CN-121982668-A - Road image processing method, device, computer equipment and storage medium

CN121982668ACN 121982668 ACN121982668 ACN 121982668ACN-121982668-A

Abstract

The invention discloses a road image processing method, a device, computer equipment and a storage medium. The method comprises the steps of obtaining a plurality of road images to be processed and field contents corresponding to each road image to be processed, carrying out semantic candidate retrieval on the plurality of road images to be processed to determine a plurality of first candidate images, screening the plurality of first candidate images according to space-time consistency and category consistency based on the field contents corresponding to the plurality of first candidate images to determine a plurality of second candidate images, carrying out image clustering processing on the plurality of second candidate images, and determining a plurality of road image clusters and clusters corresponding to each road image cluster to represent road images. The method carries out duplicate removal screening treatment on the road image to be treated by a semantic candidate retrieval, space-time and category consistency screening and multiple screening modes of image clustering, and can achieve the purpose of accurately removing duplicate of the road image to be treated.

Inventors

ZHANG LEICHAO
WU JINYONG
YU XIAOTIAN
LI AIJUN
CHEN NING

Assignees

深圳云天励飞技术股份有限公司
成都天舟锦成科技有限公司
成都芯星励飞机器人技术有限公司

Dates

Publication Date: 20260505
Application Date: 20251226

Claims (10)

1. A road image processing method, characterized by comprising: acquiring a plurality of road images to be processed and field contents corresponding to each road image to be processed; Carrying out semantic candidate retrieval on a plurality of road images to be processed, and determining a plurality of first candidate images; screening the plurality of first candidate images according to space-time and category consistency based on field contents corresponding to the plurality of first candidate images, and determining a plurality of second candidate images; And carrying out image clustering processing on the plurality of second candidate images, and determining a plurality of road image clusters and clusters corresponding to each road image cluster to represent road images.
2. The road image processing method according to claim 1, wherein the acquiring a plurality of road images to be processed and field contents corresponding to each of the road images to be processed includes: Acquiring a plurality of original road images; preprocessing a plurality of original road images to obtain a road image to be processed corresponding to each original road image; and carrying out field marking on the plurality of road images to be processed to obtain the field content of each road image to be processed.
3. The method for processing a road image according to claim 1, wherein the performing semantic candidate retrieval on the plurality of road images to be processed to determine a plurality of first candidate images includes: Carrying out semantic feature recognition on a plurality of road images to be processed, and determining global semantic vectors corresponding to each road image to be processed; and carrying out standard vector distance retrieval processing based on global semantic vectors corresponding to any two road images to be processed, and determining a plurality of first candidate images.
4. The road image processing method according to claim 1, wherein the field contents include an image longitude, an image latitude, a road density, and an image category; The step of screening the plurality of first candidate images according to space-time and category consistency based on field contents corresponding to the plurality of first candidate images to determine a plurality of second candidate images, including: Performing distance calculation based on the image longitude and the image latitude between any two first candidate images, and determining the geographic neighbor distance between any two first candidate images; determining an adaptive neighbor radius based on the road density corresponding to the first candidate image; And screening the first candidate images based on the image category, the geographic neighbor distance and the self-adaptive neighbor radius, and determining a plurality of second candidate images corresponding to the image category.
5. The method according to claim 1, wherein the performing image clustering processing on the plurality of second candidate images to determine a plurality of road image clusters and clusters corresponding to each of the road image clusters represent road images, comprises: Matching any two second candidate images to determine the edge connecting weight corresponding to any two second candidate images; And carrying out image clustering processing on the plurality of second candidate images based on the continuous edge weights corresponding to any two second candidate images, and determining a plurality of image clusters and clusters corresponding to each image cluster to represent road images.
6. The method for processing a road image according to claim 5, wherein the matching processing of any two of the second candidate images to determine the edge weights corresponding to any two of the second candidate images includes: Processing any two second candidate images in the same image category by adopting a preset dense feature matching model, and determining a pixel point confidence coefficient map corresponding to a pixel point set of the any two second candidate images; determining the matching number of the pixels, the confidence mean value of the pixels and the coverage rate of matched pixels corresponding to any two second candidate images based on the pixel confidence map; Obtaining a geographic neighbor distance and a time difference between any two second candidate images; and determining the edge connecting weight corresponding to any two second candidate images based on the pixel point matching quantity, the pixel point confidence coefficient mean value, the matching pixel coverage rate, the geographic neighbor distance and the time difference corresponding to any two second candidate images.
7. The method according to claim 5, wherein the performing image clustering on the plurality of second candidate images based on the bordering weights corresponding to any two second candidate images, determining a plurality of image clusters and a cluster corresponding to each image cluster represents a road image, includes: Performing graph construction based on the continuous edge weights corresponding to any two second candidate images, and determining a target continuous edge graph; sub-graph extraction is carried out on the target edge graph, and a plurality of image clusters are determined; And selecting the image of each image cluster by adopting a preset image selection standard to obtain a cluster representing road image corresponding to each image cluster.
8. A road image processing apparatus, characterized by comprising: the image and field content acquisition module is used for acquiring a plurality of road images to be processed and field content corresponding to each road image to be processed; the first candidate image determining module is used for carrying out semantic candidate retrieval on the plurality of road images to be processed and determining a plurality of first candidate images; the second candidate image determining module is used for screening the plurality of first candidate images according to the consistency of space time and category based on field contents corresponding to the plurality of first candidate images to determine a plurality of second candidate images; And the image clustering processing module is used for carrying out image clustering processing on the plurality of second candidate images and determining a plurality of road image clusters and clusters corresponding to each road image cluster to represent road images.
9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the road image processing method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the road image processing method according to any one of claims 1 to 7.

Description

Road image processing method, device, computer equipment and storage medium Technical Field The present invention relates to the field of image processing technologies, and in particular, to a road image processing method, a road image processing device, a computer device, and a storage medium. Background Road damage refers to damage and deformation of a road surface caused by various factors during the use of the road, and common road damage includes cracks, pits, ruts, and the like. The road image is usually obtained by collecting the image of the road disease, the type of the road disease is determined by identifying and analyzing the road image, and the road is repaired and maintained according to the type of the road disease. In the prior art, in the process of collecting images of road diseases, a large number of repeated road images are usually collected, so that the recognition efficiency of road image recognition is reduced. Therefore, how to perform the de-duplication processing on the collected road image is a technical problem to be solved currently. Disclosure of Invention The embodiment of the invention provides a road image processing method, a device, computer equipment and a storage medium, which are used for solving the technical problem of how to perform de-duplication processing on an acquired road image. A road image processing method, comprising: acquiring a plurality of road images to be processed and field contents corresponding to each road image to be processed; Carrying out semantic candidate retrieval on a plurality of road images to be processed, and determining a plurality of first candidate images; screening the plurality of first candidate images according to space-time and category consistency based on field contents corresponding to the plurality of first candidate images, and determining a plurality of second candidate images; And carrying out image clustering processing on the plurality of second candidate images, and determining a plurality of road image clusters and clusters corresponding to each road image cluster to represent road images. A road image processing apparatus comprising: the image and field content acquisition module is used for acquiring a plurality of road images to be processed and field content corresponding to each road image to be processed; the first candidate image determining module is used for carrying out semantic candidate retrieval on the plurality of road images to be processed and determining a plurality of first candidate images; the second candidate image determining module is used for screening the plurality of first candidate images according to the consistency of space time and category based on field contents corresponding to the plurality of first candidate images to determine a plurality of second candidate images; And the image clustering processing module is used for carrying out image clustering processing on the plurality of second candidate images and determining a plurality of road image clusters and clusters corresponding to each road image cluster to represent road images. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the road image processing method described above when executing the computer program. A computer-readable storage medium storing a computer program which, when executed by a processor, implements the road image processing method described above. According to the road image processing method, the device, the computer equipment and the storage medium, firstly, semantic candidate retrieval is carried out on a plurality of road images to be processed in the aspect of semantic difference, and a plurality of first candidate images with larger semantic difference are accurately screened out. And performing de-duplication screening on the plurality of first candidate images according to the field content corresponding to each first candidate image and the consistency of space-time and category and in the space-time and category angles of image acquisition, and determining a plurality of second candidate images. And finally, carrying out image clustering on the plurality of second candidate images to obtain a plurality of road image clusters, further precisely screening the second candidate images in the plurality of road image clusters, and determining that the clusters corresponding to each road image cluster represent road images. The method carries out duplicate removal screening treatment on the road image to be treated through semantic candidate retrieval, space-time and category consistency screening and multiple screening modes of image clustering, and can achieve the purpose of precisely duplicate removal of the road image to be treated, so that when road disease identification is carried out according to a plurality of road image clusters after precise duplicate removal and the road image represented by