Technique And Application Of Smart Camera Based On C64x DSP

With the development of pipeline, the production efficiency is increased tremendously. As those productions go out in a high-speed, detection system’s speed must be increased correspondingly. That poses enormous challenge to the traditional detection way. It’s a new detection method that puts smart camera in pipeline. As it developed rapidly those years for vision system, smart camera can meet the needs of pipeline on its detection accuracy and speed. Based on the research foundation of smart camera with 5509 DSP and VGA system in the laboratory, further research has been done on C64x DSP core with higher speed. In order to solve the problem of big-scale noise, target dim and slowly changed parameters, better image processing algorithm is worked out with robustness and adaptability. This can make the system smarter. After the camera is finished, calibration is done to acquire the parameters of detection system. With the calibrated system, a new scheme is designed to measure the diameters of bearings.The main work is shown as follows:1.A production pipeline detecting model is built by researching on bearings detection. Based on it, a dynamic positioning method is analyzed for Machine vision in the production line. The recursive formula of finding target on pipeline is also worked out. It is proved that the formula can effectively improve the target positioning speed on the pipeline.2.A smart camera prototype is made. In the system, MT9M001 is used as vision sensor which makes system resolution increased by 4 times to the VGA system. The FPGA subsystem is extended by a SDRAM which can store the image temporarily for processing. The special multimedia high speed processing chip DM642 is used whose peak processing speed is 12 times higher than the original system with 5509DSP. The ping-pang cache structure and EDMA is used for sending image data between FPGA and DSP. Making the testing speed from 12MByte / s in the past to 30MByte / s now, the method is proved to be a high speed data transmission scheme.3.The intelligent design which makes system smarter is to work out robust and adaptive algorithms. Frist, the improved seed-growth marker method is used before ticking out big-scale noises. Besides, the target matching method with weight is used to identify unclear goals. Futher more, the neural network is used in vision system to make it have self learning capbility. And experiments are made to prove them.4.The method of measuring diameter of bearing by using its end face image is independently designed. The way put forward is that, Sampling edge points first, doing the Circular curve fitting then, the Polar coordinate transformation and diameter number getting comes last. This method is proved to have better precision when the noise is strong.