Searching for plausible gas sources using SIFT features

According to many gas leak investigation statistic data, the occurrence probability of several objects or devices, e.g., tank or valve, is very high. In other words, Visual information plays an important role in gas source localization. In this paper, a simple and robust object detection technique for locating the suspicious gas sources is proposed. The scale invariant feature transform (SIFT) descriptor is invariant to image translations, rotations, scale changes, and robust to illumination changes. Numerous evaluation tests have been demonstrated that the robustness and repeatability of the SIFT descriptor outperforms other methods. Therefore, we use SIFT feature matching method to detect plausible gas sources. Experimental results show the efficiency and practicality of the approach for localizing a leaking ethanol bottle in complex indoor environments.