Weighted Pattern Vector for Object Tracking with the Use of Thermal Images

Abstract

Listen

Translate article

The goal of the following paper was to devise a method of object tracking in a case when we have preliminary knowledge of an object. Tracking the object which state or some properties are known may be conducted with the pattern vector modification basing on this knowledge. The object which has the same visual images and different thermal images, depending on the time the object was moving, may be an example of the abovementioned case. One of such situations, presented as an example in the paper, was the change of the car engine temperature which modified the thermal image whereas the visual image was the same for different engine temperatures. In each moment during object tracking image processing system used thermal and visible images. The authors assumed that we acquire the visual and thermal images in each moment of object tracking. The pattern vectors of the tracked object based on the visual and thermal images were computed separately. The pattern vector and current feature vector for an image of a given type were used to compute the distance between the object pattern vector and feature vector calculated for a given location of the aperture. It was calculated for both: the visual and thermal images. The crux of the proposed method is the algorithm of setting the correct value of the weight with which the visual and thermal pattern vectors are used to calculate the distance. The authors proposed the method based on the knowledge about the object location in its initial position. The pattern vector calculated in this position was treated as a correct and unchangeable one. Assuming that the thermal pattern vector was more useful for object recognition, we proposed the method of increasing weight for a part of the pattern vector which part was based on a thermal image. Finally, this study presented the examples of the object recognition by means of the developed method.