Robot Target Localization and Visual Navigation Based on Neural Network

Deep neural network has been widely used in image analysis, speech recognition, target detection, semantic segmentation, face automatic driving, and other fields due to its excellent algorithm performance. In this research, the is mainly simulate robot localization visual navigation. Firstly, data set based on ICP constructed, then, navigation positioning of are simulated by using network. The results show that combination direct method can effectively solve problem pose loss expand indoor range collection. artificial effective better robustness stability than frame matching. From perspective vision, a single RGB feasible, processing time relatively short, which meet requirements real higher practical value general scenes.