Intelligent deep learning-based vision system for human action recognition in drone-based videos
Loading...
Date
2019-06
Authors
Mohamad Safwan Bin Sahimi
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Human action recognition from a drone-based videos has many potential applications such as anti-terrorism activity or in a simple way can be used for human activity recognition such as running, walking, jumping and others. Drones have been widely adopted for many useful applications such as infrastructure inspection, agriculture monitoring, rescue, reconnaissance, surveillance and construction site monitoring. With deep- learning based computer vision now powering these drones, more unprecedented use in previously unimaginable applications. A very challenges computer vision problem to be tackled are related to many aspects including the variations in camera view, the distance from the camera, the changes in illuminations and weather conditions, the variation in the surrounding objects, as well as the present of object alike human. The implemented system is using approach which known fast R-CNN detection algorithm. This algorithm responsible to recognition human action. The system implementation on Python language, Anaconda and Google Tensorflow. The number of volunteer people will be recorded by a drone-based camera at various places and weather conditions to assess the performances of the proposed vision system. A volunteer will be recorded from different views while they are performing different activities such as walking, running, hand boxing, hand waving, and standing. The system will investigate the effectiveness of using fast R-CNN approach to locate and recognize the activity of the pedestrian inside the captured images with various video were recorded at different places, various views and daytime.