Improved Human Detection Using Image Fusion

Image fusion is fundamental to several modern day image processing applications. It is often a vital preprocessing procedure to many computer vision and image processing tasks which are dependent on the acquisition of imaging data via sensors, such as infrared and visible. One such task is that of human detection. In this paper, we present improvements to our shape and heat flow-based technique of detection and classification of humans in unrestricted poses with the addition of image fusion. We focus on both rural and urban environments and demonstrate the effectiveness of using image fusion as a preprocessing procedure for improved human detection and classification. Extensive simulations using MWIR images were conducted and promising results are obtained. Receiver Operating Characteristic (ROC) analysis also showed excellent performance of the SVM-based human classification.