A Research Of High-speed Image Capture And Processing System For Surface Defect Detection Of TFT Panel

In the field of AOI detection for TFT-LCD panel surface defects, the act is that we start late, slow in develop and the technology is backward at present. In this situation, this article studies the techniques of surface defect detection, extraction and analysis for TFT panel. Including the research of high-speed image capture system, the construction of the experimental image capture system, the high-speed image processing methods and the implementation on FPGA, and the design of interface software for display and control.The size of TFT-LCD panel at generation 8.5 is 2200mmx2500mm, the quantity of image data will be about 150GB. With so much mage data, a single computer is obviously powerless; This task is to use local area network (LAN) to form a virtual super computer with huge computing and storage resources. Therefore we need to study the parallel processing architecture, including hardware architecture and software parallel algorithms. At this present, this task is to study a distributed parallel high-speed image capture and processing system to reach the target that "High-speed parallel processing and display".Image capture system consists of 4 DALSA P3-80 series of 16k of linear scan camera that vertically in parallel mode, scan the TFT panel on the floating cargo platform below at a high speed, the image through the Camera-link bus transmit to the image processing system.Image processing system is a capture card consists of an integrated with image capture, image analysis and processing functions, the trigger module and the corresponding software based on win7. Image receiving and processing performed by the FPGA, prior written algorithm in image processing, FPGA, such as filtering processing, fast Fourier transform, wavelet transform, image segmentation and edge detection and so on, to preprocess the input image. In the process of processing after the show, the PC interface software will analyze the processing result after image processing, labeling defects and classify the defects.Through the experiments, the image processing algorithm and improved PC interface software for many times, detection algorithm scheme ware finally determined. After repeated experiments, at present, we achieved the requirements of project.