Abstract
This paper presents a robust machine learning based computational solution for human detection. The proposed mechanism is specifically applicable for pose-variant situations in video frames. In order to address the pose variance problem, features are extracted using an improved variant of Histograms of Gradients (HoG) and local Binary Pattern features (LBP). The two feature sets are combined to form a feature vector based on different poses and human shapes, while a support vector machine (SVM)-based classifier is used for detection. Common issues faced by current approaches include false and missed detections in frames with robust feature-sets consisting of improved HoG features and LBP features with rotation information. The proposed detector model performs efficiently; the miss rates are reduced, the true positives are increased, and the accuracy is improved. Some false detections for human look alike objects are also observed. A diverse dataset depicting different poses is used for training purposes. A challenge test dataset is used to test the performance of the proposed approach against current state-of-the-art detectors to verify the performance. Receiver operating characteristic (ROC) curves are plotted to compare and evaluate the results based on miss rates and true positives, which demonstrate the proposed model achieves optimal results.