An In-Depth Analysis of Different Segmentation Techniques in Automated Fruit Disease Recognition

Md. Tarek Habib


Image segmentation is a significant processing stride in a large number of machine vision applications. Although there exist different techniques for segmentation, it is effortful yet to determine whether one technique performs better than another in a given context. Automated, i.e. machine-vision-based fruit disease recognition is a typical instance of such a context. In this context, image segmentation plays a very important role for extracting features, because estimation of size and location of  area of the object of interest is requisite for ascertaining the presence of disease and extract the accurate values of features. So, region-based performance metrics are required in this regard. In this research work, we perform an in-depth analysis of four prominent segmentation techniques, namely Otsu’s method, k-means clustering, fuzzy c-means clustering and region growing in the index of six region-based metrics to thoroughly evaluate the quality of defective area detection and feature extraction for four diseases of each of the three common local fruits of Bangladesh, namely guava, jackfruit and papaya. k-means clustering segmentation technique is found outperforming all other segmentation techniques in terms of all of the metrics used other than under segmentation measure by attaining an aggregate accuracy of 81.65%. On the contrary, region growing technique shows not only the poorest results in terms of all of the metrics used other than over segmentation measure by attaining an aggregate accuracy of more than 75%.


Fruit Disease, Machine Vision, Image Segmentation, Region-Based Metric, Accuracy, Discriminatory Features



  • There are currently no refbacks.

Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Bulletin of EEI Stats