Detection of HOG Features on Tuberculosis X-Ray Results Using SVM and KNN

Abstract

Image processing is one of the sciences in image processing which can involve several other techniques such as data mining techniques, in this case the detection of an image. Images are generally carried out classification which results in accurate detection wherein the detection of an image is carried out by extracting the features so that the image can be recognized by computation. One of the extract features that are superior and easy to apply in computational techniques is HOG (Histogram OF Oriented Gradients). The HOG feature can be useful in helping detect images in the form of Tuberculosis xray. After extracting the features, then the classification is carried out using 2 methods that are good for learning levels such as KNN (K-Nearest Neighbor) and SVM (Support Vector Machine). The results of this paper in the detection of HOG Tuberculosis X-ray with KNN for positive images got an accuracy of 77.95% while the negative ones got an accuracy of 78.65%. The results of HOG detection on Tuberculosis X-ray results with SVM on images that were positive got an accuracy of 65.75% while those who were negative were 79.39%.