Cross Camera Object Tracking in High Resolution Video Based on TLD Framework

TLD is used to fulfill single and cross camera tracking in high resolution video. We adopt down-sample method in high resolution video and propose a mechanism of shared online classifier, which is possessed and updated by all cameras simultaneously and online. Experiment results show that, compared to original TLD algorithm, down-sample method can help reduce false detection rate and improve the track speed effectively in high resolution video. The shared online classifier can make full use of multimedia big data to boost the performance of classifier. It is an efficient method and tool for multimedia big data analytics.