Histogram of oriented gradients

About: Histogram of oriented gradients is a research topic. Over the lifetime, 2037 publications have been published within this topic receiving 55881 citations. The topic is also known as: HOG.

Real-time DSP implementation of Pedestrian Detection algorithm using HOG features

Abstract: Pedestrian Detection is the most critical safety application in automotive driver assistance systems. Histogram of Oriented Gradients (HOG) features is known to produce the state of the art results for this application. This feature is very compute-intensive and it is difficult to achieve real-time performance by direct porting of community software like OpenCV. In this paper, we discuss an efficient DSP implementation of this algorithm and also demonstrate how architecture aware design choices can lead to huge performance improvements. The algorithm was implemented and profiled on a Texas Instruments' C674x DSP, achieving a performance of 20 fps for a VGA resolution video sequence. Compared to OpenCV's HOG function, the proposed implementation is 130X faster without a significant loss of accuracy.

Real-time combustion state identification via image processing: A dynamic data-driven approach

Abstract: This paper proposes a framework for analyzing video of physical processes as a paradigm of dynamic data-driven application systems (DDDAS). The algorithms were tested on a combustion system under fuel lean and ultra-lean conditions. The main challenge here is to develop feature extraction and information compression algorithms with low computational complexity such that they can be applied to real-time analysis of video captured by a high-speed camera. In the proposed method, image frames of the video is compressed into a sequence of image features. Then, these image features are mapped to a sequence of symbols by partitioning of the feature space. Finally, a special class of probabilistic finite state automata (PFSA), called D-Markov machines, are constructed from the symbol strings to extract pertinent features representing the embedded dynamic characteristics of the physical process. This paper compares the performance and efficiency of three image feature extraction algorithms: Histogram of Oriented Gradients, Gabor Wavelets, and Fractal Dimension. The k-means clustering algorithm has been used for feature space partitioning. The proposed algorithm has been validated on experimental data in a laboratory environment combustor with a single fuel-injector.

3D spatial-temporal view based motion tracing in human action recognition

Abstract: The paper presents an extended approach of Motion History Image to trace the human motions in a video for recognizing the human actions. The video is represented as a 3D volume space and the trace of human motions are projected onto the three different views called 3D spatio-temporal plane. The extended view traces both the human shape and movement in different directions over the time. The Histogram of Oriented Gradients (HOG) features are extracted over all the projection plane which gives more distinct features for the action classification. Since HOG features data has high dimensionality, the optimal feature subset is selected by using the feature selection techniques. Finally, the Support Vector Machine (SVM) of multi-class classifier is used to identify the various actions of human. The various experiments are conducted on benchmark dataset KTH and results shows that the proposed method improves the action recognition rate compared to the existing methods.

Hand Gesture Recognition Using Kinect's Geometric and HOG Features

Abstract: Hand gesture recognition plays an important role in human computer interaction (HCI). Despite the recent progress, the accuracy of up-to-date methods is still not satisfactory. In this work, we proposed a comparative study to recognize six hand gestures in real time using the Kinect sensor. First, we developed a tracking method of the hand in the scene where the center of the palm is detected using depth data and projected into the color image. Second, geometric features were extracted from depth image and Histogram of Oriented Gradients (HOG) descriptors from the color image. Finally, based on those extracted features, a support vector machines (SVM) and an artificial neural network (ANN) are trained and compared.

Phase-Based Classification for Arm Gesture and Gross-Motor Activities Using Histogram of Oriented Gradients

Abstract: Micro-Doppler spectrograms are a conventional data representation domain for movement recognition such as Human Activity Recognition (HAR) or gesture detection. However, they present the problem of time-frequency resolution trade-offs of Short-Time Fourier Transform (STFT), which may have limitations due to unambiguous Doppler frequency, and the STFT computation may be onerous in constrained embedded environments. We propose in this paper an alternative classification approach based on the radar phase information directly extracted from high-resolution Range Map (RM). This novel approach does not use the aforementioned micro-Doppler processing, and yet achieves equivalent or even superior classification results. This shows a potential advantage for low-latency, real-time applications, or computationally constrained scenarios. The proposed method exploits the Histogram of Oriented Gradients (HOG) algorithm as an effective feature extraction algorithm, specifically its capability to capture the unique shape and patterns present in the wrapped phase domains, such as their contour intensity and distributions. Validation results consistently above 92% demonstrate the effectiveness of this method on two independent datasets of arm gestures and gross-motor activities. These were classified with three algorithms, namely the Nearest Neighbor (NN), the linear Support Vector Machine (SVM), and the Gaussian SVM classifiers using the proposed phase information. Feature fusion of different data domains, e.g. the modulus of the RM fused with the RM phase information, is also investigated and shows classification improvement specifically for the robustness of activity performances, such as the aspect angle and the speed of performance.