Deteksi Pornografi pada Karakter Animasi 2D dengan KNN (K-Nearest Neighbors) Menggunakan Fitur HSV

Abstract

The development of animation technology is very rapid, both in 2D and 3D forms. Most animators often create female animated characters for various fields such as games, commercials or other anime. Along with the development of animation, there are also negative and positive impacts, where the negative impact is the presence of symbols that lead to pornography. Most of the animations containing negative symbols are spread through the internet, which can be accessed easily by anyone, regardless of age. Which can result in addiction to viewing pornography to other negative behaviors. Artificial intelligence, which is currently developing rapidly, also allows this research to aim at early detection of animated characters that contain pornographic elements. By classifying 2D female animation using the KNN algorithm method with HSV feature extraction. The HSV feature is quite good at detecting complex colors in 2D animated characters. Where the HSV feature can distinguish skin color from other colors in 2D animation. The dataset used consists of 3 categories, namely pornographic images, semi-pornographic images, and non-pornographic images. Using the KNN method, it is possible to classify pornographic, semi-pornographic and non-pornographic images. In this study, the results of HSV feature extraction using the KNN classification method obtained the highest accuracy value of 63.16%.