The Research And Development Of Sorting Inspection System Based On Image Matching

With the rapid development in the domestic logistics industry, the sorting efficiency is higherand higher requirements. The sorting inspection system based on image matching technology canautomatically identify the objects on sorting line. If the system is used in logistics industry, it hasthe many advantages, for example, ensuring the accuracy of the sorting, improving the sortingefficiency, reducing labor costs and so on. This paper design the inspection system based on objectrecognition and use the software platform of Visual Studio2005and computer vision libraryOpenCV to complete the interface design and function of each module of the system.Firstly the system composition, working principle and the design scheme of system areexpounded in detail. The system hardware mainly includes the light source, camera, lens, and IPC.Software programming platform and image processing software are also carried on the simpleintroduction.Then, the method of foreground extraction based on Gaussian mixture model is used to solvethe problem of goods image capture in testing area. We can determine whether the goods have goneinto the testing area by the characteristics of foreground. Thus images can be captured.Besides, the core algorithms of the system, including template matching algorithm based onfeature and SIFT feature matching algorithm are studied in this paper. Considering identificationspeed and precision requirements of the system, these two kinds of algorithms are used to realizetesting process. When the position of goods would have offset and rotation relative with thestandard one, the algorithm of affine transformation is applied to rectify image position to makethe images alignmen.And an optimization algorithm is proposed based on the research of SIFTfeature matching algorithm. The experimental results show that it can effectively improve theaccuracy of the matching, reduce workload, and improve detection efficiency of the system.Finally the system is divided into four modules: image acquisition module, detection module,template module and database module. Paper expounds function and design of each module andshowed interface design of the system. Multithreading technology is applied to inspection system isproposed. The experiment shows that the multithreading can obviously increase the detectionefficiency. The application of system in tobacco distribution center shows that it can obtain a moreaccurate test results, and has a certain application value.