3D Haar-Like Features for Pedestrian Detection

Abstract

One basic observation for pedestrian detection in video sequences is that both appearance and motion information are important to model the moving people. Based on this observation, we propose a new kind of features, 3D Haarlike (3DHaar) features. Motivated by the success of Haarlike features in image based face detection and differentialframe based pedestrian detection, we naturally extend this feature by defining seven types of volume filters in 3D space, instead of using rectangle filter in 2D space. The advantage is that it can not only represent pedestrian's appearance, but also capture the motion information. To validate the effectiveness of the proposed method, we combine the 3DHaar with support vector machine (SVM) for pedestrian detection. Our experiments demonstrate the 3DHaar are more effective for video based pedestrian detection.

Facial expression recognition using encoded dynamic features

“…Figure 4 shows some examples. In order to efficiently evaluate the performance of the dynamic binary patterns, we compare it with the haar-like volume features [25]. For simplicity, we denote our method as DBP and the haar-like volume features as 3D haar.…”

“…Similar features are proposed for video based face recognition in [11]. The volume haar-like features obtained an encouraging performance on the pedestrian detection and action analysis in [25]. Similar to the volume features, [27] designed the ensemble of the haar-like features in the temporal domain and combined them with a coding scheme for facial expression recognition.…”

Extracting Edge Voxels from 3D Volumetric Maps to Reduce Map Size and Accelerate Mapping Alignment

“…Sources of inspiration for volumetric edge detection come from the computer image processing literature including Haar features [7] and bilateral filtering which is mentioned in [6] and used in [8]. Experiments with many edge extraction methods (PCA, Difference of Gaussians, template matching, Haar filter responses) did not produce edge voxels corresponding to our intuition about them, did not remove planar voxels well due to noisy point data, or were limited to axis aligned edges.…”

Post-processing approaches for improving people detection performance

“…Although the vast majority of approaches are mainly based on appearance information, there are some approaches that combine appearance and motion information in order to improve the detection results. Some authors combine appearance and motion expanding previous detectors based on appearance to more than one frame [14], [17], [19]; in this way they are able to easily introduce motion information in the person model and add robustness to the detector.…”