| The "Made in China 2025" strategy raises the requirements for the independent innovation ability of the national manufacturing industry,and also promotes the development of energy-saving and new energy vehicles.The preparation process and quality control of key battery parts are one of the main problems and constraints faced by the development of the automotive industry.As the key substrate for preparing power battery,Nickel plated punched steel strip has good conductivity and corrosion resistance.The parameters such as punching hole diameter,transverse hole spacing and longitudinal hole spacing are the key points of product quality.At present,the surface parameter detection of steel strip in China still adopts manual method,which is time-consuming,laborious and easy to damage the surface of steel strip.With the continuous development of industrial modernization and the maturity of machine vision technology,in order to improve the problems of false detection and missed detection caused by manual detection and meet the requirements of steel strip detection speed and accuracy in the manufacturing industry,it is urgent to replace the traditional method with intelligent systems and methods.Combined with the actual inspection requirements of the factory,this paper puts forward a parameter measurement method of punched nickel plated steel strip based on machine vision.This paper mainly completes the following work:1.According to the production environment,detection requirements and characteristics of the punched nickel plated steel strip factory,starting with the actual project design and detection requirements,this paper establishes a punched nickel plated steel strip visual detection system composed of light source,lens,industrial camera and motion platform,which can effectively collect and detect the images of punched nickel plated steel strip of various sizes,quickly feed back the punching circle parameters in the image and analyze whether there are defects in time.2.The classification algorithm of punched nickel plated steel strip based on improved lightweight network is studied,the efficient bottleneck layer structure is used,and the position correlation module is proposed to enhance the relationship between steel strip image pixels.At the same time,the improved focalloss is used as the loss function to balance the difficult and easy samples and large samples in the classification process,which improves the average accuracy and detection speed of image classification of punched nickel plated steel strip.An improved incremental method for punching nickel plated steel strip is proposed.While learning the image classification features of the new steel strip,the average accuracy of the original steel strip classification is maintained.3.The surface image preprocessing and edge detection algorithm of punched nickel plated steel strip are studied,and the improved Canny algorithm is used to improve the connectivity and integrity of the edge.In addition,according to the specific inspection requirements,scientific research was carried out on the multi-circle inspection optimization algorithm,and a perfect least squares optimization algorithm was clearly proposed.Compared with the traditional Hough and least square method,the improved optimization algorithm has a faster main performance in terms of linear fitting time and precision.Finally,the optimized calculation method is applied to the calculation of technical parameters such as the center point and diameter of stamping processing on the collected images.4.After the above basic research,the on-site hardware platform and the deployment of parameter measurement algorithm are built.Through long-time engineering verification,it is confirmed that the system can quickly and stably detect parameters,meet the requirements of 0.01 mm detection accuracy,meet the conditions of practical industrial application,effectively replace manual detection,and realize the original intention of this design. |
