Abstract—In this paper, we present an algorithm to detect people with both color and thermal images. This technique is an extension of the modified Implicit Shape Model algorithm that we had developed earlier. The algorithm can process both color and thermal images in two modes of White-Hot and Black-Hot using a single codebook generated from samples of thermal images obtained from people. The number of the samples used in the codebook is very small compared with other techniques. In the first step, the location of people is defined based on their proposed centers using the Implicit Shape Model (ISM). Then, an auto generated threshold process is used to detect people from the concentrated proposed center's density. The implementation of this technique does not need high computing power and results in improvements in the speed performance and reduction in the hardware cost. The system was evaluated using 12 image sets, six for indoor environments and six for outdoor environments. Each case included three sets of color images plus three sets of thermal images. This algorithm was implemented on a mobile robot prototype designed for rescue assist missions and the tests conducted provided promising results for detecting individuals in image sets with difficult scenarios.

Index Terms—thermal image, color image, ISM, people detection, rescue

Cite: Laith A. H. Al-Shimaysawee, Ali H. A. Aldabbagh, and Nasser Asgari, "A Flexible Method for Detecting Individuals with Color and Thermal Cameras," Journal of Image and Graphics, Vol. 5, No. 2, pp. 39-46, December 2017. doi: 10.18178/joig.5.2.39-46