Research Of Image Input And Lens Distortion Correction Technology On DM6437

Machine vision is a hot spot of image processing technology research. Smart camera is a rug in the field of machine vision, because it combines several high technologies include IC design, embedded hardware design, embedded software design, optical design, mechanical design, machine vision algorithm design, and so on. The smart camera research and design has been last for a long time in foreign countries, and been made some great progress. Some foreign high-tech companies have built up their long product lines of smart camera. However, result from Technical barriers and other factors, smart camera research in China has been gained few achievements, and the products are so rare that the technical support and value-added services do not exist totally.This research is about the image input and processing in smart camera, basing on DM6437 DSP platform which is offered by Texas Instruments Incorporated (TI). Hardware design, software framework design and algorithm design are included. The image input interface, the algorithm optimization are the research emphasis, and the result has been shown.To choose a suitable image sensor and proposal is the first step in the image input interface design. CCD sensor, specialized Timing generator and analog front end were chosen to assemble the image sample circuit. The hardware schematic diagram has been finished, and the driven clock of CCD sensor, the sample control circuit and the DSP interface were realized.In software design, a flexible software framework was designed for the feature of DM6437 and the extension requirement. TI's driver libraries were made good use of managing and operating the low level hardware. Besides the framework, an algorithm, which is compatible of TI's DSP Algorithm Standard, has been design for the lens distortion correction.The optimization technology is an important part of this research. Several type optimization methods were applied to the distortion correction algorithm and they were discussed in this paper. The architecture of DM6437 is so complicated that some optimization methods were limited and the others require hardware support. Compare to the original one, the speed of the algorithm has been improved more than 500 times.A simple testing platform was assembled to test the software and the algorithm. Some phenomenon which occurred during the test has been discussed and some suggestions were given.