Application of Human Action Recognition in Outdoor Environment

• Main Authors: HUANG, YU-PING, 黃煜斌
• Other Authors: Tsai, Yao-Hong
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/36863672255150874225

碩士 === 玄奘大學 === 資訊管理學系碩士班 === 105 === With the tremendous growth of science and technology, human action recognition is getting popularized, as well as being widely applied on various aspects such as surveillance at public areas, home care safety and other relevant applications. According to the past research, human action recognition mostly applied in indoor environment compared to outdoor environment. Due to more variability among outdoor environment, the recognition rate is comparatively low relative to indoor environment. For example, influence of climate, such as sun’s brightness changes or cloudy day. Furthermore, there are other variables such as scene objects’ disturbance, roadside falling leaves, or garbage left by others which are uncontrollable and hard to handle. However, the same variables happens in indoor environment are controllable and able to reduce inappropriate disturbance. In this paper, the main purpose is to design a human action recognition system that able to be applied in outdoor environment. We use unmanned aerial vehicle (UAV) to capture images and find out the local features via space-time interest points (STIP), then classify the local features through k-nearest neighbors algorithm (KNN) in order to successfully identify object’s movement in the image. With our proposed system, fall down and hand waving both action attained 77.7% and 78.2% of recognition rate respectively while running shows result of 93.0% recognition rate. Our approach shown good result of recognizing action in the circumstances of whether is there any happening, hence we hope that this proposed system can be applied in variant circumstances.