Research On Obstacle Avoidance Of Plant Protection UAV Based On Vision

In recent years,due to the continuous development and improvement of UAV technology,UAV has gained a lot of demands and applications in both civil and military fields.In agricultural plant protection,because of its small size,high efficiency and safety,uniform application of pesticides,low cost and other advantages,plant protection UAV has been widely used in agricultural plant protection,gradually replacing artificial plant protection,occupying a certain market share.When plant protection UAV carries out plant protection task in farmland,the environment in farmland is more complex.The plant protection UAV will encounter obstacles such as trees,houses,electric poles and so on.The obstacles in farmland will affect the flight safety of plant protection UAV,and cause damage to plant protection UAV.It will not only fail the plant protection task,but also cause additional menstruation.Therefore,the plant protection UAV needs to have obstacle avoidance function,to avoid obstacles,so that the plant protection UAV can complete the plant protection task safely and effectively.In this paper,a vision-based obstacle avoidance method for plant protection UAV is studied.Further research is carried out on open source flight control UAV.The obstacle avoidance algorithm is added to plant protection UAV to realize obstacle avoidance function.The main research contents of this paper are as follows:1.Through reviewing a large number of literatures on visual obstacle avoidance of UAVs,the research background and significance of UAV visual obstacle avoidance are expounded,and the research status of UAV visual obstacle avoidance at home and abroad is summarized.2.Knowledge of video and image processing.The plant protection UAV uses the CCD camera on board as a visual sensor to obtain the environmental information in the field of vision.This paper briefly introduces the CCD camera.Because of the internal components of vision sensor and the influence of the external environment of plant protection UAV,the digital image acquired by plant protection UAV not only has poor contrast,but also has noise,which makes the quality of the image decline.Histogram equalization and median filtering are used to improve the contrast and remove the image,respectively.The principle of histogram equalization and median filtering is introduced in detail,which is verified by simulation experiments.The commonly used image segmentation methods are:image segmentation based on threshold,region growing and region splitting and merging,image edge detection.In this paper,the principles of these three image segmentation methods are introduced in detail,and the performance of several edge detection operators to achieve edge detection is compared and analyzed through simulation experiments.3.Research on vertical obstacle avoidance strategy based on feature detection.Obstacle detection is an important part of visual obstacle avoidance for plant protection UAV.In this paper,HOG feature detection algorithm,SIFT feature detection algorithm and SURF feature detection algorithm are studied.Through experiments,the three algorithms are compared and analyzed,and SIFT feature detection algorithm is used for obstacle detection,using millimeter wave radar to measure the distance information of obstacles,the principle of millimeter wave radar ranging is introduced,vertical obstacle avoidance of UAV based on the information of obstacle characteristic points.4.Research on left-right obstacle avoidance strategy based on genetic algorithm.The principle of genetic algorithm is briefly introduced,and the feasibility of the left-right obstacle avoidance strategy based on genetic algorithm is verified by simulation experiments.