基于LeNet-5卷积神经网络的物探野外手写数字识别方法与实现

Method and implementation of handwritten digit recognition in geophysical prospecting field based on LeNet-5convolutional neural network

物探队野外采集作业时,往往需要野外操作手拍摄作业过程,其中包括当前作业位置的桩号信息,以此来协助室内质检人员对采集的视频数据进行质检工作。传统质检工作常常是由质检人员观看野外作业视频,通过查看视频中所展示的桩号信息是否正确,从而完成该项质检内容,上述质检流程浪费了大量的人力、物力,质检效率较低。本文详细介绍了LeNet-5卷积神经网络模型,并在此基础上利用Matlab软件,实现了对LeNet-5模型的训练和测试。通过训练效果图和测试结果可以看出,LeNet-5卷积神经网络模型可以快速、准确的识别物探工区手写数字,具有较好的准确性、时效性和可重复性。

In the field acquisition operation of the geophysical prospecting team,the field operator is often required to shoot the operation process,including the stake information of the current operation position,so as to assist the indoor quality inspection personnel to carry out the quality inspection of the collected video data.The traditional quality inspection work is often carried out by the quality inspectors to watch the field operation video,and to complete the quality inspection content by checking whether the stake information displayed in the video is correct.However,the above quality inspection process wastes a lot of manpower and material resources,and the quality inspection efficiency is low.This paper introduces the LeNet-5convolutional neural network model in detail,and realizes the training and testing of LeNet-5model by using Matlab software.Through the training effect diagram and test results,it can be seen that the LeNet-5convolution neural network model can quickly and accurately identify the handwritten numerals in the geophysical prospecting work area,with good accuracy,timeliness and repeatability.