Indoor People Counting Method Based on Fingerprinting Localization with Kernel Fuzzy C-Means Clustering

Efficient and accurate people counting has become significant as it is essential in many applications such as smart guide and intelligent building. In this paper, a people counting approach based on fingerprinting localization is proposed for multiple regions in indoor environment. We obtain the number of people in each subarea by localizing each target in the environment. The approach is composed of the offline training phase and online counting phase. In offline phase, we construct the fingerprint database which is also called radiomap, and transform it through DIFF method to solve the problem of device diversity. Then we divide the whole region into different subareas utilizing Kernel Fuzzy C-Means (KFCM) Clustering algorithm to reduce the computational complexity. In online phase, we first select the nearest subarea of the target and then estimate its location. The location is compared with the boundaries of all subareas to determine the subarea it belongs to. When applied to multiple targets, the number of people in each region will be counted. Experimental results show that our proposed method is effective and the average people counting accuracy can achieve 90.24%.