Optimized Scale Invariant HOG Descriptors for Object and Human Detection

Abstract

This article presents a scale variation approach to identify objects and humans in video sequences using histogram of gradient descriptor. A significant restriction in HOG descriptors is its variations with scale changes and illumination changes, as is frequently the considered case. We recommend unique SIO-HOG descriptors that are figured to be invariant to scale changes. The system associates the benefits of adoptive bin selections and sample resizing in the object recognition process. We analyze the effect of PCA transform based feature selection process on object detection performance, ultimately the finite scale range, adoptive orientation binning in non-overlapping descriptors is all main thing for nominal detection rate. HOG feature vector over complete search window is computationally more exclusive, to acquire more precise set of features with finite Euclidean distance to classify them using KNN classifier. This new approach provides near-perfect ways of separating humans from other objects. The whole object detection system was assessed on a few test samples from real-world data sets and compared beside a publicly available pedestrian detection data base.

Target Detection in Video Images Using HOG-Based Cascade Classifier

“…The gradient magnitude and orientation for each block's individual pixels are constructed. A 9-bin (ranging from 0° to 180° with a step of 20°) histogram of the oriented gradient computation for each block in the image based on the orientation and magnitude matrices Each bin is a vector feature [27]. It is also helpful to contrastnormalize the local responses before utilizing them for improved invariance to light, shadowing, etc.…”

“…It is used to detect objects by extracting set of object features using multiple levels. Where, each level is an ensemble of weak classifiers [27]. The ROI detected by sliding window over the entire input image.…”