PDF文档 114 Video object matching across multiple independent views using local descriptors and adaptive learnin