Indoor Navigation for a Complex Parking Building Based on Computer Vision

With the increasing number of vehicles, more and more complex indoor parking lots are built to fulfil the booming parking needs. Consequently, people nowadays spend more and more time inside the parking lot to find an available spot. There are some existing approaches in resolving this problem, including the usages of GPS, Bluetooth beacons, radar imaging. However, these existing solutions have various technical limitations. In order to reduce the cost of the indoor parking system and make it easy to be deployed, in this paper, an accurate, low-cost and intelligent navigation system is developed based on computer vision algorithms. The system shows its strengths from three pespectives: 1) It can achieve a good accuracy in terms of mapping and localization. 2) The system is infrastructure-free except a server and phone cameras. 3) The system is able to update the map automatically and retrieves navigation information in real time. Unlike other existing indoor localization frameworks, this system could be easily deployed. Based-on evaluation experiments conducted in several different parking buildings, the results show an accuracy of 98%.