Dsp-based Digital Image Measure System

Besides the characteristics of dependability, non-contacting and automation, modern industrial production requires product detection to be very fast and precise, so it needs the real-time measurement system. Some times, artificial detection is better than machine detection , but there are some problems with artificial detection ,for example, artificial detection is slow and the operator should be tired. These problems should reduce the precision of detection, so we need a faster auto-detecting method which is also more precise. In this day, image processing technology is gradually perfect and reliable. It is possible to use image processing to solve this problem.Along with the quick development of image processing and digital signal processor, common DSP chips were widely utilized to image processing. Extracting digital image from video signal inputted by CCD camera and processing the data stored in buffer to get the final result are the main works of image processing system. The processed data should been stored in DSP' storage or PC, and the result should be displayed on the screen. In this system, hexadecimal data should be stored in buffer, so it is convenient for programmer to use the data directly. Now we can see that image processing algorithm and software designing are the main work in our research.Firstly, we introduced the significance and development of image measurement system in a whole, then introduced video signal ,the theory and work method of CCD camera, and explained the system's hardware and store structure. Image processing algorithm is the main problem in our research. We must process and analyze the original image before extract dimension character from image to get the dimension data. Edge detection is the precondition and basis of image segmentation and feature extraction. So, firstly we used edge detection to get edge image. Usually, we use template to implement edge detection. The most ordinary templates are prewitt and sobel operators, and so on. We chose sobel operator for that sobel operator is betterat noise reduction than prewitt operator. We must indicate that it's inevitable that there should be noise during the process of image collection for the reasons of environment , quantization, etc. So we need to filter noise before process image. Furthermore, it is necessary to transfer edge image into binary image. It is impossible to get perfect edge whether the image processing algorithm is very perfect, so we need to process edge tracing to compensate edge in subsequent process. If we get integrated edge image, then we can calculate dimensions such as area, perimeter though some rules. Finally, we ran our code on DSP platform, the result indicated that image measurement system is a real-time and precise system which can meet common application's requirements.