Visual Detection Technologies Of Dimension Of Circular Saw Blade Based On Convolutional Neural Network

With the rise of a new round of industrial revolution,visual detection technology has become one of the main bottlenecks in the development of industrial detection.As a typical cutting tool,the diameter of circular saw blade is still at the manual level.However,the traditional visual detection technology is easily affected by the surface texture and the external environment of the workpiece,so the detection accuracy and efficiency of large-dimensions workpiece cannot meet the production requirements.With the rapid development of convolutional neural network,it has a strong generalization ability in image feature extraction.Therefore,this paper research on the visual detection method based on the convolutional neural network,and works on the detection system design,edge feature extraction and refinement processing,sub-pixel edge detection and experimental analysis,so as to realize the high-precision measurement of dimensions of the circular saw blade.According to the measurement accuracy of the circular saw blade,the design requirement and the structure of the inspection system are determined.The working principle of camera and other important parts are expounded and the parameters are calculated and selected.The reflection characteristics of matrix surface are analyzed and the forward lighting scheme is proposed.Based on the module division,the software function module is designed,and the overall design of the detection system is completed,which lays a foundation for the subsequent experimental research.The traditional edge detection method is analyzed,and the experiment shows that the robustness of the traditional operator to the matrix edge detection is poor.The edge detection model based on RCF is established,and the residual structure is added to improve the model,so as to realize the fusion of multi-scale edge feature graphs.Image acquisition and creation of the corresponding edge label map,model training to the circular saw blade matrix edge feature extraction,on this basis through non-maximum suppression,double threshold segmentation and morphological refinement,to obtain a complete and fine edge image,to further improve the edge point positioning accuracy provides a basis.The principle of Zernike moment sub-pixel edge detection is studied,and the mathematical model of Zernike moment 7x7 template is established.By convolving each pixel in the image with the template,the positioning accuracy of edge point pixel is improved to sub-pixel level.The method of camera calibration is studied and the relevant parameters such as camera radial distortion are calculated.The algorithm proposed in this paper is verified through experiments,and the results show that:the outer diameter error is about 0.08mm,which meets the detection requirements of the circular saw blade matrix,and the effectiveness and robustness of this algorithm are verified.The research of this topic has solved the visual measurement of the outer diameter of the circular saw blade and realized the high-efficient and high-precision detection target of large-dimensions workpiece,which is of great significance and value for improving the degree of intelligence of production and the ability of product quality control.