ANALYSIS OF CLASSIFICATION PROCESS FOR VIDEO SCENE UNDERSTANDING

Scene understanding is a process observing scenes which humans are used as models to understand them. This project is focused on analyzing the classification process for video scene. The algorithm used in this project will later be evaluated to see its performance. The process starts with human action video obtained from KTH dataset as the input for video processing. In the frame process, important information will be extracted from the video images accordance to frame sequence where Spatial Temporal -Interest-Point (STIP) is used based on Harris’ Corner detection. The human motions will then be classified by utilizing K-Nearest Neighbor (K-NN) method into their desired group of actions such as walking, running, or clapping. K-NN is an effective classifier since it works well with small datasets. However, K-NN does not work well with large datasets because it required longer timeframe.