Detection of Small Target image based on Spatio-Temporal Entropy Operator

The recent development of the infrared cameras, the technologies of detection and tracking the target are under progress in the surveillance field, intelligent robot field, and the military field. Detection of infrared threat in IR defense systems are very important to use IR cameras. Many scholars have paid much attention to the study on spatial and temporal detections. Existing methods are focusing to detect the small targets in the infrared images and in case of very small targets, they are focusing on the algorithm to distinguish the target and the noise. The method using the mean-subtraction filter obtains the temperature difference with the target by measuring the mean temperature of the surrounding environment, Although it can increase the threshold value of the target temperature, the constant false alarm rate detector represents high and it is not suitable for the target detection in the low contrast infrared image and for the target expressed with dot in the long distance. Generally, the method using morphological filter detects the target by reducing the 3D temporal detection technique to 2D space detection level after performing the preprocessing of the raw image using top-hat operator. At this time, the target was detected with the constant false alarm rate detector. This method is useful for the small target having diverse speeds but represents high false alarm rate detector due to surrounding environments [1, 2]. Laplacian of gaussian (LOG) filter is useful to detect the blob shape. Since the small target represented in blob shape in the long distance, the small target is detected by applying this method. Therefore there is difficulty in detecting the small target directly. The state-of-the-art spatial target detection methods are the local entropy operator has been applied to suppress backgrounds of small target images [3]. Deng et al [4] purposed the above-mentioned great computation and target growing effect, a different background suppression method based on the fast local reverse entropy operator (FLREO). The concept of FLREO derives from the idea of local entropy operator, reverse entropy and local reverse entropy operator (LREO). Then a local reverse entropy map (LREM) is built on FLREO, and it is adopted to suppress complex backgrounds of small target images. In this paper, an efficient method of detecting small targets in IR images under various environments is proposed, by using spatiotemporal information entropy. The rest of this paper is as follows. First, the proposed method is explained in Section 2. The experiment results evaluate the conventional method and the proposed method in Section 3, and finally this paper is concluded in Section 4.