Modified multi-target recognition based on CamCom

Targets recognition methods based on Camera Communication (CamCom) face these challenges: low communication bandwidth, high Bit Error Rate (BER) due to bad frames and long response time. In this paper, a modified recognition method based on CamCom is proposed. In the method, Light-Emitting Diode (LED) which embedded in targets emits ON-OFF keying periodic light signals, basic image processes segment and capture the signal from sequential images, the features of signals are extracted by Fast Fourier Transform (FFT), and then Support Vector Machine (SVM) is used to classify each signal and to identify the corresponding target. Experiments showed that compared with CamCom based or vision based recognition methods, this method can locate and recognize targets fast and accurately, has short response time to new targets and can endure occasional error data.