Classification of chest X-ray images using Machine Learning and Histogram of Oriented Gradients

Este trabalho propõe um modelo de aprendizado de máquina para classificar e detectar a presença de pneumonia a partir de uma coleção de amostras de radiografias do tórax. Ao contrário da maioria dos trabalhos que utilizam abordagens de aprendizado profundo para classificar se a imagem é de um pulmão com pneumonia ou não, ou seja, duas classes para assim alcançar um desempenho de classificação notável, este modelo utiliza Histograma de Gradientes Orientados para extrair características de uma determinada imagem de raio-X de tórax e classificá-la em três classes, determinando se uma pessoa está ou não infectada com pneumonia viral ou bacteriana. Apesar de uma maior complexidade e utilização de modelos tradicionais de aprendizado de máquina, a maior acurácia alcançada foi de 91.32% superior a de trabalhos que utilizam redes profundas e buscam resolver o mesmo grau de complexidade.

Investigation on Recognition Performance of Harvesting Robot Using Regions of Interest Histogram of Oriented Gradients Feature Based on Improved Fuzzy Least Square Support Vector Machine

In order to solve the problems such as big errors, lack of universality, and too much time consuming occurred in the recognition of overlapped fruits, an improved fuzzy least square support vector machine (FLS-SVM) is established based on the fruit ROI-HOG feature. First, the RGB image is transformed into saturation and value (HSV) image, and then the regions of interest (ROI) are detected from HSV color information. Finally, the histogram of oriented gradients (HOG) feature of ROI will be used as the input of FLS-SVM pattern recognizer to realize the recognition of picking fruit. In addition, the verified FLS-SVM is used to investigate the recognition performance of harvesting robot using regions of interest histogram of oriented gradients feature. The results reveal that the vector sizes are effectively reduced and a higher detection speed is achieved without compromising accuracy relative to conventional approaches. Similarly, the detection accuracy for the learning samples, the isolated fruit, the overlapped fruit, and the background can achieve 99.50%, 96.0%, 89.9%, and 97.0%, respectively, which shows the good performance of the proposed improved ROI-HOG feature recognition method.

Food Intake Vision-Based Recognition System via Histogram of Oriented Gradients and Support Vector Machine for Persons With Alzheimer’s Disease

Due to cognitive decline, individuals with Alzheimer’s often suffer from malnutrition, forgetting to eat, even if food is presented. Therefore, assistance with feeding is needed. In this paper a vision-based system for monitoring of eating patterns is presented. Upper Body Region (UBR) is detected using Viola-Jones method, a histogram of oriented gradients (HOG) is generated for feature extraction, and a support vector machine (SVM) is used to distinguish eating versus non-eating. To reduce false positive results, Haar-like features are used to detect hands while moving between served food and mouth within the identified upper body region (UBR). A combined template image (CTI) method is proposed in this work to eliminate false positive hand detections where 30 hand eating posture images have been selected and combined into one template image. Matching implemented using CTI is 2.86 times faster than matching the subject to the 30 images separately. Experimental simulation used 33 videos of 163840 frames indicates that the proposed method achieves a high accuracy of 90.65%.

p-norms of histogram of oriented gradients for X-ray images

<span>Lebesgue spaces (</span><em><span>L^p</span></em><span> over </span><em><span>Rⁿ</span></em><span>) play a significant role in mathematical analysis. They are widely used in machine learning and artificial intelligence to maximize performance or minimize error. The well-known histogram of oriented gradients (HOG) algorithm applies the 2-norm (Euclidean distance) to detect features in images. In this paper, we apply different </span><em><span>p</span></em><span>-norm values to identify the impact that changing these norms has on the original algorithm. The aim of this modification is to achieve better performance in classifying X-ray medical images related to of COVID-19 patients. The efficiency of the </span><em><span>p</span></em><span>-HOG algorithm is compared with the original HOG descriptor using a support vector machine implemented in Python. The results of the comparisons are promising, and the </span><em><span>p</span></em><span>-HOG algorithm shows greater efficiency in most cases.</span>

Object Detection and Tracking using Multiple Features Extraction

In most of the video analysis applications, object detection and tracking play vital role. Most of detection and tracking algorithms fail to predict multiple objects with varying orientation. In this paper, the goal is to identify and track multiple objects using different feature extraction methods like Locality Sensitive Histogram, Histogram of Oriented Gradients and Edges. These features are subjected to train classifier that can detect the object of different orientations. Experimental results and performance evaluation depicts the proposed method which uses LSH performs well with an increased accuracy of 98%. This method can precisely track the object and can be utilized to track under different scale and pose variations.