Radon Transform Based Real-Time Weed Classifier Muhammad Inam ul Haq 1 , Abdul Muhamin Naeem 2 , Irshad Ahmad 3 , Muhammad Islam 4 1 Center of IT, Institute of Management Sciences, Peshawar. 2 Department of Computer Sciences, Farabi College Peshawar 3 Department of Computer Sciences, Islamia College Peshawar. 4 Department of Telecom Engineering, FAST-NU Peshawar. {inamix@imsciences.edu.pk, (abdulmuhamin, iakhalil2003)@yahoo.com, m.islam@nu.edu.pk} Abstract A machine vision system to detect and discriminate crop and weed plants in a commercial agricultural environment was developed and tested. Images are acquired in agricultural fields under natural illumination were studied extensively, and a weed classifier based on Radon Transform is developed. This classifier is specifically developed to classify images into broad (having broad leaves) and narrow (having narrow leaves) classes for real-time selective herbicide application. The developed system has been tested on weeds in the lab; the results shows reliable performance and significantly less computational efforts on images of weeds taken under varying field conditions. The analysis of the results shows over 93.5% classification accuracy over a database of 200 sample images with 100 samples from each category of weeds. Keywords--- Ecology, Image Processing, Radon Transform, Real-Time Recognition, Weed detection. 1. Introduction 1.1. Background Increasing farm sustainability and protecting water quality are two major goals of current agricultural research. Total variable costs in 2002 for U.K were within a range of £1,720/ha and £1,870/ha for main crop potatoes, of which herbicides accounted for between 3% and 4% of costs, fungicides accounted for about 8% of variable costs and nematicides accounted for about 14%- 16% of variable costs. United States farmers applied about $16 billion of herbicides in 2005 (The Value of Herbicides in U.S. Crop Production: 2005 Update, CropLife Foundation), representing a significant portion of the variable costs of agricultural production. Although weeds are not evenly distributed on field, weed treatments are mostly applied with the same dose over the entire field, while the herbicides are used more efficiently by adjusting the herbicide dose to specific weed [10]. In order to perform a localized weed treatment, local weed population must be evaluated on the field. Only automatic weed detection can be economically justified [7, 8]. Species based identification of weeds is not necessary for its precise treatment. All weeds with the same treatment can be placed in one class, but this might create classes with high variance in spectral response, because weeds might have very different optical characteristics. A well considered choice of weed categories should be based on treatment and morphological properties of weeds. Classification into narrow and broad leaves seems particularly interesting because these weed groups are treated with different herbicides and have different optical characteristics [11]. The treatment of weeds, i.e. use herbicides, only where it is needed and not on the whole field, will be a great environmental and economical benefit. 1.2. Related Work Much research has investigated strategies to control weeds with less herbicide to reduce production costs and to protect the environment. A verity of visual characteristics that have been used in plant identification can be divided into three categories: Spectral Reflectance, Morphology and Texture. The photosensor-based plant detection systems [2] can detect all the green plants and spray only the plants. A machine-vision guided precision band sprayer for small-plant foliar spraying [3] demonstrated a target deposition efficiency of 2.6 to 3.6 times that of a conventional sprayer, and the non-target deposition was reduced by 72% to 99%. A system that could make use of the spatial distribution information in real-time and apply only the necessary amounts of herbicide to the weed-infested area would be much more efficient and minimize environmental damage. Therefore, a high spatial resolution, real-time weed infestation detection system Computer Graphics, Imaging and Visualisation (CGIV 2007) 0-7695-2928-3/07 $25.00 © 2007