Investigation On Interest Detection And Silhouette Extraction Algorithm Of Imaging Ladar

Laser radar is a technology developed from 70s of 20 century.As the laser radar technology developed, the image processing of ladar(laser radar) becomes more and more important. The system of ladar was divided in 5 parts: preprocess, detection, extraction, decomposition, matching. In this paper, the first three parts are the focal point, especially the detection and extraction parts. The five parts of the laser radar is associated with each other, in order to search the extraction and detection, the preprocess should be searched before.The kinds and causes of the noise are introduced firstly in this paper for better noise suppression, it is concluded that the speckle noise is the main component of intensity image noise, and the outlier is the main component of the range image noise. The distribution of probability of speckle noise is analyzed, the relation of the probability of range anomaly and the threshold was introduced, the increasing of threshold will decrease the probability of range anomaly, but it will cause the loss of the valued information. It is approved the outlier should be cleared by the statistic method.In chapter 3, the relation between RMSE and threshold was reseached.The existent outlier filtering methods was researched, then compared the methods filtering affect and the running time with each others, then analyzed in the frequency field.Based on the studies of existent methods, a new filtering methods was brought out, it was named adjusted histogram method, the combination of filter methods was also studied. At the end of the chapter, the new methods was tested by the real range image data, the size and shape of template was also researched. The result of new methods was compared with the best existent method—order mean-weight method, the compared result showed that the new methods had the better affect both in filtering and the running time.In last chapter, the studies about detection and silhouette extraction was made, in order to research, the 2 scenes were constructed, named tank simulated environment and the carrier simulated environment. By detection, the CI image was obtained, and the WRS was used in the next subsystem. In the step of silhouette extraction, a new estimation standard was brought out, named ESE system. The ESE was used 3 kinds of data to estimate the result of silhouette extraction. The relation between CNR, distance and the result of extraction was researched, the relation showed that the net hit ratio could represent the result of extraction. At the end of the chapter, the algorithm was analyzed, the algorithm has a broad application field, good perfoemance in aging test, the time complexity was focal point. It is the base of algorithm selection.