The human detection in images using the depth map

In today world the necessity for the autonomous mobile robots and vehicles is increasing. The safety autonomous moving demands the reliable and fast detection algorithms. The Histogram of Oriented Gradients (HOG) descriptors show significantly outperforms the existing feature sets for a human detection. Though the given method has a lot of type I errors. The amount of these errors can be decreased by using the object distance information. This paper presents a robust human detection method using pairs of color frame and depth map. During the experiment, we used color images and maps of depth received from the Kinect v2 visual sensor. During the first step in our detection experiment we processed the whole frame with the HOG descriptor and received regions of interest. Then on the second step we determined the approximate distance to this region and compare its value to the range of possible human height and width values on that distance. The experimental results show that the new proposed method of HOG and distance restriction combining provides lower false positive and increase the precision in comparison to the HOG method without using the depth map. It gives opportunities to train more sensitive classifiers, which can provide the higher recall values. Consequently, we can increase the safety moving of the autonomous mobile robots and vehicles.