Research Of Weed Recognition Method Using Color And Shape Features

With the popularity of computer technology and precision agriculture model proposed,it's possible that the application of computer vision and image processing technology are useful for weed identification and control in crops.For herbicide spraying technology,the primary question is to how to capture the information of weeds distribution in field accurately,and determine the weeds situation in the field (location,density or species,etc.) to achieve fine spray.Based on studying the interrelated research both at home and abroad,the weeds in farmland as the research object,this article focus on the extraction and identification methods which use color characteristics of weeds and the combination of color and shape features to achieve weed identification effectively.The image pretreatment for the primitive farmland weed image gathered in the natural illumination conditions is mainly removing the influence of noise.Aiming at the mixed growth characteristic of weed and wheat in seedling stage of wheat field, this paper presents a weed identification method that takes OHTA as color space and I2' as characteristic variant,uses automatic threshold selection method on the basis of genetic algorithms for the feature component I2' to separate weed and wheat preliminarily,and then uses color clustering and morphological filtering to achieve the accurate weed regions.The experiment results show that separating directly in the color space would improve the segmentation effect of color images,and the average correct identification rate reaches 90.47%using the method proposed before. Considering the instance of cotton field,it's discovered that the color of cotton and common weed is close,and it's not easy to recognize weed effectively only using the color characteristic.This paper studies the weed recognition in cotton field combined with color and shape characteristic,and calculates shape characteristic parameter of common weeds,which prepares for the establishment of weed characteristic database. Various weeds of the particular cotton fields were analyzed,calculate four RST invariance shape parameters of weed leaf:Roundness,Elongation,Dispersal and Compactness,and the shape parameters of weed leaf:Area,Perimeter,etc.The characteristic data provided by the database can be used to carry on the combination classified training,which is useful to weed recognition effectively.This paper proposes some methods of feature extraction and recognition for infield weeds,which will provide the feasible plan to locate the weed accurately and viable project for variable herbicide spraying.It has vitally academic significance and practicality to develop automatic killed-weed system based on machine vision,and raise automation level of agricultural production in our country.