Research Of Moving Targets Detection Technology In Infrared Image

The moving target detection and recognition in infrared images is one of the keytechnologies of modern weapons and equipment, it also has been a hot issue in thefield of image processing in recent years. As the world’s local war weapon againstescalating intelligence on weapon systems have become increasingly demanding,which hopes defense and attack weapon system capable of long-range conditions in theinfrared target detection and recognition. Since the remote infrared imaging andtransmission exist different factors such as illumination changes, background motion,target shadows, target occlusion and overlap, camera jitter and so on, which bringsome difficulties to the moving target identification.The main contents of this paper include infrared image enhancement research andmoving target detection research.As to the poor contrast and low signal-to-noise ratio of infrared weak target, anew algorithm is proposed based on a mathematical morphology which combines thelocal linear transformation and median filtering. Firstly, using a local lineartransformation for the original image to achieve the purpose of enhancing the contrastof target and background; secondly, using median filter for image smoothing; finally,introducing the idea of Top-Hat filter to filter out the background and target noise. Thealgorithm not only improves the image and contrast, but also maintains the target edgeinformation, which is conducive to the following target detection.In view of the problem that the global optical flow algorithm cannot acquireaccuracy motion parameter estimation at a low-gradient value, and is shaped with ahigh false alarm rate, an improved global optical flow technology has been proposed,which combined with mean shift algorithm and morphological operations to detect themoving targets from image sequences. Firstly, the Gaussian filter is used to predict thenoise of the background; secondly, by studying the basic principles of the globalconstraint algorithm proposed by Horn and Schunck, the weighting function in thebrightness of the conservation constraints has been improved in order to acquire themoving target area; finally, morphological filtering approach is used to simplify theimage data, and the mean shift segmentation method is used to achieve accuratedetection of the moving targets. The improved method not only has a strongself-adaptive ability in target detection, but also reduces the false alarm rate and increases the detection rate.