基于SVM的井下水仓淤泥识别系统的研究

Research on sludge recognition system of underground water silo based on SVM

针对煤矿井下水仓经常出现的淤泥中含水量过多而导致的涌仓问题,该文设计了一种基于SVM的井下水仓淤泥表面含水量识别系统。以井下水仓的淤泥与水面对光线反射的不同情况作为识别特征,通过对数及对比度拉伸变换法提高图片局部亮度;采用自适应阈值分割法将灰度图进行二值化处理。在SVM模型中进行前景、背景分割时,将二值化图片作为输入而不是像传统SVM分类直接输入原图,有利于提高识别精度。分别计算将彩色图、二值图作为输入图像的相似度、正确率、错分率和漏分率,得出结果:采用二值化图片作为输入图片的各项分割属性均优于彩色图作为输入图像,为井下水仓涌仓预警系统的搭建提供了基础。

Aiming at the problem of flooding caused by excessive water content in the silt that often occurs in underground sumps in coal mines,a SVM-based system for identifying the surface water content of silt in underground sumps is designed.Taking the silt in the underground sump and the reflection of light on the water surface as the identification feature,the local brightness of the image is improved by the logarithmic and contrast stretching transformation method;the adaptive threshold segmentation method is used to binarize the gray image.When segmenting the foreground and background in the SVM model,the binary image is used as input instead of directly inputting the original image like the traditional SVM classification,which is beneficial to improve the recognition accuracy.Calculate the similarity,correct rate,error rate and omission rate of the color image and the binary image as the input image.The result is that the segmentation attributes of the binary image as the input image are better than the color image As the input image,it provides the basis for the construction of the early warning system of underground water storage inrush.