An Improved RANSAC Algorithm Based on Adaptive Threshold for Indoor Positioning

The smart city is an important direction for the development of the highly information-based city, and indoor navigation and positioning technology is an important basis for the realization of an intelligent city. In recent years, indoor positioning technology mainly relies on WiFi, radio frequency identification (RFID), Bluetooth, and so on. Yet, the implementation of the above method requires the relevant equipment to be laid out in advance, and it is only suitable for indoor positioning with low accuracy requirements owing to interference and fading of the signal. The visual-based positioning technology can achieve high-precision positioning in enclosed, semienclosed, and multiwalled indoor environments with strong electromagnetic interference by means of epipolar geometry and image matching. The visual-based indoor positioning mostly uses the random sample consensus (RANSAC) algorithm to estimate the fundamental matrix to acquire the user's relative position. The traditional RANSAC algorithm determines the set of inliers by artificially setting a threshold to estimate the model. However, since the selection of the threshold depends on experience and prior knowledge, the reliability of the positioning results is not robust. Therefore, in order to improve the universality of the algorithm in indoor environments, this paper proposed an improved RANSAC algorithm based on the adaptive threshold and evaluated the real-time and accuracy of the algorithm by using an open-source image library. Results of the experiment show that the algorithm is more accurate than the traditional RANSAC algorithm in an enclosed and semienclosed multiwalled indoor environment, with fewer iterations.