Longitudinal Tearing Detection Algorithm Of Conveyor Belt Based On Visual Saliency

Conveyor belt is an important equipment for mining and shoulders the responsibility of safe production.In the production process,due to the harsh environment under the mine and the sharpness of the conveying materials,the conveyor belt is often accompanied by longitudinal scratches and even tears,which poses a huge threat to the safety of people and property in the mine..Therefore,it is a research hotspot and topic that has practical value to study how to detect the longitudinal tear on the conveyor belt accurately and quickly.At the beginning,tearing detection often relies on additional equipment such as pressure sensors,which is costly,difficult to set up,and prone to false positives.With the gradual development of machine vision,researchers began to use image-taking devices,such as CCD cameras,to replace the human eye for the measurement and evaluation of belt tears.Combined with image processing technology,non-contact tearing detection came into being.This type of detection pays more attention to the characteristics of the image itself and has higher realtime performance.However,the current detection method can only detect one type of damage for each image,which is not conducive to the detection and maintenance of the conveyor belt.Damage is a small part of the entire conveyor image,which is very different from the conveyor belt background in terms of grayscale.Therefore,the paper introduces visual saliency for belt tearing detectionVisual saliency is based on the characteristics of the human eye,focusing more on a small area or some targets that contain the main information of the image,that is,a salient area.Visual saliency detection is generally based on the underlying visual characteristics of the image,and mostly uses global or local features of the image.However,the current mainstream detection methods often have problems in which the edge of the region is blurred,and the salient target cannot be highlighted overall.Aiming at the above problems,this paper proposes a visual saliency detection method based on multi-feature,called RCS(Rectangle & Color & Spatial).Which extracts multi-layer visual features,and extracts a complete salience targets accurately.Then,the RCS algorithm is applied to tearing detection,realized location and classification of multiple types of damage in the same image in the end.The work of this article is as following :(1)For the case of salient target edge information loss,this paper first uses the bilateral denoising filtering algorithm to preserve the edge and denoise.Then based on the rectangular particles,combines the local and global features of the image effectively,and further preserves the edge information of the image.Through experimental comparison,the bilateral filtering has the highest PSNR(Peak Signal to Noise Ratio)value and significant edge-preserving effect.Calculated by rectangular particles,the background suppression is obvious,and the edge information is preserved completely.(2)For the problem that the target content is incomplete or unevenly highlighted,this paper proposes a spatial feature extraction method based on boundary point knowledge,and then uses SVM classifier to realize foregroundbackground classification based on pixel points.The experimental results show that the background suppression effect of the spatial feature saliency map is obvious,and the significant target is accurately extracted.The salient map generated by the SVM binary classification fully realizes binarization and highlights the significant target.(3)Aiming at the disadvantages of single-type damage detection for single-conveyor images,this paper proposes a method based on SVM multiclassification.After simplifying the proposed RCS algorithm,pixel point classification is performed to realize classification and localization of multiple damages of a single image.The experimental results show that the tearing detection algorithm proposed in this paper can effectively classify the damage type of the conveyor belt,locate the damage location,and improve the detection accuracy.