| Car-body inline visual inspection system could complete the inline measuringtask with high precision and plays an important role in modern automobile assemblyprocess. In the system, the core component is the visual sensor which defines theefficiency and precision. But traditional visual sensor has poor signal processingcapability, leading to centralized processing in the measuring computer and limitingthe inspection speed and precision. Also, the output signal of traditional sensor is inanalog form with bad immunity to interference. These defects restrain thedevelopment of visual inspection system. To solve the problem, this paper studies thekey techniques of smart digital visual sensor. At the core of a fast template matchingalgorithm, the new sensor integrates the functionalities of data collection, coarsepositioning of characteristics, communication and controlling. And it pushes thevisual inspection system forwords to do processing in distributed and digital formwith higher measuring speed，precision and stability.The main works of this paper are summarized as follows:1. To improve the efficiency and precision of the visual inspection system, a digitaldistributed processing solution was established and the research subject， Studyon Key Techniques of Smart Digital Visual Sensor was made clear.2. The overall project of smart digital visual sensor was proposed.3. With S3C6410as core processor, the peripheral funciton modules were designed.4. With Linux as OS, Bootloader, kernel, file system and drivers were developed.5. On ARM Linux platform, NNPROD was adopted as the similarity measurementwith a search strategy of multi-resolution Pyramid.6. A new2D summation method called Running Sum with Complementing LostPixels was proposed, optimizing the calculation of NNPROD.7. The visual sensor had been proved experimentally to run the image processingfunction with high speed，high precision and high robustness, and to improve theefficiency and precision of the visual inspection system. |
PDF Full Text Request