Machine Learning Algorithms for Categorization Of Agricultural Dust Emissions Using Image Processing of Wheat Combine Harvester

India is the second largest wheat producer in the world after Russia. Wheat harvesting in the country was traditionally done using a sickle, a hand tool. However, in the last two decades, combined harvesters have been extensively used. The rapid development of mech­anization has resulted in the production of dust and straw particles during the harvesting operation of wheat. These particles have severe health hazards for the machine operator. Exposure to various types of particulate matter has a variety of effects on human health. Such an effect can be minimized if the concentration of the generated particle is maintained within a permissible limit. Hence, the present study has been conducted to evaluate and categorize dust and straw particles in the workspace of a combine harvester operator during wheat harvesting. An image-processing technique was used to study a field data sample collected on sticky paper. It describes a novel method of collecting dust and straw particles while harvesting wheat. Few studies have been conducted in developing countries to analyse the characteristics of dust and wheat straw exposure of combined harvester operators. The number of dust and straw particles deposited per square millimetre was 9–12, with sizes ranging from 10 to 1400 mm. The extracted data were divided into three groups, viz. thoracic, inhalable and straw and modelled using machine learning algorithms, including support vector machine (SVM) and k-nearest neighbor. With an accuracy of 96%, SVM outperformed the other methods for categorising dust and straw particles, whereas linear discriminant analysis performed poorly with an accuracy of 88%.