Road image intelligent acquisition and identification method based on deep learning

Abstract

The invention provides a road image intelligent acquisition and identification method based on deep learning. The road image intelligent acquisition and identification method comprises the following steps: shooting a road video image through road image acquisition hardware, extracting key frames in the video image through a program, detecting and identifying the key frames through an image intelligent identification algorithm, and determining a disease position, a disease category and a disease grade in the road image. According to the invention, positioning of the road disease area and classification tasks of types and grades can be completed, and manual intervention is not needed, and unmanned operation is completely realized, and the automation degree is high, and the fault tolerance ishigh, and the road maintenance patrol efficiency can be greatly improved, and the intelligent management level of the industry is improved.

Description

technical field [0001] The invention relates to the technical fields of video images and computer vision, in particular to a method for intelligent collection and recognition of road images based on deep learning. Background technique [0002] Target detection and recognition in images has a wide range of practical needs in applications such as intelligent video surveillance, and it is also a popular development direction in the field of computer vision. Convolutional neural network (CNN) has a strong advantage in extracting deep-level features of images by performing operations such as convolution and pooling on images, and is widely used in the field of computer vision. [0003] The current mainstream deep learning-based target detection algorithms include Faster-RCNN, SSD, YOLO, etc. The two-stage algorithm Faster-RCNN has precise positioning and recognition effects. However, when the number of categories is large, the model training time is long, the recognition speed d...

AI Technical Summary

Problems solved by technology

The two-stage algorithm Faster-RCNN has precise positioning and recognition effects. However, when the number of categories is large, the model training time is long, the recognition speed decreases proportionally, and the false recognition rate also increases.

The one-stage algorithm SSD and YOLO have significantly improved the recognition speed compared with the two-stage algorithm, but their positioning is not accurate enough, the detection rate of small objects is low, it is easy to miss detection, and the recognition accuracy is lower than Faster-RCNN

Images