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.

Human Action Recognition: Contour-Based and Silhouette-Based Approaches

Abstract: Human action recognition in videos is a desired field in computer vision applications since it can be applied in human computer interaction, surveillance monitors, robot vision, etc. Two approaches of features are investigated in this chapter. First approach is a contour-based type. Four features are investigated in this approach such as Cartesian Coordinate Features (CCF), Fourier Descriptors Features (FDF), Centroid-Distance Features (CDF), and Chord-Length Features (CLF). The second approach is a silhouette-based type. Three features are investigated in this approach such as Histogram of Oriented Gradients (HOG), Histogram of Oriented Optical Flow (HOOF), and Structural Similarity Index Measure (SSIM) features. All these features are simple to compute, efficient to classify, and fast to calculate. Therefore, these features demonstrate a promising field for human action recognition. Moreover, the classification is achieved using two classifiers: K-Nearest-Neighbor (KNN) and Support Vector Machine (SVM). The experimental results demonstrated that these features have a promising potential and useful for the human action recognition in videos.

Intensity self similarity features for pedestrian detection in Far-Infrared images

Abstract: Pedestrian detection is an important but challenging component of an Intelligent Transportation System. In this paper, we describe a pedestrian detection system based on a monocular vision with a Far-Infrared camera (FIR). We propose an original feature representation, called Intensity Self Similarity (ISS), adapted to pedestrian detection in FIR images. The ISS representation is based on the relative intensity self similarity within a pedestrian region of interest (ROI) hypothesis. Our system consists of two components. The first component generates pedestrian ROI hypothesis by exploiting the specific characteristics of FIR images, where pedestrian shapes may vary in large scale, but heads appear usually as light regions. Pedestrian ROI are detected, with high recall rate, due to a Hierarchical Codebook (HC) of Speeded-Up Robust Features (SURF) located in light head regions. The second component consists of pedestrian hypothesis validation, by using a pedestrian full-body classification based on the ISS representation, with Support Vector Machine (SVM). For classification, we retained two feature descriptors: the Histogram of Oriented Gradients (HOG) descriptor and the original ISS feature representation that we proposed for FIR images. The early fusion of these two features enhances significantly the system precision, attaining an F-measure for the pedestrian class of 97.7%. Moreover, this feature fusion outperforms the state-of-the-art SURF descriptor proposed previously. The experimental evaluation shows that our pedestrian detector is also robust, since it performs well in detecting pedestrians even in large scale and crowded real-world scenes.

An efficient multi-scale overlapped block LBP approach for leaf image recognition

Abstract: In this paper, an effective method based on multi-scale overlapped block LBP is proposed for plant leaf image recognition. Firstly, multi-scale pyramid is employed in order to improve the leaf data utilization. For each scale, each training image is divided into several equal overlapping blocks to extract the LBP histograms. Then, the PCA method is used for LBP feature dimension reduction. Finally, the recognition experiments are performed by using the SVM classifier. We compare the proposed method with Histogram of Oriented Gradients (HOG) method and Inner-Distance Shape Context (IDSC) method on Swedish leaf dataset and our ICL leaf dataset. The experimental results show that the proposed method achieves better performance than IDSC and HOG.

Improved multi-classification of breast cancer histopathological images using handcrafted features and deep neural network (dense layer)

Abstract: Breast cancer (BC) classification has become a point of concern within the field of biomedical informatics in the health care sector in recent years. This is because it is the second-largest cause of cancer-related fatalities among women. The medical field has attracted the attention of researchers in applying machine learning techniques to the detection, and monitoring of life-threatening diseases such as breast cancer (BC). Proper detection and monitoring contribute immensely to the survival of BC patients, which is largely dependent on the analysis of pathological images. Automatic detection of BC based on pathological images and the use of a Computer-Aided Diagnosis (CAD) system allow doctors to make a more reliable decision. Recently, Deep Learning algorithms like Convolution Neural Network have been proven to be reliable in detecting BC targets from pathological images. Several research efforts have been undertaken in the binary classification of histopathological images. However, few approaches have been proposed for the multi-classification of histopathological images. The classification accuracy produced by these approaches are inefficient since they considered only texture-based extracted features and they used some techniques that cannot extract some of the main features from the images. Also, these techniques still suffered from the issue of overfitting. In this work, handcrafted feature extraction techniques (Hu moment, Haralick textures, and color histogram) and Deep Neural Network (DNN) are employed for breast cancer multi-classification using histopathological images on the BreakHis dataset. The features extracted using the handcrafted techniques are used to train the DNN classifiers with four dense layers and Softmax. Further, the data augmentation method was employed to address the issue of overfitting. The results obtained reveal that the use of handcrafted approach as feature extractors and DNN classifiers had a better performance in breast cancer multi-classification than other approaches in the literature. Moreover, it was also noted that augmentation of data plays a key role in further improvement of classification accuracy. The proposed method achieved an accuracy score of 97.87% for 40x, 97.60% for 100x, 96.10% for 200x, and 96.84% for 400x for the magnification-dependent histopathological images classification. The results also showed that the proposed method for using the Handcrafted feature extraction method with DNN classifier had a better performance in multi-classification of breast cancer using histopathological images than most of the related works in the literature.