Plant Health Monitoring Using Digital Image Processing

In this paper a way of disease detection in cashew plants using digital image processing and smartphone is given. Usually it is either difficult or not economically viable for farmers to get proper diagnosis of their plants. So, with the help of digital image processing quick, easy and cost effective diagnosis for disease detection can be done. And use of android application makes it even more versatile.

Description: The method described makes use of leaves for disease detection and identification. Usually the diagnosis is done by observing plants and identifying the disease (done by experts). Brown and yellow spots, or early and late scorch, and fungal, viral and bacterial diseases are some very common plant diseases. Image processing makes use of image segmentation to identify the type of disease a plant has. This is not very first approach in the direction of ‘Plant Health Monitoring’. The first step is image acquisition of plant leaves (cashew leaves in this case), next step is feature extraction from the leaves, then statistical analysis and finally classification of the disease. Figure-1 and Figure-2 are flow charts but the blocks used are all rectangular; whereas a proper flowchart shows workflow or process and makes use of proper starting, input, processing, output and ending blocks encapsulating details.

The colour image captured using any camera is converted to a device-independent colour space transformation is applied. After applying some noise removal filters image segmentation is done using K-means Clustering, converting RGB image into HSI model etc. Algorithm for K-means Clustering used in this paper is J=|xn - µj|2 Where xn is a vector representing the nth data point and µj is the geometric centroid of the data points in Sj. The feature (such as contrast, energy, correlation etc.) then extracted are used for classification. The classifier used in this case is Support Vector Machine (SVM) this is normally used for classification and recognition of patterns. The accuracy of SVM increases with the number of samples used in the training dataset (supervised machine learning).

Detecting diseases in plants is crucial. However, it can be cumbersome because of the large areas the plantations cover. And hiring professionals is a good but costly option. The method proposed in this paper is simple and easy; anyone with a smartphone and computer with proper softwares installed can use digital image processing can monitor large fields/plantations for potential diseases that threaten the produce. This paper talks about cashew plants but in general this method can be used for any other plant, provided the system is trained properly with the dataset of features for the intended plant.