基于图像识别技术的井下重大事故报警方法研究  

Study on Alarm Method of Underground Serious Accident Based on Image Recognition Technology

作　　者：李晓丽[1] 王庆福[2] LI Xiao-li;WANG Qing-fu(Institute of Disaster Prevention,Sanhe 065201,China;Liaoning Academy of Governance,Shenyang 110161,China)

机构地区：[1]防灾科技学院,河北三河065201 [2]辽宁行政学院,沈阳110161

出　　处：《煤炭技术》2021年第11期174-176,共3页Coal Technology

基　　金：中国地震局建筑物破坏机理与防御重点实验室开放基金资助项目(ZY20210312)。

摘　　要：针对基于图像识别的报警系统研究,采用了DSP技术、图像处理技术、RFID技术等,实时勘测裂缝的动态变化,并迅速定位发出警报。图像处理方面,对图像完成PCA降维后进行OTSU分割,然后通过Wiener算法完成复原,最后经泊松融合后提取HOG特征,完成裂缝特征提取;最后对裂缝特征进行分析,设计了裂缝面积占比、区域裂缝面积占比、裂缝面积变化速率3种预警准则,当3种准则均满足时认定存在危险,实时上报进行报警。Aiming at the study of alarm system based on image recognition,DSP technology,image processing technology and RFID technology are adopted to complete the real-time survey of dynamic changes of cracks,and quickly locate the alarm.In the aspect of image processing,OTSU segmentation was carried out after PCA dimension reduction,and then restoration was completed by Wiener algorithm.Finally,HOG feature was extracted after Poisson fusion,and crack feature extraction was completed.Finally,the characteristics of cracks are analyzed,and three early-warning criteria are designed,namely,the ratio of crack area,the ratio of regional crack area and the rate of change of crack area.When the three criteria are met,the danger is identified and reported in real time for alarm.

关 键 词：图像识别 阈值分割 HOG特征 预警准则 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TD76[自动化与计算机技术—计算机科学与技术]