Development of a Computer application for detection and classification of Citrus Black Spot (CBS) through machine learning and deep learning models

Citrus black Spot (CBS) disease is a disease of economic importance worldwide. It is on a Red List of the European and Mediterranean Plant Protection Organization (EPPO). The Red list is a list of plant diseases or pests that are considered high risk to a particular region and Phytosanitary Regulatory measures are put in place to ensure that these are pests are not introduced into the country through product importation from countries where these disease and pests have been identified. CBS has recently been identified and confirmed to occur in South Africa.  In order to retain the European market for citrus it is important that this disease is eradicated. Efforts are therefore being undertaken to develop technology for early detection of this disease and this would enable farmers and pathologist to effectively control the disease.

Regina Analytics Pvt Ltd  partnered with North West University Mafikeng Campus to develop a prototype. Regina Analytics Pvt Ltd  provides data science techniques and support, modelling algorithms that are used for machine learning. While North West University is providing plant pathologist who will work as subject expects and researchers. In this process the joint venture managed to develop  Tortoise Application Version 1, that classify/detect the CBS disease using Tensorflow.

The computerized CBS detection technology uses a given image or a video stream, an object detection model can identify the disease by use of a camera on mobile phone. Taking a picture of a diseased citrus fruit in a particular location and captures the symptomatic information and confirms the disease. The technology also gives the GPS location where the disease has been identified using Tensorflow. The models will be used to also classify other diseases in citrus. The model will be on a mobile phone as an application software (app) and will be able to run without internet. The user will take a picture on the phone and run the app, the app will give instant results in real time form a phone running on both android/iOS.

The application runs on android and is still being developed to other platforms like X-code for iPhone. The basic model is below:


The research has been adapted by a PHD student, to further studies in forecasting the spread of CBS diseases using statistical analysis. In this way, the partnership is exploring ways of detecting the disease before it spreads. That is being able to identify the disease and contain it before the affected plants can go to the market.



The research has got interest from Microsoft through the AI for Earth grand for the period of 2020 April to 2021 April where, funding was provided in form of Microsoft Azure Credit.  This made it possible to develop models for the reporting system from cloud to PowerBI.



The project also got interest from the Council for Scientific and Industrial Research (CSIR) as representative from both institution were chosen to represent South Africa in the Innovation Mission Climate Smart Agriculture, which is a collaboration of South Africa and Netherlands agriculture project that was held in the year 2020. This gave exposure and insight for both ventures on other projects in the same field of agriculture being show cased and possibilities of collaborations with these companies and institutions.

This project is ongoing, as there are projection to include unmanned aerial vehicle or uncrewed aerial vehicle. More cameras with capability of Visible-Near Infrared (VIS-NIR) multispectral (MS) imagining system are required. The analysis from the plant pathologist involves working with live CBS samples, which are to be grown and injected in crops and use the results to model algorithms for forecasting. The project is open to interested parties for collaboration.