The Research Of The Battery Bottom Defect Decection Aigorithm

In recent years, the speed of alkaline battery production line becomes much faster. It’s product quality monitoring also becomes more and more stringent. Therefore, higher requirement for the product quality detection system must be put forward. However, the traditional detection methods are almost based on manual testing. The human eyes have so many disadvantages, such as slow speed and being easy to be fatigue, which make it hardly adapt to the requirements of modern high speed batter production line. As a new type of industrial automatic detection technology, machine vision combined with the calculation ability of computers can greatly improve the efficiency of battery defect detection and reduce labor costs. Based on these requirements, this thesis has studied machine vision defect detection algorithm and a detection system to defect on the bottom surface of alkaline batteries is designed which realized the real-time detection of battery defects. The mainly contents are as follows:The noise characteristics of defects imaging on the metal bottom surface of battery is analyzed. A mathematic model of scratch defects for further analysis is also built. Then, we make an image processing algorithm flow of the entire system and analyzed all the factors that may affect the quality of images.A battery bottom image preprocessing algorithm which includes image filtering, adaptive image binarization method, mathematical morphology processing algorithm and so on are studied. The effects of defect detection on the battery bottom surface are enhanced obviously through the preprocessing of the original images which are obtained from machine vision sample system.The segmentation and rotation algorithm of the bottom images of alkaline battery are researched. According to the specific situation of the battery defect detection, the Hough circle segmentation method is analyzed and a fast segmentation method based on image pixel traversal is proposed. A fast rotation method of the battery bottom image through analyzing the rotation angle based on the image similarity is proposed., The rotation angle is obtained and good effects is achieved through using two local characteristics of battery image. Two characteristic parameters that are overall characteristics and local ones of the bottom images of alkaline battery selection and extraction methods are studied. The image correlation is used as the overall characteristics and the geometrical characteristics of the connected regions are selected as the local ones. The principle and design method of classifier is studied and a two-stage classifier for battery defects detections is designed.The battery defect detection algorithm studied in this thesis can be applied to all the alkaline batteries, including AA, AAA, D batteries and so on. It also solved the problems of low speed and easy to be fatigue of human eye detection and the limitation of disability of real-time detection. This method also provides a reference for other metal surface detection. It has good universality and practicality according to lots of experiments.