Counting Method of Wheatear in Field Based on Machine Vision Technology

Fan Mengyang;Ma Qin;Liu Junming;Wang Qing;Wang Yue;Duan Xiongchun;College of Information and Electrical Engineering,China Agricultural University;Key Laboratory of Agricultural Information Acquisition Technology,Ministry of Agriculture,China Agricultural University;

Wheatear;;Machine vision;;Image feature;;Support vector machine;;Skeleton

Wheat is a main crop in China and the timely and accuracy estimation of wheat yield is significant. The number of wheater in certain area is an important element in wheat yield estimation. The counting method of wheatear based on machine vision technology was studied,which was cheap and suitable for local area. The method was very significant for wheat growth monitoring and yield estimation.Firstly,the counting method for wheatear in field based on machine vision technology was studied by collecting images of wheatear colony with cameras deployed in the field. The analysis method for wheatear image feature,the thinning method for wheat ear outline and wheatear counting method based on skeleton were realized. The low resolution images of wheat plant were collected with cameras deployed in field.Then the color features and texture features of images were extracted. The outline of wheatear was extracted to get binary image of wheatear by using learning method of SVM. The database of wheatear feature was constructed at the same time and wheatear skeletons were generated by thinning the wheatear binary image. Finally,the number of wheatears was calculated by calculating the number of skeletons and skeleton intersection points. The method was tested in Zhaohe Demonstration Area,Fangcheng County,in May of 2014 and 2015. As a result,it took averagely only 1. 7 s to calculate the number of wheatears and the accuracy was 93. 1%,which means the wheatear counting method presented meets the requirement of both speed and accuracy,and it can provide reliable data for wheat yield estimation.