Abstract:Because of the extensive flexibility and accuracy, visual navigation technology has been widely used in the field of agriculture intelligent navigation, and many effective machine vision navigation application cases were developed. But under the condition of heavy fog, visual navigation precision is greatly decreased and the processing time in the front image is largely increased, which due to unable to obtain clear front image recently. If the front image interference by the fog is bigger, and image enhancement and recovery effect is not obvious, then it will cause navigation function failure, which results in unable to effectively positioning and navigation. And even it cannot work in serious. In order to solve this problem, this paper proposed an agricultural robot visual dehazing method based on image segmentation map. First of all, this paper adopted the front end image blurring vision and regional segmentation, and got the atmospheric scattering function prediction value based on the segmentation map through the image brightness information. Second, the method optimized the atmospheric scattering function estimation value based on the orientation filter, which enhanced the image edge information, and further improved the fog residual problem caused by the large sky background. Finally, the frontend image dehazing experiment was conducted based on the actual agriculture intelligent navigation platform, and the results were compared with traditional dehazing method. The results showed that the method had high precision and realtime performance. The image dehazing integrated indicators were improved by 28.9% and 29.1% respectively of two part of the video, and the time consumption was improved by 34.4% and 53.9% respectively.