate recognition is carried out on the candidate area image or the enhanced image from the direct recognition path to obtain a first recognition result; s5, combining the first recognition result and the second recognition result, analyzing the risk correlation between the first recognition result and the second recognition result, and calculating to obtain a comprehensive risk value representing the risk degree of the current supervision scene through a risk assessment model; And S6, triggering external responses of different grades according to the preset range of the comprehensive risk value. In a preferred embodiment of the vehicle supervision method according to the present invention, in S2, the evaluating the quality includes analyzing a composite quality score of the candidate region, the composite quality score being determined by a luminance uniformity, an edge sharpness, and a color saturation of the candidate region, and comparing the composite quality score with a dynamic threshold. In the S3, the loss function used for training comprises a task guide loss term used for restricting the representation of the enhanced image in a preset license plate recognition network feature space by taking the improvement of the accuracy rate of the subsequent license plate recognition as an image enhancement model of an optimization target. In the S4, the first recognition result comprises license plate number information and corresponding first confidence coefficient, and the second recognition result comprises driver behavior category and corresponding second confidence coefficient. In the preferred scheme of the vehicle supervision method, in the S5, analyzing the risk correlation between the first recognition result and the second recognition result comprises judging whether the risk event indicated by the first recognition result and the risk event indicated by the second recognition result have correlation on a time sequence or not, and generating a correlation indicating factor. As a preferable scheme of the vehicle supervision method, the risk assessment model analyzes according to the first confidence coefficient, the second confidence coefficient and the relevance indicating factor to generate the comprehensive risk value with the value range between 0 and 1. In a second aspect, the present invention provides a vehicle supervision system comprising, The synchronous acquisition module synchronously acquires an external environment image of the vehicle and a driver state image of a driver; the quality routing module is used for positioning candi