Environment Perception And Path Planning Of Unmanned Agricultural Machinery In Complex Environment

Environment perception capability is the basis for unmanned agricultural machin es to realize automatic driving.In order to make the unmanned agricultural machine p erceive the environmental information in time and avoid safety accidents in the produ ction process of complex environment,this paper proposes the environment perceptio n algorithm and path planning algorithm based on multi-sensor fusion for the problem of automatic driving when the unmanned agricultural machine performs special tasks in the low visibility and complex environment.The algorithm in this paper can make t he unmanned farming machine can avoid pedestrians in time and ensure the safety of operation in the case of low visibility or obscured objects.First,for the different characteristics of visible and infrared images,this paper pr oposes a proven image pre-processing method.The image after graying,denoising an d sharpening can meet the requirements of subsequent algorithms,enhance the feature extraction ability of subsequent image processing,and exclude the interference infor mation.Secondly,this paper proposes an infrared and visible image fusion algorithm co mbining Poisson image editing and conditional generative adversarial networks.First,the infrared image and its corresponding significant region of the infrared image are u sed to train the conditional generative adversarial network;then,the infrared image is fed into the trained network to obtain the significant region mask;after its morphologi cal optimization,the Poisson image-editing-based image fusion is performed;finally,t he fusion result is processed for contrast enhancement.The algorithm can achieve fast image fusion to meet the needs of unmanned agricultural aircraft to sense the environ ment in real time,and the algorithm can highlight important information such as pede strians and animals in infrared images while retaining visible image detail information,and performs well in objective indicators such as standard deviation and information entropy.Finally,to address the problems of slow convergence speed of ant colony algorit hm,easy to fall into local optimum and many path folds in the process of unmanned f arming machine performing special tasks,this paper proposes an improved ant colony fusion dynamic window algorithm to solve the path planning problem of unmanned f arming machine in dynamic environment.In order to avoid blind search in the early st age of ant colony algorithm,the algorithm designs an adaptive parameter and combine s elite ant strategy and adaptive pheromone concentration volatility factor to avoid fall ing into local optimum and accelerate the convergence speed of the algorithm,improv es state transfer rules to reduce path inflection points by constructing dynamic adjust ment factors,and uses the key points in the optimal path generated by the improved a nt colony algorithm as sub-target points of the dynamic window algorithm to Perform dynamic obstacle avoidance.The simulation results show that the global optimal path can be planned while achieving local dynamic obstacle avoidance to ensure the safety of unmanned agricultural machine operation.