A More Effective Way of Target Segmentation of Multispectral Imaging Data Using Kmeans Algorithm

Conventional multispectral system are carried in aerial or satellite systems and are limited in temporal resolution and can only measure static targets. In this paper, we design a novel multispectral device that uses a camera array with different filters to continuously and dynamically capture 12-bit images with $960 \times 540$ resolution in 35 bands. Algorithms are also designed to correct for pixel shifts in different channels of images, as well as a method to cluster the captured multispectral images using the K-means algorithm.