Enhancing the recognition of medication intake using a stereo camera

In this paper, we propose a simple method for estimating the depth of objects with a stereo camera in a medication intake recognition context. A template for each detected object in the left image is matched against the right image, for calculating its horizontal disparity. Then the disparity is used for estimating the distance from camera, so that we can decide whether two objects in occlusion are in genuine contact or not. Experiments show that, the proposed approach is simple and efficient for detecting a true contact between objects in occlusion, a necessary step to further improve a video surveillance system for detection and recognition of medication intake.