Algorithm of binary image labeling and parameter extracting based on FPGA

For the real-time detection and identification requirements of the rail profile image, A quickly algorithm using connectivity labeling of binary image and parameter extracting to remove the speckle disturbing was presented in this paper, based on FPGA, it can use the limited hardware resources of the system to realize high-speed and accurate binary image labeling and feature extraction, obtain the bright band image and region. This method overcomes the shortcoming of previous methods which must scan pixel repeated and require a large memory to record the relationship of the connected components, and a large computation to merge the labeling. Also, it is rapid, simple, rules, and extensibility, the processing speed will be increased by providing a quick effective way to identify and record the complex relationship between regions, and by completing the label merge or parameter extraction during the line or field blanking. realizing of FPGA can accurately and effectively identify the complex connectivity between images, produce the correct label results and extract the characteristic parameters of each connected component which can limit the subsequent processing in a rectangle and save the time and resource greatly, provide guarantee for subsequent identification. The algorithm is used and we have a good effect on the scene testing of rail measurement, it is enough to meet the requirement of real-time image recognition system.