Research On Surface Defect Detection Technology Of Non-ferrous Metal Nickel Sheet Based On Machine Vision

As an important basic raw material,nickel is widely used in many industrial fields.In the process of preparing nickel plates,it is affected by many factors,and surface defects are inevitable.The impact of these surface defects on the automatic packaging process and sales of the plate cannot be underestimated.At this stage,most non-ferrous metallurgical enterprises use manual visual inspection to detect surface defects,which requires a lot of labor costs and is highly subjective,and factors such as human eye fatigue and distraction can lead to a decrease in inspection efficiency and accuracy.In view of the disadvantages of manual visual inspection,it is of great significance and value to assist inspectors to detect defects based on machine vision-based sheet surface defect detection technology solutions.Based on this,this paper takes the surface defects of nickel plates as the research object,and based on machine vision technology,develops a system for assisting inspectors to detect surface defects of plates.The main research content of this paper is as follows:(1)By analyzing the formation mechanism and characteristics of each defect on the surface of nickel plate,the overall system design is carried out according to the actual requirements.According to the surface defect characteristics of the plate,the field of view of the plate,and the inspection requirements,the key hardware selection of the camera,lens and light source of the machine vision system is completed.Combined with software system requirements analysis,the overall structure design and testing process design of the software system are realized.(2)Aiming at the bulge defects such as granulation and bubble,an improved Canny edge detection algorithm is proposed based on image processing technology.Through the experimental comparison of this algorithm with other classical algorithms and some existing literature algorithms,it is proved that the proposed algorithm is superior and effective in detecting plate bulge defects.(3)Aiming at the surface ablation defects of nickel plate,based on the semantic segmentation technology of deep learning,an improved Unet network is designed,and a semantic segmentation network model is proposed to segment the ablation surface defects.The model ablation experiment shows that the fusion of ECANet and ASPP module has higher detection accuracy.By comparing the algorithm proposed in this paper with other classical semantic segmentation network algorithms,the pixel accuracy and average intersection ratio of ablation surface defects are higher,which is5.07% and 3.54% higher than that before improvement.(4)The development of the non-ferrous metal surface defect detection system based on machine vision is completed,the calibration and parameter correction of the hardware system industrial camera are realized,and the man-machine interface of the detection software system is designed.and the related functions are tested and verified.