Sand Residue Detection Of Mobile Phone Partition Based On Machine Vision

Now mobile phone is the most influential consumer electronic products. And itrequires extremely high quality appearance in the process of mass production. Thetraditional inspection of mobile phone accessories is usually done by manualeyeballing. Manual inspection is convenient, practical and strong adaptability, butalso has high labor intensity, low efficiency, and easy to produce visual fatigue.The partitions of mobile phone often present scratches, sand residues, stains andother quality problem during the production, which will affect the assembly ofmobile phone and the final product quality. The partition is an important componentof mobile phone assembly, the detection of it is important to control the quality ofmobile phone. Therefore, based on image processing and pattern recognition, thedetection of residual sand in the mobile phone partition is studied in this work, themain work is as follows:(1) Firstly, we have maken an analysis of the current mobile phones andaccessories consumption market in China. And a brief introduction of machinevision technology and mobile phone accessories quality testing status, explains thenecessity of testing residual sand of mobile phone separator.(2) Considering the dot object detection problem, we use top-hat transform,template matching, double loop filter to inspect sand residue, and the advantages anddisadvantages of these detection algorithms are analysized.(3) Based on line scan and local peak analysis, a novel method is proposed.Sand residues appear as round local peaks, and they reflect omnidirectionalsymmetry and Gaussian distribution characteristics. First, the paper uses a rotatingline scan operator to scan image. Second, by using binarization technology, therelationship of adjacent pixels are described. Third, combined with local peak andthe pixel gray level information, sand residue are detected. Finally, the sand pointsare segmented with double threshold technology. Experiments show that theproposed algorithm is superior to traditional top-hat transformation, Gauss templatematching and double ring filtering method. The method is robust to illuminationvarietion, and has lower false negative and false positive results. It presents goodadaptability and practicability.(4) Based on machine vision, the visual system for the detection of mobilephone residue is designed. It mainly includes three modules,i.e., the user management module, parameter setting module and image processing module. Thesystem can accurately detect the sand, the whole process is completely automatic,and avoid the interference of human, thus greatly improves the precision andefficiency of production.