Motion Recognition Based on Angle-reg Network
Zhuoqun Yu, Guhong Zhang, Wenjun Zhang, Jianliang Shen, Zhonghe Jin · 2019
Motion recognition of moving target has become one of the hotspots of object detection. Recognition of moving target based on vision is one of the most important methods to solve problems about mention recognition. However, it is complex to achieve real-time recognition for traditional moving target recognition, which needs human assistance, and more external conditions. In order to solve the problems above, we propose a new method for moving object attitude recognition based on CNN. The method has a simple network structure, and can complete the tasks of classification and recognition of attitude angle through one detection. The advantages of our method such as light, high precision and fastness ensure that our method is excellent for intelligent application of mobile terminals.