Research On Obstacle Recognition And Edge Contour Acquisition Of Grasshopper Bionic Robot

Grasshoppers have the characteristics of small size and concealment,as well as their superb jumping ability,so their bionics research has been uninterrupted.Grasshopper bionic robots can operate in a variety of harsh and complex environments,and can even use their concealed characteristics to penetrate behind enemy lines to complete military reconnaissance targets.Under the circumstances of the above-mentioned complicated operating environment and the execution of special task objectives,the vision system of a grasshopper bionic robot was studied and designed.It is very meaningful to enable it to achieve autonomous obstacle avoidance and path planning.The vision system is an indispensable part when researching the design of bionic robots.By comprehensively using advanced technologies such as image processing,pattern recognition,and artificial intelligence,obstacle detection and moving target tracking can be performed.It has important theoretical value and practical significance.In order to achieve the above goals,this article mainly conducts design research from the following aspects:The hardware structure of the grasshopper bionic robot vision system was designed and constructed,and the hardware was selected and analyzed.When constructing the hardware of the grasshopper bionic robot vision system,considering that the grasshopper bionic robot has a small space,a compact structure,and a compact form,and cannot install too many hardware modules,the hardware system is designed with the most simplified structure.The three YOLO models are studied,and the principle and training methods of YOLOv3 are analyzed.Detection of processed images.In terms of target detection,it is a deep neural network algorithm based on yolov3,which has fast detection speed and high accuracy.Obstacle detection was performed using the traditional YOLOv3 model.Aiming at its low detection speed,the model was optimized.Mainly pruning optimization of the YOLOv3 network model reduces the number of layers of the network and increases the recognition speed;optimizing the yolov3 parameters makes the training speed improve and the training model converges faster.The obstacles detected in the image are segmented and extracted,and the extracted obstacles are image processed.The main purpose is to obtain the outline of the outline of the obstacle,and polygonally approximate the outline of the outline to form a convex closure.It prepares for the subsequent research on the path planning of grasshopper bionic robot.The grasshopper bionic robot is a typical high-end intelligent bionic robot.It not only integrates advanced technologies such as robotics,bionics,and control,but also needs to be able to adapt to very complex operating scenarios.Therefore,there are many technical difficulties in its research and development,among which the typical technical points are the identification and positioning of obstacles.In this paper,the vision system of grasshopper bionic robot is taken as the research object,and the theoretical research and experimental verification around obstacle recognition,image processing and so on.