CMOS Image Matching Sensor And Applied Research

Image matching measurement belongs to the category of artificial intelligence, and it iswidely used in modern automated production industry. Combined with the requirement ofquick sorting for particular shape of parts in production line, the key technologies of CMOSimage measurement sorting system is researched in this paper. The main work of this papercan be divided into the following four parts.Firstly, analyzes the background of the subject, and discussing the common algorithm ofimage matching. Manual measurement has many shortcomings, such as its measurement isslow, and the gauge needs to contact with the object and so on. In response to theseshortcomings, the CMOS image matching sensor is proposed for the non-contactmeasurement of cycle devices in this dissertation. Especially for concentricity measurementthat manual is not easy to achieve, this system has a unique advantage.Secondly, according to the analysis of the functional requirements for this system, andthe convenient of debug, determining the overall program and measurement principles for thissystem, and analysis the key technologies of matching measurement. System choosesOV7670, FPGA, TFT and touch screen to make up the hardware, and program to drive them.Because the shape of testing object is circular, choosing three points to determine a cycle andcounting pixel to calculate diameter as principles. Key technologies related to this systemconclude the selection of image processing algorithms, and the implement of evolutionfunction.Thirdly, to realizing algorithm of digital image processing and matching the cycle inpicture, using program to achieve those function. According to the characteristic of parallelprocess data in FPGA, using algorithm which has the features of parallel and easy toimplement to achieve the function of digital image processing, including median filteringalgorithm, Sobel algorithm and Zhang-Suen fast parallel thinning algorithm. According thecharacteristics of round, system detect points on cycle and process points to calculate theparameters of cycle by using state machine. If the automatic detection fails, there is manualcan complete the measurement.Fourth, calibrate the system, measure the actual objects and error analysis. Throughcompare the actual size of object and the result which system gets, it could obtain the errorwhen measure, and then making a detail analysis for the error.If the system is used in automatic detection with telecentric lens, for example, steppedshaft concentricity measurements, diameter filters, etc., not only can significantly improve the detection efficiency, but also reduce labor costs and improve labor productively; it also canprovide an effective technical means for online testing of automation for enterprise.