论文部分内容阅读
人工检测公路边坡灾害,费时费力,效率和识别率较低,随着以深度卷积神经网络为代表的深度学习技术在图像目标识别方面取得的卓越成绩,为实现公路边坡灾害图像自动检测提供了新的思路。有鉴于此,针对传统人工方案存在的问题,本文设计了基于深度卷积神经网络实现公路边坡灾害自动识别系统。系统由预处理、深度学习网络模型及后处理三部分构成,在深度学习开源Caffe开发环境下,实现了Alex Net和Goog Le Net深度学习网络模型并采用大量公路边坡数据完成模型训练。通过对公路边坡实测数据的分类试验,本方案的边坡灾害识别率达到90%左右,表明基于深度学习的公路边坡检测方案可有效完成公路边坡灾害识别任务,有效替代传统人工检测方式。
Artificial detection of roadside slope disasters is time-consuming and labor-intensive, with low efficiency and recognition rate. With the advanced achievements of deep learning technology represented by deep convolutional neural networks in image recognition, automatic detection of roadside slope disaster images Provide a new way of thinking. In view of this, in view of the problems existing in the traditional artificial scheme, this paper designs an automatic identification system of highway slope disaster based on deep convolution neural network. The system consists of preprocessing, deep learning network model and postprocessing. Under the condition of deep learning open source Caffe development environment, the system realizes Alex Net and Goog Le Net deep learning network model and uses a large amount of highway slope data to complete model training. Through the classification test of measured data of highway slope, the recognition rate of slope disaster of the scheme reaches about 90%, which shows that the scheme of highway slope detection based on depth learning can effectively accomplish the task of highway slope hazard identification and effectively replace the traditional manual detection methods .