With the continuous development of intelligent transportation systems, automatic detection and recognition of road images has become an important aspect of current urban traffic management and research on autonomous vehicles. The purpose of this study is to establish an image-based road detection and recognition system using computer vision technology. This article conducts research and analysis on existing road detection technologies to identify their limitations in various road conditions. It studies a multi-layer feature extraction and analysis method based on deep learning to improve the accuracy of road image detection and recognition. It conducts algorithm optimization research to improve the generalization ability and operational efficiency of the model by improving the network structure and adjusting parameters. The proposed method can be experimentally validated on real road surfaces to evaluate its correctness and reliability. The performance of the model in this article on different types of roads (urban, suburban) shows a success rate of 85% for urban navigation, an accuracy rate of 90% for road edge detection, and an accuracy rate of 92% for obstacle recognition. The image road detection and recognition algorithm based on machine vision in this article helps to improve the accuracy of image road detection and enhance road safety.