| With the rapid development of science and technology, infrared small targetdetection technology has been widely used in military, medical science, aerospace andother fields, such as precision-guider system, early warning system, Unmanned AerialVehicles. It has very important theoretical significance and practical application value toresearch the infrared small target detection.Infrared small targets have features of weak energy, small size and low SNR. Theyare easy to be submerged under complex background. So it is a challenging subject todetect infrared small targets. The main contents and work in this paper could besummarized as follows:(1) With the help of mask image and morphological method, we extract thebackground of infrared image, which will be used in Multi-scale Local Contrast Measure(MLCM). In the image with the background of sea, Hough transformation is used toextract the sea-horizon, the interested area of targets.(2) Based on local feature, MLCM can brighten the target and impress thebackground. Then, a proper threshold value will be chosen to get the target. In multipletarget detection, through analyzing a dozen of threshold values and its correspondingresults, a value will be chosen as the best threshold value.(3) In the practical application of infrared small target detection, it requires that thealgorithm must have high real-time capability. Therefore, it is programmed underGraphic Processing Unit (GPU). In image processing each pixel has the same dealing,which shows the feature of Single Instruction Multiple Data. Data processing with thiskind of feature is suitable to be programmed on GPU. The experimental results show thatthe speed of the algorithm is much faster by GPU.(4) At last, the experiments are designed to check the accuracy and the speed of the algorithm. Through the experiments, we can see that the algorithm not only has highrecognition rate but also shows a good performance in real-time on GPU platform. |
PDF Full Text Request