Algorithm of Robot Visual Perception and Environment Interaction Based on Deep Learning

With the rapid development of science and technology, the visual perception and environmental interaction of robot technology in complex and changeable environment has become a research hotspot. In this paper, an algorithm of robot visual perception and environment interaction based on deep learning is proposed, aiming at solving the limitations of traditional methods in dealing with complex and dynamic scenes. The algorithm uses pre-trained ResNet as feature extractor, combined with YOLO7 target detection algorithm, and realizes fast and accurate detection and recognition of objects in the image. The detected targets are classified by softmax function, and the scene is deeply understood by using conditional random field (CRF) technology, which improves the robot’s overall ability to grasp the environment. In the aspect of path planning, the algorithm uses reinforcement learning technology and continuously optimizes the decision-making strategy through Q-learning method, thus realizing the autonomous navigation and obstacle avoidance of the robot in the dynamic environment. Experimental results show that this algorithm has high accuracy and robustness in complex scenes such as multiple similar objects recognition, illumination change recognition and partial occlusion recognition, and the path planning accuracy is better than Dijkstra algorithm, and the overall performance is better than other deep learning algorithms (Faster R-CNN, SSD, YOLOv5). This study not only verifies the effectiveness of the robot visual perception and environment interaction algorithm based on deep learning, but also provides new ideas and methods for the future development of robot technology.