China
Africa
Explore chapters and articles related to this topic
Real Time Face Detection using Raspberry Pi 3B+
Published in Rajesh Singh, Anita Gehlot, P.S. Ranjit, Dolly Sharma, Futuristic Sustainable Energy and Technology, 2022
V Sai Ganesh Reddy, Navjot Rathour, G Siddhu Ganesh, Yaswanth, Satish Kumar
Histogram of oriented gradients also known as HOG, is a feature descriptor which helps to detect the objects in a computer vision and for the image processing and extract the features from image data. This HOG descriptor have some unique features from the remaining descriptors. The Histogram of oriented gradients mainly concentrates on the shape and structure of the object. HOG descriptor identifies the edge direction. This can be processed by extraction of orientation as well as gradients of the edges. Apart from these the orientations are processed in a localized portions which is the full image is divided into small parts and for every part the orientation and gradients will be calculated. This HOG will create a histogram for every small parts individually. And there histogram are formed with the orientations and gradient of a pixel values, so this is named as a Histogram of oriented gradients.
EEMS2015 organizing committee
Published in Yeping Wang, Jianhua Zhao, Advances in Energy, Environment and Materials Science, 2018
In this paper, a pedestrian detection algorithm based on histogram of oriented gradients and support vector machine was proposed by Dalal Navneet and Triggs Bill in 2005. The Histogram of Oriented Gradients (HOG) is a feature descrip- tor used in computer vision and image processing for the purpose of object detection. The tech- nique counts occurrences of gradient orientation in localized portions of an image. This method is similar to that of edge orientation histograms, scale-invariant feature transforms descriptors, and shape contexts, but differs as it is computed on a dense grid of uniformly spaced cells and uses over- lapping local contrast normalization for improved accuracy. Navneet Dalal and Bill Triggs, research- ers for the French National Institute for Research in Computer Science and Control (INRIA), first
A classification method for artistic images on feature computation
Published in Xiaoling Jia, Feng Wu, Electromechanical Control Technology and Transportation, 2017
Raoshan Xu, Zhengxing Sun, Chen Ma
Spatial features were used to describe the relative spatial position of objects in an image or area, such as the vertical and horizontal relations between areas. Spatial features are used on the basis of segmentation, which are generally developed for detecting the contents of the image by area identification. The Histogram of Oriented Gradients (HOG) is used to calculate the gradient direction histogram to form local area features, which is mainly used to detect objects in an image. In implementation, it divides the image into small areas, namely unit cells, and then calculates the histogram for each pixel of the unit cell, at the direction gradient or edge. The results obtained from these steps are then passed on to the final stage of combination, thereby putting these cell histograms together to form HOG features of the whole image. The Space Pyramids Histogram of Oriented Gradients (PHOG) (Bosch, 2007) extended the principle of HOG by adding the space pyramid matching method for achieving better spatial characteristics. In this paper, we divided 360 degrees space into 8 bins for features extraction, according to the gradient direction, and let L=3 for spatial pyramid matching. In this way, each image can be extracted as a 680-dimensional feature vector.
A Systematic Review of Recent Machine Learning Techniques for Plant Disease Identification and Classification
Published in IETE Technical Review, 2023
In [6], S. R. Maniyath et al. (July 2019) applied the Random Forest algorithm on papaya leaf images and compared the results with the other classifiers, i.e. SVM, K-NN, Logistic Regression, CART, Gaussian Naive Bayes. The Histogram of oriented gradients (HOG) is applied as a feature descriptor for object detection, which utilizes three essential components as a descriptor which are: Color Histogram, haralick texture, and Hu moments. The feature vector is given as input to distinguish the classifiers; the results show that the Random Forest Algorithm provides the highest accuracy rate 70.14%. The application of extended image processing techniques and machine learning algorithms to detect plant illnesses at various stages of the plant's life cycle was investigated in this research study [29]. In [30], four soybean phenotypes (two susceptible and two moderately resistant) were collected using hyperspectral technology at five different time points following post infection. The research study used both spatial and spectral information for disease identification. An optimization problem was formulated to determine the optimal waveband combination from a collection of 240 wavebands for discriminating between normal and diseased charcoal rot stems. The objective of the optimization was to find the best waveband fusion that maximizes the classification performance. Using six different hyperspectral waveband combinations, the GA-SVM model achieved an F1 score of 0.97 and a classification accuracy of 97% for the entire test set.
Tyre pattern image retrieval – current status and challenges
Published in Connection Science, 2021
Liu Ying, Liu Qiqi, Fan Jiulun, Wang Fuping, Fu Jianlong, Yuan Qingan, Chiew Tuan Kiang, Ling Nam
Combined with the unique texture patterns and HOG features in tyre pattern images, Liu, Ge, et al. (2019) uses circular feature extraction method to extract texture gradient features of the image, and then uses these texture gradient features to match against dataset images. The HOG-DG (Histogram of Oriented Gradients – Dominant Gradient), an improved version of HOG, has superior robustness to illumination variation, camera’s shooting distance and angle. This advantage is more profound in tyre tread pattern classification, along with lower computation requirement and higher classification accuracy. This paper tested on the CIIP-TPID tyre tread pattern data set and obtained 81.5% classification accuracy. The HOG-DG feature extraction algorithm flow is shown in Figure 6:
Intelligent System Utilizing HOG and CNN for Thermal Image-Based Detection of Wild Animals in Nocturnal Periods for Vehicle Safety
Published in Applied Artificial Intelligence, 2022
Yuvaraj Munian, Antonio Martinez-Molina, Dimitrios Miserlis, Hermilo Hernandez, Miltiadis Alamaniotis
In the last decade, image processing and machine learning tools in automobiles have led to several advanced systems like self-driving cars, pedestrians’ autodetection, and cruise control, reducing fatalities. This research focuses on detection application by combining an image processing technique with a deep learning tool, and more particularly, the Histogram of Oriented Gradients (HOG) with a convolutional neural network (CNN), as it has been preliminarily introduced in (Munian, Martinez-Molina, and Alamaniotis 2020). Notably, deep learning models are the most effective in image processing applications since they distinguish the high-level features from the low-level features (Wu, Sahoo, and Hoi 2020).