论文部分内容阅读
研究开发了一套基于机器视觉的路面质量检测系统,并将改进的神经网络方法应用到道路路面缺陷检测中。分析了系统的基本组成和总体结构,介绍了软件设计流程以及网络设计与训练过程。同时考虑到传统图像分割算法的局限性,设计了一种检测图像内任意区域实时检测算法。可以适应路面龟裂、横裂、纵裂、块裂等多种缺陷以及缺陷并存的复杂道路样本图像。该检测系统具有很强的灵活性,检测速度较快,完全满足实时检测的要求。
Research and develop a set of pavement quality inspection system based on machine vision, and apply the improved neural network method to the detection of road pavement defects. The basic composition and overall structure of the system are analyzed. The software design process and the process of network design and training are introduced. Considering the limitation of the traditional image segmentation algorithm, a real-time detection algorithm for detecting arbitrary region in the image is designed. Can be adapted to the pavement crack, transverse crack, longitudinal crack, block crack and other defects and defects coexist complex road sample images. The detection system has a strong flexibility, detection speed, fully meet the requirements of real-time detection.