Camera-Based Light Emitter Localization Using Correlation of Optical Pilot Sequences

Visual identification of objects using cameras requires precise detection, localization, and recognition the in field-of-view. The visual problem is very challenging when look identical features between distinct are indistinguishable, even with state-of-the-art computer vision techniques. becomes significantly more themselves do not carry rich geometric photometric features, for example, tracking light emitting diodes (LED) visible communication (VLC) applications. In this paper, we present a camera based solution where or regions interest tagged an actively transmitting LED. Motivated by concept pilot symbols, typically used synchronization channel estimation radio systems, LED transmits unique symbols which detected across series image frames our proposed spatio-temporal correlation algorithm. We setup as localization on image, involves identifying (pixels) unique ID corresponding to algorithm trace-based evaluation accuracy under real-world conditions including indoor, outdoor, static mobile scenarios. addition micro-benchmarking technique different parameter configurations, show that outperforms comparative techniques, including, color support-vector machine (SVM) learning, you only once (YOLO), convolutional neural network (CNN) deep learning object tool.