To overcome the limitations of object recognition based on appearance, we integrate a spatial relationship between the local key points and the position of the object's center. A voting technique is applied to estimate the area of the object and then construct a bounding box to capture the object. A blended appearance model is introduced by a recovery image to help deal with false detection problems. The experimental results show that our method can improve the detection time of objects while preserving the results of average precision. In addition, our method can improve the precision of the classification of the view.
In recent years, there has been a growing interest in the use of RGBD video sequence information for object detection. Online learning and real-time object detection are important and challenging tasks in the area of computer vision research. People, especially the elderly, tend to forget where important things are and spend a great deal of time looking at them. In this article we propose an RGBD information system for the robust detection of object localization and help people to find the target object immediately with a simple stage of online training in advance.