Object Detection using Gabor Responses and Texture Information in Low Contrast Infrared Images

S.-G. Sun, S.-S. Park, J.-M. Lee, and Y.-S. Kang (Korea)


Automatic target detection, clutter rejection, co occurrence matrix, forward-looking infrared, Gabor response


A new object detection method in military army application is proposed. The main purpose of this study is to identify target locations with low false alarms in thermal infrared images of natural field. The proposed method is different from the previous research because it uses morphology filters on Gabor response images instead of an intensity image in initial detection stage. And the method does not need to extract precise target silhouette to distinguish true targets or clutters. It comprises three distinct stages. First, some morphological operations and adaptive thresholding are applied to the summation of four Gabor responses of an image frame to find out salient regions. The locations of extracted regions can be classified into targets or clutters. Then, local texture features are computed from salient regions of an input image. Finally, the local texture features from an input image are compared with the training data to distinguish between true targets and clutters. The MLP (Multi-Layer Perceptron) with three layers is chosen as a classifier. Experiment using natural infrared images shows that the proposed method is superior to previously reported approaches.

Important Links:

Go Back