A Multi-view Recognition Technology Based on the Prediction of Active Robot’s Optimal Viewing Angle

In task of multi-robots exploring unknown environments, the collaborative information between them can be used to improve the accuracy of object recognition. This paper proposes and implements a multi-view active learning model, which uses images obtained from different perspectives to improve the performance of target detection and recognition. Using robot clusters to obtain multi-view information of the target, through the promotion of the target detection network based on deep learning, the algorithm based on multi-view learning comprehensively utilizes the complementarity of multi-view information, combined with the initiative of multiple robots to estimate the best perspective, Constantly adjust the best observation angle of the cluster to improve the accuracy of target detection. The final experiment shows that the recognition rate of the active multi-view recognition technology proposed in this paper is 12% higher than that of a single view.