Search

CN-122023288-A - Rail vehicle appearance image quality evaluation method and device and electronic equipment

CN122023288ACN 122023288 ACN122023288 ACN 122023288ACN-122023288-A

Abstract

The invention relates to the technical field of image quality evaluation, and provides a method, a device and electronic equipment for evaluating the appearance image quality of a railway vehicle, wherein the method comprises the steps of obtaining an appearance image to be processed of the railway vehicle; and obtaining the overall evaluation result of the appearance image to be processed based on the multiple dimension evaluation results. The invention realizes the efficient, accurate and visual image quality evaluation of the appearance image to be processed, thereby laying a foundation for the accuracy of subsequent fault diagnosis based on the appearance image to be processed.

Inventors

  • LI CHANGGENG
  • SUN XIAODONG
  • LIU ZHEN
  • BAI JINLING
  • WAN YANMING
  • WANG JIAN

Assignees

  • 中车青岛四方机车车辆股份有限公司

Dates

Publication Date
20260512
Application Date
20260113

Claims (20)

  1. 1. A method for evaluating the quality of an appearance image of a rail vehicle, the method comprising: obtaining an appearance image to be processed of the railway vehicle; Determining dimension evaluation results of the appearance image to be processed under a plurality of image evaluation dimensions based on the appearance image to be processed; And obtaining the overall evaluation result of the appearance image to be processed based on the plurality of dimension evaluation results.
  2. 2. The method of claim 1, wherein the image evaluation dimensions include an image contrast dimension, an image sharpness dimension, an image brightness dimension, and a reference image feature similarity dimension, wherein the reference image feature similarity is used to characterize feature similarity of the appearance image to be processed and a standard reference image: Respectively determining a first weight value of the image contrast dimension, a second weight value of the image definition dimension, a third weight value of the image brightness dimension and a fourth weight value of the similarity dimension with the reference image characteristic; The step of obtaining the overall evaluation result of the appearance image to be processed based on the plurality of dimension evaluation results comprises the following steps: And carrying out weighted summation on the first dimension evaluation result of the image contrast dimension, the first weight value, the second dimension evaluation result of the image definition dimension, the second weight value, the third dimension evaluation result of the image brightness dimension, the third weight value, the fourth dimension evaluation result of the similarity dimension with the reference image characteristic and the fourth weight value to obtain the overall evaluation result of the appearance image to be processed.
  3. 3. The method for evaluating the quality of the apparent image of the railway vehicle according to claim 2, wherein the first weight value of the image contrast dimension, the second weight value of the image definition dimension, the third weight value of the image brightness dimension and the fourth weight value of the similarity dimension with the reference image feature are determined by the following ways: Obtaining a target component type of a target component in the appearance image to be processed; Invoking a preset mapping table, wherein the mapping table comprises corresponding relations between different component types and different weight value duty ratio distributions, and the weight value duty ratio distributions are used for representing duty ratio conditions of the first weight value, the second weight value, the third weight value and the fourth weight value; Obtaining a target weight value duty ratio distribution matched with the target component type based on the target component type and the corresponding relation in the mapping table; and determining a first weight value of the image contrast dimension, a second weight value of the image definition dimension, a third weight value of the image brightness dimension and a fourth weight value of the similarity dimension with the reference image characteristic based on the target weight value duty ratio distribution.
  4. 4. The method according to claim 2, characterized in that after the overall evaluation result of the appearance image to be processed is obtained based on the plurality of the dimensional evaluation results, the method further comprises: Mapping the first dimension evaluation result of the image contrast dimension, the second dimension evaluation result of the image definition dimension, the third dimension evaluation result of the image brightness dimension and the fourth dimension evaluation result of the similarity dimension with the reference image feature to four axial directions of a radar chart so as to intuitively display a plurality of dimension evaluation results of the appearance image to be processed.
  5. 5. The method according to any one of claims 1 to 4, wherein the dimension evaluation result includes a first dimension evaluation result of an image contrast dimension, wherein determining, based on the appearance image to be processed, the dimension evaluation result of the appearance image to be processed in the image contrast dimension includes: obtaining gray values of all pixel points in the appearance image to be processed based on the appearance image to be processed; averaging processing is carried out based on the gray values of all the pixel points, so that the image average gray value of the appearance image to be processed is obtained; performing gray value standard deviation processing on the appearance image to be processed based on the gray value of each pixel point and the image average gray value of the appearance image to be processed to obtain the gray value standard deviation of the appearance image to be processed; and determining a first dimension evaluation result of the appearance image to be processed under the dimension of the image contrast based on the gray value standard deviation of the appearance image to be processed.
  6. 6. The method according to claim 5, wherein before the gray value standard deviation of the appearance image to be processed is performed based on the gray value of each pixel and the image average gray value of the appearance image to be processed, the method further comprises: Acquiring the number of first pixels of the appearance image to be processed in the image width direction and the number of second pixels of the appearance image to be processed in the image height direction; The gray value standard deviation processing of the appearance image to be processed is performed based on the gray value of each pixel point and the image average gray value of the appearance image to be processed, so as to obtain the gray value standard deviation of the appearance image to be processed, and the method comprises the following steps: and carrying out gray value standard deviation processing on the appearance image to be processed based on the first pixel point number, the second pixel point number, the gray value of each pixel point and the image average gray value of the appearance image to be processed, so as to obtain the gray value standard deviation of the appearance image to be processed.
  7. 7. The method for evaluating the quality of the appearance image of the railway vehicle according to any one of claims 1 to 4, wherein the dimension evaluation result includes a second dimension evaluation result of an image definition dimension, wherein determining the dimension evaluation result of the appearance image to be processed in the image definition dimension based on the appearance image to be processed includes: obtaining gray values of all pixel points in the appearance image to be processed based on the appearance image to be processed; averaging processing is carried out based on the gray values of all the pixel points, so that the image average gray value of the appearance image to be processed is obtained; carrying out Laplace calculation on the gray value of each pixel point to obtain the gray value of the pixel point after Laplace calculation; carrying out Laplace calculation on the image average gray value of the appearance image to be processed to obtain the image average gray value after the Laplace calculation; Carrying out gray value Laplacian standard deviation processing on the appearance image to be processed based on the pixel gray value after the Laplacian calculation processing and the average gray value of the image after the Laplacian calculation processing to obtain the gray value Laplacian standard deviation of the appearance image to be processed; and determining a second dimension evaluation result of the appearance image to be processed under the dimension of image definition based on the Laplacian standard deviation of the gray value of the appearance image to be processed.
  8. 8. The method according to claim 7, characterized in that before the gray value laplace standard deviation processing of the gray value of the appearance image to be processed is performed based on the pixel gray value after the laplace calculation processing and the average gray value of the image after the laplace calculation processing, the method further comprises: Acquiring the number of first pixels of the appearance image to be processed in the image width direction and the number of second pixels of the appearance image to be processed in the image height direction; The step of carrying out the gray value Laplacian standard deviation processing on the gray value of the pixel point after the Laplacian calculation processing and the average gray value of the image after the Laplacian calculation processing to obtain the gray value Laplacian standard deviation of the appearance image to be processed comprises the following steps: And carrying out gray value Laplacian standard deviation processing on the appearance image to be processed based on the first pixel number, the second pixel number, the pixel gray value after the Laplacian calculation processing and the image average gray value after the Laplacian calculation processing to obtain the gray value Laplacian standard deviation of the appearance image to be processed.
  9. 9. The method according to any one of claims 1 to 4, wherein the dimension evaluation result includes a third dimension evaluation result of an image brightness dimension, wherein determining the dimension evaluation result of the appearance image to be processed in the image brightness dimension based on the appearance image to be processed includes: obtaining gray values of all pixel points in the appearance image to be processed based on the appearance image to be processed; averaging processing is carried out based on the gray values of all the pixel points, so that the image average gray value of the appearance image to be processed is obtained; and determining a third dimension evaluation result of the appearance image to be processed under the image brightness dimension based on the image average gray value of the appearance image to be processed.
  10. 10. The method for evaluating the quality of an exterior image of a railway vehicle according to claim 9, characterized in that before the averaging process is performed based on the gradation value of each pixel point to obtain the image average gradation value of the exterior image to be processed, the method further comprises: Acquiring the number of first pixels of the appearance image to be processed in the image width direction and the number of second pixels of the appearance image to be processed in the image height direction; The average processing is performed based on the gray value of each pixel point to obtain the image average gray value of the appearance image to be processed, which comprises the following steps: and carrying out averaging processing based on the number of the first pixel points, the number of the second pixel points and the gray values of the pixel points to obtain the image average gray value of the appearance image to be processed.
  11. 11. The method according to any one of claims 1 to 4, wherein the dimension evaluation result includes a fourth dimension evaluation result of a feature similarity dimension with a reference image, wherein determining, based on the appearance image to be processed, a dimension evaluation result of the appearance image to be processed in the feature similarity dimension with the reference image includes: A reference image feature library is built in advance, wherein the reference image feature library comprises reference image features of reference images under different feature dimensions, and the reference images are images with image quality meeting preset requirements; Obtaining the to-be-processed image characteristics of the to-be-processed appearance image under different characteristic dimensions based on the to-be-processed appearance image; Performing similarity matching on the image features to be processed and the reference image features to obtain feature similarity of the image features to be processed and the reference image features; And determining a fourth-dimension evaluation result of the appearance image to be processed under the dimension of the feature similarity with the reference image based on the feature similarity.
  12. 12. The method for evaluating the quality of the appearance image of the railway vehicle according to claim 11, wherein the reference image features of the reference image in different feature dimensions comprise a reference image feature in a low-dimensional feature, a reference image feature in a medium-dimensional feature and a reference image feature in a high-dimensional feature, the to-be-processed image features of the appearance image in different feature dimensions comprise a to-be-processed image feature in a low-dimensional feature, a to-be-processed image feature in a medium-dimensional feature and a to-be-processed image feature in a high-dimensional feature, and the matching the to-be-processed image feature with the reference image feature to obtain the feature similarity between the to-be-processed image feature and the reference image feature comprises: matching the similarity between the reference image feature under the low-dimensional feature and the image feature to be processed under the low-dimensional feature to obtain a first feature similarity between the image feature to be processed and the reference image feature under the low-dimensional feature; matching the similarity between the reference image feature under the middle-dimensional feature and the image feature to be processed under the middle-dimensional feature to obtain a second feature similarity between the image feature to be processed and the reference image feature under the middle-dimensional feature; matching the similarity between the reference image feature under the high-dimensional feature and the image feature to be processed under the high-dimensional feature to obtain a third feature similarity between the image feature to be processed and the reference image feature under the high-dimensional feature; and obtaining the feature similarity of the image feature to be processed and the reference image feature based on the first feature similarity, the second feature similarity and the third feature similarity.
  13. 13. The method for evaluating the quality of the apparent image of the railway vehicle according to claim 11, wherein the reference image features of the reference image in different feature dimensions are obtained by: invoking a pre-trained image feature recognition model, wherein the image feature recognition model is used for extracting reference image features of a reference image under different feature dimensions based on the reference image; Inputting the reference image into the image feature recognition model, and extracting a plurality of reference image features of the reference image under different feature dimensions; Clustering a plurality of reference image features under any feature dimension to obtain clustered reference image features; And taking the clustered reference image characteristics as the reference image characteristics of the reference image in the characteristic dimension.
  14. 14. The method for evaluating the quality of the exterior image of the railway vehicle according to claim 13, wherein the step of inputting the reference image into the image feature recognition model and extracting a plurality of reference image features of the reference image in different feature dimensions comprises: inputting the reference image into the image feature recognition model, and extracting a plurality of reference image features of the reference image under low-dimensional features based on a low-dimensional feature layer in the image feature recognition model; Inputting the reference image into the image feature recognition model, and extracting a plurality of reference image features of the reference image under the medium-dimensional features based on the medium-dimensional feature layer in the image feature recognition model; and inputting the reference image into the image feature recognition model, and extracting a plurality of reference image features of the reference image under high-dimensional features based on a high-dimensional feature layer in the image feature recognition model.
  15. 15. The method according to claim 13 or 14, characterized in that before the inputting the reference image to the image feature recognition model, extracting a plurality of reference image features of the reference image in different feature dimensions, the method further comprises: carrying out format standardization processing on the reference image to obtain a reference image with a preset format; The step of inputting the reference image into the image feature recognition model, extracting a plurality of reference image features of the reference image under different feature dimensions, comprises the following steps: inputting the reference image with the preset format into the image feature recognition model, and extracting a plurality of reference image features of the reference image under different feature dimensions.
  16. 16. The method for evaluating the quality of an appearance image of a railway vehicle according to claim 1, characterized in that after the overall evaluation result of the appearance image to be processed is obtained, the method further comprises: under the condition that the overall evaluation result meets the preset requirement, performing fault detection on the railway vehicle based on the appearance image to be processed; And under the condition that the overall evaluation result does not meet the preset requirement, initiating an alarm prompt, wherein the alarm prompt is used for prompting a user to acquire the appearance image to be processed of the railway vehicle again, so that under the condition that the overall evaluation result of the acquired appearance image to be processed meets the preset requirement, fault detection is carried out on the railway vehicle based on the appearance image to be processed.
  17. 17. An apparatus for evaluating the visual image quality of a railway vehicle, comprising: the acquisition module is used for acquiring an appearance image to be processed of the railway vehicle; The determining module is used for determining dimension evaluation results of the appearance image to be processed under a plurality of image evaluation dimensions based on the appearance image to be processed; And the evaluation module is used for obtaining the overall evaluation result of the appearance image to be processed based on the plurality of dimension evaluation results.
  18. 18. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the rail vehicle appearance image quality assessment method of any one of claims 1 to 16 when the computer program is executed.
  19. 19. A non-transitory computer-readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the rail vehicle appearance image quality evaluation method according to any one of claims 1 to 16.
  20. 20. A computer program product comprising a computer program which, when executed by a processor, implements the rail vehicle appearance image quality evaluation method according to any one of claims 1 to 16.

Description

Rail vehicle appearance image quality evaluation method and device and electronic equipment Technical Field The present invention relates to the field of image quality evaluation technologies, and in particular, to a method and an apparatus for evaluating an appearance image quality of a railway vehicle, and an electronic device. Background In an intelligent maintenance system of rail transit, a trackside system based on visual detection has become a core technical platform for monitoring the states of key components of a motor train unit. The system collects real-time image data of the components such as the bogie, the pantograph, the vehicle body connecting part and the like through the high-precision camera device arranged beside the track when the vehicle passes at high speed, and provides an original basis for subsequent fault diagnosis, life prediction and operation and maintenance decision. However, multiple factors in the actual operating environment may cause significant degradation in image quality, and performing fault diagnosis based on low-quality appearance images may impair the accuracy of subsequent fault diagnosis. Therefore, finding a method for evaluating image quality efficiently, accurately and intuitively has become a current research hotspot. Disclosure of Invention The invention provides a method, a device and electronic equipment for evaluating the quality of an appearance image of a railway vehicle, which realize the efficient, accurate and visual image quality evaluation of the appearance image to be processed, thereby laying a foundation for the accuracy of subsequent fault diagnosis based on the appearance image to be processed. The invention provides a method for evaluating the appearance image quality of a railway vehicle, which comprises the steps of obtaining an appearance image to be processed of the railway vehicle, determining dimension evaluation results of the appearance image to be processed under a plurality of image evaluation dimensions based on the appearance image to be processed, and obtaining an overall evaluation result of the appearance image to be processed based on the plurality of dimension evaluation results. According to the method for evaluating the quality of the appearance image of the railway vehicle, the image evaluation dimension comprises an image contrast dimension, an image definition dimension, an image brightness dimension and a reference image feature similarity dimension, wherein the reference image feature similarity is used for representing the feature similarity of the appearance image to be processed and a standard reference image; before the overall evaluation result of the appearance image to be processed is obtained based on the multiple dimension evaluation results, the method further comprises the steps of respectively determining a first weight value of the image contrast dimension, a second weight value of the image definition dimension, a third weight value of the image brightness dimension and a fourth weight value of the similarity dimension with the reference image characteristic, obtaining the overall evaluation result of the appearance image to be processed based on the multiple dimension evaluation results, and weighting the first dimension evaluation result of the image contrast dimension, the first weight value, the second dimension evaluation result of the image definition dimension, the second weight value, the third dimension evaluation result of the image brightness dimension, the third weight value, the fourth dimension evaluation result of the similarity dimension with the reference image characteristic and the fourth weight value to obtain the overall evaluation result of the appearance image to be processed. According to the method for evaluating the appearance image quality of the railway vehicle, the first weight value of the image contrast dimension, the second weight value of the image definition dimension, the third weight value of the image brightness dimension and the fourth weight value of the image brightness dimension are determined, the method is obtained by acquiring the target component type of the target component in the appearance image to be processed, calling a preset mapping table, wherein the mapping table comprises corresponding relations between different component types and different weight value proportion distributions, the weight value proportion distributions are used for representing the proportion conditions of the first weight value, the second weight value, the third weight value and the fourth weight value, and the corresponding relation between the target component type and the mapping table is used for obtaining the proportion distribution of the target weight value matched with the target component type, and the corresponding relation between the weight value proportion distribution of the first weight value, the second weight value, the third weight value and the image brightness