Research On Key Technologies Of Obstacle Avoidance For NAO Humanoid Robot With Monocular Vision Vision

With the development of humanoid robots,the demand for core technology such as obstacle avoidance and path planning of humanoid robots has gradually increased.At this time,it is of great significance to study the obstacle avoidance technology of humanoid robots.In this paper,we use the monocular vision system of NAO humanoid robot to identify obstacles,obtain the position of obstacles,then map the indoor environment to grid map,use grid map and use ant colony algorithm to avoid obstacle path planning,and finally complete the monocular.Obstacle avoidance exercise experiment.The main work of this paper is as follows:Basic experimental system design:Firstly,according to the use environment and scene of the humanoid robot,the rules and accuracy requirements of obstacle avoidance are determined.Then,Zhang Dingyou calibration method is used to complete the calibration and coordinate conversion of the camera's internal and external parameters,and at the same time,the deep information acquisition is completed..Research on obstacle extraction method:Firstly,the robot walking environment is analyzed,and the NAO robot camera is used to construct the data set of obstacles for the indoor environment,and the rotation is selected according to the characteristics of different obstacle sizes and less data in the data set.Noise,cropping and Gaussian blurring are used to enhance the dataset.Then Frater R-cnn+Grabcut is used to form a two-segment obstacle recognition strategy to separate the obstacle region from the background region.It is difficult to identiy small objects for convolutional neural networks.The problem is to optimize the Faster R-cnn using the feature pyramid structure.Finally,the image is post-processed using a binarization and expansion algorithm,and the image is converted into a binarized image for generating a raster map.Research on the monocular obstacle avoidance path planning method of humanoid robot:Firstly,computer vision is used to transform the obtained obstacle coordinates into a grid map.Secondly,the ant colony algorithm is used to form the NAO humanoid robot obstacle avoidance walking path,and the ant's path information is used to complete the acceleration fitting of the walking path,and the path optimization is completed.NAO Robotic Obstacle Avoidance Experiment:Based on the above research results,an experiment was developed,and the NAO robot was used to complete the indoor obstacle avoidance experiment of the humanoid robot,so that the robot successfully bypassed the obstacle to reach the destination.