Realtime Feature Tracking and Object Recognition on Mobile Platforms

[摘要] We propose an efficient and robust feature tracking algorithm for mobile platforms, and demonstrate an object recognition mechanism that can be used for interactive mobile applications. Corner features are detected from captured video frames in a multi-scale image pyramid, and are tracked between consecutive frames efficiently using an optical flow-based tracking algorithm. In order to provide object recognition, SIFT descriptors are calculated on the tracked features, and quantized using a vocabulary tree. For each object, a bag-of-words model is learned by a simple user interaction to acquire multiple views. The learned objects are recognized as computing TF-IDF score for the set of features for a video frame. We describe the methods in detail and show the experiment results with a prototype realtime tracking and recognition application on a mobile phone.

[发布机构] UCLA Henry Samueli School of Engineering and Applied Science

[学科分类] 计算机科学（综合）