Title: Efficient Prediction Model for Detection of Black Spot and Downy Mildew Diseases in Rose Flowers


Authors:

Deepika Kanwar

rathoredeepika1998@gmail.com
Department of Computer Science & Engineering, Swami Keshvanand Institute of Technology, Management & Gramothan, Jaipur-302017 (INDIA),

Mukesh Kumar Gupta

ijskit@skit.ac.in
Department of Computer Science & Engineering, Swami Keshvanand Institute of Technology, Management & Gramothan, Jaipur-302017 (INDIA),

Neha

ijskit@skit.ac.in

Pages: 5-9

DOI:

Abstract:

Flower also known as bloom or blossom often believed as natural item to show the love, peace, affection and complement. Among various countries in the world, India has fourth largest production of flowers. It is also found that rose flower are most preferable flower by people to exchange their feeling on different events. However, rose flower plants are suffered by different diseases. The reasons of these diseases are occurrence of bacteria, virus or fungi in these plants. Every farmer wants to get increase in flower production and concentrate more on the detection of disease, if any, in the early phase of plant growth. 

This paper propose an approach based on image processing and machine learning to correctly detect the plants which are affecting by a particular disease. We have used a dataset of digital images that contains a large set of healthy and unhealthy images of rose plants. We have used color threshold algorithm in order to segment leaf from its background. We have used image processing methods to extract a set of features and develop various classification models. In this work, we have used recall, precision and accuracy parameters for performance evaluation. We have performed various experiments by considering different ratios of training and testing samples on original and augmented datasets. From the all experiments, it is found that the precision, recall and accuracy of proposed approach is highest for random forest based prediction model and lowest for SVM based prediction model in detection of flower plant diseases. In summary, we have achieved the 98.57% accuracy for the detection of Black Spot and Downy Mildew diseases in rose flowers.

Keywords: