Camera system long wave infrared for automatic door

Abstract

The core of this project is based on image processing and the implementation of algorithms capable of reliably detecting the trajectory and the presence of persons within a predefined perimeter for the opening of automatic doors using an infrared camera. The detection of the trajectory is based on the movement. The direction of the person is then used for controlling the opening of the door. Trajectory information can also be used to count incoming or outgoing persons. The algorithm presented here uses a series of methods well defined from each other. Each frame goes through a series of modules which extract and compute several data based on what is observed. The overall algorithm is thus easy to understand since each step is well defined. First of all, a background subtraction is performed in order to keep only relevant pixels. Then features are extracted by using Shi–Tomasi corner detection algorithms. Features represent important groups of pixels as corners or edges. To make the features move with the person where they came from, the optical flow method is used. The method is based on the Lucas-Kanade method. To define only one tracking point for each person walking in the field of view, a data clustering combines with an orient object view has been implemented. Position, speed, and direction are now known for each person walking in front of the camera, and can be used to control the opening of the door. Passive infrared sensors for automatic door are a good solution, most specifications are reached by the system presented here. All results in this report demonstrate the potential of this technology in the framework of this application. The picture below shows a person being tracked by the system.