Machine Vision Based Abnormal Detection Of Bolt Support In Coal Mine Roadway

Nowadays,most of the coal mines in China are underground coal mines,in which roadway is the only channel for coal mining and transportation.Therefore,the safety of roadways must be ensured to guarantee the safety of coal mine production.The safety of roadway is maintained by supporting systems,among which the mostly used one in coal mines is bolt supporting system.With the passage of time and the progress of production,bolt support failure may occur,causing the decline of the supporting effect of bolt supporting system.When the support capacity of the supporting system drops to the point where it cannot support the stability of the surrounding rock of the roadway,accidents such as roof fall may occur,damaging the personal safety of miners and the production safety of coal mines.To maintain the safety of roadways,it is of great significance to discover the abnormal bolt in time.The existing bolt abnormality detection is a contact detection method,which requires operators to carry out pull-out or acoustic wave experiments on each bolt using handheld detection equipment.However,the length of roadways supported by bolts is usually measured by kilometers and the density of bolts is high.Therefore,every bolt detection task needs a lot of manpower.In this context,it is urgent to build up a non-contact and rapid detection method for bolt abnormality in unmanned intelligent mines.Based on the platform of roadway inspection robot and the machine vision technology,this paper proposes a bolt anomaly detection method composed of bolt scene matching and bolt region extraction module and bolt feature extraction and anomaly judgement module.According to the morphological characteristics of the exposed part of the bolt,it judges whether the bolt is loose or not by measuring the change information of the bolt length and angle in the field of vision,so as to realize the non-contact rapid detection of the bolt.The main work of this paper is as follows:(1)The bolt scene matching and bolt region extraction method is proposed.Firstly,according to the significant difference between the bolt loose anomaly and the normal in image representation,the pre-acquisition map of the same scene is established by sensing hash matching and acquisition map,and the anomaly is judged by comparing the change of the bare section of the bolt in the two maps.Secondly,aiming at solving the problem of uneven illumination in roadway,an image enhancement method of bolt scene based on histogram equalization is proposed.In addition,in view of the complex roadway environment,there are many pipelines and other equipment similar to the morphological characteristics of the bolt rod,which is difficult to extract the characteristics of the bolt rod directly from the global vision.The YOLOv3 algorithm optimized by k-means anchor frame is used to locate the bolt area,which lays the foundation for the subsequent feature extraction and loose anomaly detection.(2)The bolt feature extraction and anomaly judgement method is proposed.Firstly,the improved Canny edge extraction method is adopted,and the bilateral filtering is used to better protect the edge information of the bolt area map.The image entropy is introduced to adaptively determine the optimal Canny parameters.Secondly,a bolt feature extraction method based on LSD line detection and parallel line group screening is proposed.LSD algorithm is used to detect the straight-line segment in the bolt area,and the bolt line screening mechanism based on parallel features is established to obtain the characteristics of the exposed segment of the bolt.Finally,the characteristics of the exposed section of the bolt detected in the acquisition map and the pre-acquisition map are compared,and the bolt anomaly detection based on the length and angle change information is realized.The experimental results show that the method proposed in this paper is stable and reliable.By detecting the change of the exposed section of the bolt,the bolt loosening anomaly detection in the scene can be carried out without contact,unmanned and long time,which has good practical significance for maintaining the safety of the roadway.