Farmland human-shape obstacles identification based on Viola-Jones Algorithm

Abstract: When agricultural unmanned aerial vehicle (UAV) is spraying in the field or hovering on the ridge, there is a possibility that obstacles without moving characteristics represented by humans may collide with agricultural UAV.  In order to quickly identify and detect human-shape obstacles, realize the safe operation of agricultural UAV, this paper proposes a new method based on Viola-Jones Algorithm according to the head-shoulder ratio characteristics of Chinese adults and the basic spraying parameters of agricultural UAV.  The cascade object detector is used to detect people’s upper bodies area of all humanoid obstacles in the view field of the agricultural UAV’s flight direction, and the extracted upper body image is binarized to obtain the relevant projection histogram and calculate the width ratio and distance ratio of head-shoulder, then judge whether it is a human-shape obstacle, and combine the depth and position information to avoid the danger of collision.  The test results show that the method is feasible and effective in recognizing human-shape obstacles on the front or back in the farmland within the sight distance of6 meters. Keywords: agricultural UAV, binocular vision, picture processing, human-shape obstacles, MATLAB DOI:  10.33440/j.ijpaa.20200303.99 Citation: Wang L L, Xiao W W, Qi Y, Gao Q C,Li L, Yan K T, Zhang Y L, Lan Y B.  Farmland human-shape obstacles identification based on Viola-Jones Algorithm.  Int J Precis Agric Aviat, 2020; 3(3): 35–40.