IR imagery based mine detection techniques

During the past several years there has been a dramatic improvement in the IR thermal imaging systems to curtail the limitation of the background clutter aspect in the images. There are several existing techniques in detection of dynamic scene sequences of mines under solar illumination.; The objective of this thesis is to propose some new approaches to get better algorithms and a time period analysis over these techniques The mine detection technique is executed in three steps (i) preprocessing, (ii) segmentation and (iii) marker extraction. We propose several different algorithms, the first algorithm uses morphological operators in pre-processing step and in segmentation it uses watershed algorithm and morphological techniques in marker extraction mines. The second algorithm the preprocessing technique utilizes KLT, if priori knowledge of mine location exists then a mask is used at the time of pre processing otherwise just the images are used without the mask. The segmentation step in this technique involves thresholding the images followed by morphological operations. The marker extraction step involves morphological operators. The third algorithm tested for a different set of images utilizes filters and morphological operators for preprocessing step and thresholding with morphological operators for segmentation and morphological operators for marker extraction step. The mine detection algorithms on different images were found to be image dependent. The results were found to be satisfactory in most of the cases. In case of first two algorithms, they were found to give optimal performance in certain time stamps only. The time periods for best results have been tabulated. Approaches and areas for future work are pointed out.