Research On Surface Defect Detection Method Of Jade Beads Based On Edge Calculation

As jade products,the surface of jade beads has cracks,black spots and wear defects due to various factors.At present,the detection of surface defects of jade beads in the industry is still done by artificial subjective judgment.The high cost and low efficiency caused by this seriously affect the overall benefit.Therefore,the research and development of intelligent edge computing platform for surface defects of jade beads has become an urgent need.This paper focuses on the industrial demand of intelligent detection of jade beads surface defects,studies the defect detection algorithm based on deep learning,and builds an intelligent edge computing platform,which provides favorable support for the realization of intelligent detection of jade beads surface defects.The main research contents of this paper include:First of all,this paper constructs the jade bead image data set.Jade beads image data set is the premise and basis of intelligent detection of jade beads surface defects,and the deep learning task cannot be carried out without the support of data.However,since there is no open data set of jade beads images at present,this paper constructs data set based on solid jade beads.Under the guidance of industry professionals,600 images of four categories,including black spot defects,crack defects,wear defects and standard no defects,were collected and marked.The establishment of jade bead image data set provides data support for the subsequent research of jade bead surface defect detection algorithm.Secondly,this paper proposes the CAGN(Coordinated Attention and Ghost Net)neural network algorithm based on the lightweight neural network Ghost Net.CAGN neural network was optimized for Ghost Net as a whole,and by adding Coordinate attention(CA),the model not only considered the information between channels,but also paid attention to the correlation of feature position information.By replacing Leaky Re LU activation function with Leaky Re LU activation function,the problem of neuron failure caused by Relu activation function in the original network is effectively avoided.Through the above methods,a neural network suitable for the detection of jade bead surface defect features is established in this chapter.Finally,the ablation experiment and the comparison experiment before and after the improved algorithm were used to verify the experimental results.The experimental results show that the CAGN network structure can effectively improve the detection accuracy of the model compared with the original Ghost Net network structure in the detection of jade bead surface defects.Thirdly,this paper constructs the CAGN-YOLOv7 algorithm based on the target detection model YOLOv7.By replacing the backbone network of YOLOv7 with CAGN neural network,on the one hand,the problems of complex structure and heavy computing load when the YOLOv7 algorithm is applied to the deployment of edge computing platform in this project are solved.On the other hand,by combining the SPPCSPC of YOLOv7 with the improved Rep Conv structure,It improves the detail attention of multi-size jade bead defect feature map and realizes the effective enhancement of reasoning speed.After the construction of the overall algorithm model,CAGN-Yolov7 combines the advantages of CAGN network structure and YOLOv7 algorithm to simplify the structure of the original YOLOv7 and provide algorithm support for the construction of the subsequent intelligent edge computing platform.Finally,the feasibility of CAGN-YOLOv7 algorithm in the surface defect detection of jade beads is verified by the experimental comparison before and after the improvement,and the model is evaluated in terms of precision and complexity.Finally,this paper built an intelligent edge computing platform for jade bead surface defects.Through the overall design of the intelligent edge computing platform,the effective combination of CAGN-YOLOv7 algorithm and the domestic intelligent edge computing platform Ruixin Micro RV1109 is realized.Through the testing and analysis of the algorithm deployment of each platform,the feasibility verification analysis of the whole platform is finally realized.The intelligent edge computing platform of jade bead surface defect detection has been developed,which has the advantages of real-time and cost,and meets the needs of industry.