| Autonomous mobile robot is an integrated system that integrates the functions of environment perception,planning decision-making and multi-level auxiliary driving.The technology of embedded system,sensor,communication,navigation and automatic control is used centrally.It has the functions of autonomous navigation,autonomous positioning,autonomous identification of road obstacles,autonomous alarm,autonomous braking and speed control.In complex road situations,it can realize many functions such as environment awareness,path planning and path tracking.In order to meet the market demand for more accurate,more stable and more economical autonomous mobile robot.And the system complexity is reduced.This paper takes computer vision technology as the core.Halcon 18.05 is selected as the software development environment for computer vision.A road recognition detection algorithm which can satisfy autonomous mobile robot driving on structured road and unstructured road is proposed.The specific work is as follows:(1)The image preprocessing technology of the autonomous mobile robot: the quality of the image is influenced by invalid factors such as impurities in the outdoor road environment.As a result,the collected images can’t be directly used in automatic navigation technology.Therefore,the image is repaired by grayscale,morphological processing and region of interest processing algorithms.At the same time,in view of the frequent occurrence of haze phenomenon in recent years.An improved dark channel prior image defogging algorithm based on bilateral filter is proposed.Clear fogging images are obtained.The preprocessing operation can highlight the effective information in the image.It lays a foundation for the implementation of navigation and obstacle avoidance algorithm of autonomous mobile robot.(2)The structural road recognition technology of the autonomous mobile robots: aim at the lane line regions in structured roads.An adaptive lane line detection algorithm based on dynamic constraints is proposed.The detection range of the algorithm is constrained dynamically.Meanwhile,the gradient calculation method and threshold selection method of the Canny edge detection operator are improved.The accuracy of the detection is improved.The accurate lane line information is extracted as a reference for autonomous mobile robot navigation.(3)The lane line classification tracking and deviation warning technology of the autonomous mobile robot: the lane lines are classified according to color and linear characteristics.The parameters of lane line are predicted by Kalman filter.The optimal estimation of lane line is obtained by analyzing the detection value of lane line.The departure warning system of lane line deviation is proposed based on the optimal value of lane line.The driving state of autonomous mobile robot is evaluated effectively.The task of robot departure warning under dangerous driving condition is realized.(4)The unstructured road recognition technology of the autonomous mobile robots: in view of the unstructured road there are a variety of interference factors.A pyramid LK optical flow method with high real-time performance is proposed.The motion state of the target is judged by the change of pixel position of the image between frames.The obstacle region is raised.In addition,an obstacle avoidance algorithm based on neural network model with strong stability is proposed.The multilayer perceptron algorithm is the core.By learning and training the characteristics of the road region and the non-road region.The road region and the non-road region are precisely divided.At the same time,a self-supervision online correction and deviation warning mechanism is introduced.This mechanism makes the autonomous mobile robot has the ability of autonomous judgment and error correction.The safety of autonomous driving is ensured. |
PDF Full Text Request