Comparison of keypoint detection methods for indoor and outdoor scene recognition

We describe an experimental study, based on several million video scenes, of seven keypoint detection algorithms: BRISK, FAST, GFTT, HARRIS, MSER, ORB and STAR. It was observed that the probability distributions of selected keypoints are drastically different between indoor and outdoor environments for all algorithms analyzed. This paper presents a simple method for distinguishing between indoor and outdoor environments in a video sequence. The proposed method is based on the central location of keypoints in video frames. This has lead to a universally effective indoor/outdoor environment recognition method, and may prove to be a crucial step in the design of robotic control algorithms based on computer vision, especially for autonomous mobile robots.

---

Opracowanie ze środków MNiSW w ramach umowy 812/P-DUN/2016 na działalność upowszechniającą naukę (zadania 2017).