Outdoor scene understanding of mobile robot via multi-sensor information fusion

The present research on the multi-sensor information fusion technology of mobile robots aims to better understand the outdoor scene and improve the robot's perception of the environment. Firstly, a conversion algorithm is proposed based on two point-cloud-to-image algorithms, including the point cloud plane fitting and point cloud projection transformation. Moreover, the elevation map is constructed to describe the terrain characteristics of the scene based on the three-dimensional laser ranging data. Meanwhile, the conditional random field model is used to obtain landform characteristics from visual information. Besides, the projection transformation and information statistics methods are used to effectively integrate the laser information and the visual information with the grid in the elevation map as the carrier. Ultimately, the convolution neural network is used to realize the three-dimensional scene understanding. It is found that the average recognition rate of the outdoor scene understanding model based on multi-sensor information fusion is as high as 89.36%, and the image segmentation time of the proposed algorithm is not more than 180 ms.The latest research results refer to the use of SSAE in combination with the CRF algorithm. On the whole, the proposed model improves the real-time performance of the mobile robot under the premise of accuracy, and realizes the recognition and analysis ability of complex scenes through the construction of multi-sensor information. This study has important practical significance for promoting the development of the mobile robot autonomous industry.