Research On Face Detection Algorithm Combining Depth Data And AdaBoost Algorithm

Face detection is one of the most studied topics in the field of computer vision. The current research on face detection mainly uses two-dimensional color image. Yet the two-dimensional image has the inherent limitation of information loss resulted from change of illumination conditions, which can affect the performance of face detection algorithm. With the continuous development of hardware conditions, the illumination invariance of depth data has made it more robust in face detection and recognition. Based on the conventional face detection algorithm, the performance and robustness of the algorithm can be effectively improved by combining the depth data.Based on in-depth analysis of the Ada Boost face detection algorithm, the characteristics of Ada Boost algorithm, combined with the depth data, can be used to conduct face detection. Under normal illumination conditions, the Ada Boost algorithm with a lower region merging threshold can be used to ensure a high recall rate. Then, it is possible to reduce the false detection rate by using the actual size of face and the depth information standard deviation of the extended face region, which are calculated by the depth data, to filter the non-face region. Under poor illumination conditions, Ada Boost algorithm can’t effectively detect the face region; yet by using depth data with contour chamfer matching method to get candidate head regions, and then confirming the candidate head regions, it is possible to obtain the final face area.Test is conducted on the test set containing three-dimensional data acquired by the Kinect(including two-dimensional color image and depth data). The results show that compared with Ada Boost face detection algorithm for identical merging threshold under normal light conditions, the face detection algorithm which uses depth data filtering is able to effectively reduce the false detection rate and improve the accuracy of algorithm while ensuring the recall rate. Under poor illumination conditions, when it is difficult for Ada Boost algorithm to detect human face, the face detection algorithm based on depth data can improve the performance of face detection.