Source-linked AI summary
Adapted Approach for Fruit Disease Identification using Images
Shiv Ram Dubey, Anand Singh Jalal
TL;DR
Fruit diseases threaten yield and quality, while image-based defect detection is difficult amid variation in skin color, defect types, and stem or calyx presence. This paper segments fruit images with K-Means, extracts descriptors, and classifies disease; it reports support for automatic apple detection and classification.
Problem
Image-based fruit-defect detection remains difficult because skin color and defect types vary and stems or calyxes may be present, while fruit diseases threaten yield and quality.
Source: INTRODUCTION · p. 2Method
The approach segments fruit images with K-Means, extracts color and texture descriptors from segmented regions, and classifies diseases using a multiclass SVM.
Source: INTRODUCTION · pp. 2–3, FRUIT DISEASE IDENTIFICATION · p. 4Results
The authors report that their approach can significantly support automatic detection and classification of apple diseases.
Source: Result Discussion · p. 12Takeaways & Limitations
The study evaluates apple blotch, rot, and scab and targets identification when symptoms appear to support appropriate management.
Source: INTRODUCTION · pp. 2–3
Abstract
from arXiv · showhide
Diseases in fruit cause devastating problem in economic losses and production in agricultural industry worldwide. In this paper, an adaptive approach for the identification of fruit diseases is proposed and experimentally validated. The image processing based proposed approach is composed of the following main steps; in the first step K-Means clustering technique is used for the defect segmentation, in the second step some state of the art features are extracted from the segmented image, and finally images are classified into one of the classes by using a Multi-class Support Vector Machine. We have considered diseases of apple as a test case and evaluated our approach for three types of apple diseases namely apple scab, apple blotch and apple rot. Our experimental results express that the proposed solution can significantly support accurate detection and automatic identification of fruit diseases. The classification accuracy for the proposed solution is achieved up to 93%.
INTRODUCTION
Fruit diseases threaten yield and quality, while image-based defect detection is complicated by natural variation and existing monitoring can be costly or labor-intensive. The paper proposes an image-processing approach to identify fruit diseases, evaluated on three apple diseases.
INTRODUCTION
- Photographic data were difficult to process or quantify mathematically, while image analysis and processing technology offer scope for their analysis.Source: INTRODUCTION · p. 1
- Existing monitoring options include scouting, PCR, and expert visual inspection, which the paper describes as costly, time-consuming, labor-intensive, or requiring detailed sampling and processing.The introduction also reports no commercially available sensor for real-time assessment of tree health.Source: INTRODUCTION · pp. 1–2
- Fruit diseases can reduce yield and quality, and early detection may aid loss reduction and prevent spread; the authors target symptom-onset identification for proper management amid pesticide concerns.The introduction also notes pesticide residues, groundwater contamination, and production cost as concerns.Source: INTRODUCTION · p. 2
- Image-based defect detection remains difficult because fruit skin color varies, defect types vary widely, and stems or calyxes may be present.Source: INTRODUCTION · p. 2
- The proposed approach segments fruit images with K-Means, extracts features from segmented regions, and classifies fruit diseases using a multiclass SVM.The apple test case includes blotch, rot, and scab.Source: INTRODUCTION · pp. 2–3
LITERATURE REVIEW
The review frames fruit-disease identification as image categorization and surveys defect segmentation alongside spectroscopic and imaging approaches for monitoring plant and fruit diseases.
LITERATURE REVIEW
- Fruit-disease identification is presented as an instance of image categorization.Source: LITERATURE REVIEW · p. 3
- Prior fruit-defect segmentation includes simple and globally adaptive thresholding, Bayesian pixel classification, and unsupervised classification.The review notes that unsupervised classification lacks target values to guide learning.Source: LITERATURE REVIEW · p. 3
- Reviewed monitoring methods include fluorescence, multispectral or hyperspectral imaging, infrared and visible-band spectroscopy, and NMR, alongside broader reviews of pathogen-detection techniques.Source: LITERATURE REVIEW · p. 3
- Field-oriented examples include hyperspectral and fluorescence imaging for crop disease detection and airborne hyperspectral mapping of oil-palm trees affected by basal stem rot.One field system compared 550 and 690 nm fluorescence images to detect disease presence.Source: LITERATURE REVIEW · p. 3
- Other cited work applied hyperspectral imaging to citrus canker lesions and NIR spectroscopy to sugarcane resistance ratings.Source: LITERATURE REVIEW · p. 4
- Laser-induced fluorescence distinguished citrus-canker-infected orange leaves from healthy leaves, but not from another disease.Source: LITERATURE REVIEW · p. 4
FRUIT DISEASE IDENTIFICATION
The approach combines K-means segmentation with statistical color and texture descriptors to identify fruit diseases. Its workflow trains a support vector machine on features from infected regions, then classifies new images using features from their segmented regions.
FRUIT DISEASE IDENTIFICATION
- The workflow segments infected regions, extracts their features, trains a support vector machine on stored features, and classifies new images from their segmented features.Precise segmentation is used to prevent features from non-infected regions from dominating.Source: FRUIT DISEASE IDENTIFICATION · p. 4
- K-means partitions fruit-image pixels into four clusters by minimizing squared distances; one or more clusters may contain only infected regions.The experiment uses four clusters, while empirical observations report good segmentation with three or four.Source: FRUIT DISEASE IDENTIFICATION · p. 4, K-Means Image Segmentation Algorithm · p. 5
- The segmentation procedure converts images from RGB to L*a*b* and clusters colors in a*b* space.The paper uses the a* and b* channels to reduce processing time for defect segmentation.Source: K-Means Image Segmentation Algorithm · p. 5
- The method uses Global Color Histogram, Color Coherence Vector, Local Binary Pattern, and Completed Local Binary Pattern features for fruit disease identification.The method is described as exploiting statistical color and texture descriptors in a multi-class scenario.Source: FRUIT DISEASE IDENTIFICATION · p. 4, Feature Extraction · p. 7
- Unlike LBP, CLBP encodes signs and magnitudes of local differences as well as the original center gray level.CLBP combines sign, magnitude, and center-value components.Source: Completed Local Binary Pattern (CLBP) · p. 7
RESULTS AND DISCUSSIONS
The evaluation uses apple images to compare color and texture features across RGB and HSV with a multiclass SVM, and reports support for automatic apple-disease detection and classification.
RESULTS AND DISCUSSIONS
- The apple dataset contains 391 images: 104 blotch, 107 rot, 100 scab, and 80 normal apples.The authors describe variation in apple type and color as making the dataset more realistic.Source: Data Set Preparation · p. 9
- The evaluation analyzes GCH, CCV, LBP, and CLBP features derived from RGB and HSV images, using MSVM as the classifier.Test accuracy is defined as correctly classified images divided by the number of test images, multiplied by 100.Source: Result Discussion · pp. 9, 11, Figure 5 · p. 10, Figure 6 · p. 11
- The evaluation varies the number of training images per class and plots test-image accuracy for the feature comparisons.The experiment uses M training images per class, with the remaining N-4*M images used for testing.Source: Result Discussion · p. 9
- The experiments indicate that the approach can support automatic detection and classification of apple diseases.The supplied results passage gives no numerical accuracy value.Source: Result Discussion · p. 12