![• K-Means clustering algorithm extent to partition
of q images into p clusters and each images
belongs to the corresponding clusters with the
centroid , mean intensity and area [9]. This
algorithm introduces a p different clusters.The
good number of clusters p leading to give the
separation (distance) is not known as a prior and
it should be computed from the data set. The
important of K-Means clustering is to reduce the
total cluster variance or the square function :](https://image.slidesharecdn.com/presentation1-220805143421-bc3a38a9/85/Presentation1-pptx-8-320.jpg)
Presentation1.pptx
- 2. III. PREPROCESSING STAGES • It is most common practice to have the preprocessing of brinjal leaf images before it has been classified and extracted. The processing stages are (i) Reading the images. (ii) Image pre-processing includes enhancing the images and segment the images (iii) Feature extraction and classification.
- 3. • As a first step, RGB images of leaf samples were taken . The procedure for pre processing stages are given below: 1) Reading the RGB image. 2) Histogram equalization. 3) Resizing the image. 4) Create the color transformation structure. 5) K-means clustering Algorithm. 6) Applying masking in an image. 7) Convert the affected leaf (clusters) from RGB to HSI. 8) Create a spatial Gray Level Dependence Matrices (SGDM’s) for S and H. 9) Calculating the features by calling the gray-level cooccurrence matrix (GLCM) parameters. 10) Recognition using Support Vector Machine(SVM).
- 4. Histogram equalization • Different Images of healthy and disease affected leaves are collected from the dataset Histogram equalization is performed on the image in order to increase its quality prior to clustering process as shown in figure 3.Resizing of the image performed to map the images with equal size
- 5. K-means Algorithm • Computation of disease severity requires computation of diseased area of the leaf. This leads to the segmentation of diseased portion of the leaf from the healthy portion. Segmentation partitions the image in to meaningful parts for better analysis and understanding of the image. Thresholding and region growing are two basic approaches of segmentation. Thresholding is simplest where gray image can be partitioned based on threshold value. The image is converted into binary image based on whether the image pixels fall below or above
- 6. • the threshold value. This approach does not provide effective segmentation of the image and hence limits the classification of the various objects of the image if the image contains multiple regions or parts or color. Further it reduces proper detection of the required area. The region growing approach is advantageous in such situations. K- means clustering is one of the popular algorithm in this approach.the flow chart for K means clustering
- 7. Flow Chart for K Means Clustering
- 8. • K-Means clustering algorithm extent to partition of q images into p clusters and each images belongs to the corresponding clusters with the centroid , mean intensity and area [9]. This algorithm introduces a p different clusters.The good number of clusters p leading to give the separation (distance) is not known as a prior and it should be computed from the data set. The important of K-Means clustering is to reduce the total cluster variance or the square function :
- 9. • where,j=objective function, p=number of clusters, q=number of cases, a=case q, c=centroid for cluster p, • 1) Algorithm: The procedure for K-Means Clustering Algorithm is given below: 1) Classify the images into p number of groups where p should be known. 2) Mark p points at randomly in cluster centroid. 3) Mapping objects to their closest cluster centroid. 4) Calculate the mean, centroid or perimeter of all images in each cluster. 5) Repeat steps 2, 3 and 4 until the equal points are mapped to each cluster.
- 10. Result k means clustering Cluster 1 Cluster 2 Cluster 3
- 12. Final output