Beetles: A Mobile Application to Detect Crop Disease for Farmers in Rural Area Rahat Yasir, Nova Ahmed EECS Department, North South University Dhaka, Bangladesh rahat.anindo@live.com ; nova.ahmed.2011@gmail.com Abstract— We have presented our work Beetle designed to support farmers in rural area to detect crop disease. The system shows promises in terms of disease detection. We present our technique here. We have started our software development process with user’s requirement analysis. We wanted to complete this cycle by taking Beetles back among the farmers and find out about their opinion on such tool. The farmers received the tool positively and provided us with valuable insights presented in the paper. Keywords— Plant disease; Mobile Phone Application; Image processing; Human Factors; Design; Measurement; Crop Disease and Fertilizer. I. INTRODUCTION Recent proliferation of cell phone has created unique opportunities in developing countries like Bangladesh. On one side the technology has created communication and compute capabilities all around the country. On the other hand, there is a demand for support at rural places where this cell phones are available. We have focused on an application developed for farmers to detect crop disease which is a major problem among farmers. We developed a cell phone based application “Beetle” that detects crop disease from the image captured by a cell phone and detects the disease in real time. We have studied Beetle among a small group of farmers in our evaluation process along with the evaluation of the software performance itself. Disease detection is a major concern among farmers in the rural areas of Bangladesh. They make their decision based on experiences and assumptions. We have focused on image processing based techniques to detect crop disease. Current image processing algorithms require high quality images [1, 2, 5, 6] which is not the case for cell phone based images. The acquired image must be processed within the cell phone to serve farmers in absence of internet connectivity. We have developed an application that is able to detect crop disease within the limited compute capabilities of cell phone using histogram and color information of captured image. Similar approach has been proposed by [6] to detect the shapes of medicinal plants unlike our approach. There has been various cell phone based applications [add] focused to support rural people. Our approach is complementary to other cell phone based applications as our approach can gain advantage of internet connectivity in terms of information sharing. However, Beetle is able to perform independently. Beetle can detect five different crop diseases successfully and they are Narrow brown leaf spot, Bacterial blight, Brows spot, Ufra and Rice Blast. We have conducted user studies before and after our software development process. We wanted to find out about the burning problems for farmers and then we have showed our complete application to the farmers and got their feedback. We focused mainly on the accuracy and performance of Beetle which was well received among the farmers. We got suggestions about improvement of the current application in terms of its interface and features. Beetle shows promises and gives us future direction for a robust application. Our user study provided us with insights about desirable features such as use of native language and familiarity. The rest of the paper is organized as follows, we present our system design in next section, followed by the results, related work and finally, the conclusion. II. SYSTEM DESIGN We present our system design in two phases. We first present the user interface which is carefully designed to be simple keeping in mind the low literacy (or lack of literacy) level among farmers followed by the actual system development process. A. User Interface We have focused on a simple interface so that the users can select their options in minimal number of steps. Beetle has two major parts – one is disease detection and then it suggests for in suggestion for fertilizers. The disease detection process goes through four steps. First, we need to place the leaf of observation on against a white background and take a photo using a single click in the current smart phone. Then we may choose our desired step from the list of tasks named Analyze Crop Disease. Then our algorithm uses its internal mechanism to calculate the disease in real time and finally, displays the result in an output screen. Text to speech option is also available for illiterate farmers who are unable to read the output result of Beetle. The fertilizer selection process goes through the first two steps as before. The third step requires the user to select a portion from the leaf as shown in the figure. The fertilizer suggestion is based on the leaf color as shown in Figure 1. Workshop on Human And Technology (WHAT), 8-10 March 2014, Khulna, Bangladesh 11