Research Of Battery Defects Parallel Detecting Methods Based On Machine Vision

Battery is an important consumer goods, as demand growth and the development of automation technology, battery production equipment increasingly toward high-speed, automated direction. Each step of testing is an important part to determine the battery quality. Due to production equipment, technology and environmental factors, the battery in the production process occurs in zinc and coated paper deformation, missing equipment carbon rod, carbon rod eccentricity defects that can seriously affect the quality of the battery. Currently the traditional detection methods rely on manual visual inspection, not only slow detection speed, high labor intensity, but not eliminate subjective errors and testing of visual fatigue caused by difficult to meet the testing requirements of modern high-speed production lines. Therefore, how to efficiently and accurately detect defects in batteries, it’s not only for enterprises to improve the quality and efficiency of the urgent need to solve the problem, but also has important research value.Machine vision inspection method based on Lab VIEW has non-contact, high speed and high precision characteristics. In this thesis, Against to three types of battery production line defect:zinc cylinder, coated paper and carbon rod defect, machine vision of multi-channel parallel detection method research is applied, and parallel detection experimental device structures are initially realized.Firstly, requirement analysis of detecting coated and carbon rod assembly station is performed in thesis, setting up requirements and the overall design of the experimental apparatus. Build a complete hardware and portions of the image acquisition system, including lighting systems design, CMOS sensor selection, selection and installation of optical lenses, PC control computer component selection.Secondly, based on Lab VIEW queue state machine producer/consumer design pattern, multi-channel image acquisition is realized through producer/consumer queue data transmission. Image captured parallel by four CMOS camera and programs operating status are bundled in producer/consumer queue, the queue is used for transmission of image data and operating status. This mode takes full advantage of multi-threaded programming automatic computer control hardware resources, but also a combination of producer/ consumer design pattern of parallelism to improve efficiency program. Software architecture detection device is modular in design, multi-channel image acquisition module, image processing module and data storage module on Lab VIEW implementation process, and the main interface experimental device designed are introduced.Finally, the entire program’s core--image processing module’s design and optimization. Various studies batteries image pre-processing and segmentation by comparing the experimental results show that the median filter is more suitable for battery Image Noise & linear gradation conversion on an enhanced power dry objective information better contrast. Research on image edge detection, the experimental results show that zinc can,coated paper and carbon rod are without good edge information which is benefit to feature analysis. For battery zinc can defects in coated paper and carbon rod defect recognition algorithms. Gradient-based contour extraction can be efficiently extracted zinc can outside contour and achieve the positioning of the carbon rod, the actual pitch contour and contour fitting in zinc measure the degree of deformation is easy to implement and highly reliable; coated paper has obvious regional characteristics, through regional edge intensity detection can accurately identify defects; through regional carbide contour edge intensity detection and targeting can effectively realize the lack of carbon rod and an eccentric defect recognition.