Algorithm for military object detection using image data

This paper deals with problems of surface object detection in urban environmental conditions using infrared and visible part of the electromagnetic spectrum. The aim of this work is to compile an algorithm for detection and selection of objects of interest in urban built-up background, of civil automobiles which closely resemble similar military equipment. Those objects are captured by infrared camera and visible camera in different outdoor conditions - for instance during daytime in all seasons. The object detection in infrared spectrum is based on assumption that the objects of interest have sufficient contrast against the background, while the object detection in visible spectrum is based on assumption that the objects of interest are given a definite colour. The basic task of the object detection in image data is a selection of an optimal threshold value to be converted from intensity image to binary image. As a result of the variable environmental conditions and variable colours of the objects, the optimal threshold value can be changed. In this study, a compromise threshold value was used to suit the different outdoor environmental conditions. First, algorithms were tested on static images, where the optimal threshold value for conversion to binary images was determined. Then, the object detection, object selection and object tracking was done by video processing. The selection of the detected objects was based on their position in the image frame, their size, temperature distribution, motion rate, etc. This algorithm can then be applied in the vehicles using thermo camera for monitoring. For example, in the search and rescue operations in difficult weather conditions. The designed algorithms were tested in program MATLAB and MATLAB-SIMULINK. This paper also deals with new algorithm using template matching. The idea is to reduce processing rate and increase precision for the object detection and object selection when using template matching technique. This algorithm was tested in programming environment Matlab, on pictures as well as on video sequences that were acquired by cameras Ray Cam C.A. 1884, thermo IMAGER 160, Canon operated in visible band from internet source.