Machine Learning-Based Detection and Sorting of Multiple Vegetables and Fruits

Vegetable and fruit security plays a crucial role in the Indian economy. In the recent past, it has been noted that vegetables and fruits are affected by different diseases. This leads to the failure of the economy in the agriculture field. The identification of type and grading of vegetables and fruit is onerous due to the heavy production of products. The manual investigation is expensive, laborious, and inconsistent. Thus, an automated machine learning-based algorithm is proposed for the detection of type and quality grading of five different (jalapeno, lemon, sweet potato, cabbage, and tomato) vegetables and four different (apple, avocado, banana, and orange) varieties of fruits. Firstly, images are preprocessed by Gaussian filtering to enhance the quality of the image and removing of noise. Secondly, segmentation of images is done by fuzzy c-means clustering and grab-cut. Then, various features, namely, statistical, color, textural, geometrical, Laws' texture energy, the histogram of gradients, and discrete wavelet transform, are extracted (114) and selected from feature vector by PCA. The detection of vegetable and fruit types is done by color and geometrical features while all other features are considered for grading. Lastly, LR, SRC, ANN, and SVM are used to make decisions for sorting and grading. The performance of the system has been validated by the k (10) fold cross-validation technique. The proposed algorithm achieves 85.49% (LR), 87.63% (SRC), 92.64% (ANN), and 97.63% (SVM) for detection of type. Also, the system achieves 83.91% (LR), 85.00% (SRC), 89.54% (ANN), and 96.59% (SVM) for grading. The proper feature selection shows the enhanced performance of the system. Among four different classifiers, SVM shows more efficient results that are promising and comparable with the literature.

Automated summary

Vegetable and fruit security is crucial for the Indian economy, and the proposed automated machine learning algorithm aims to detect types and quality grading efficiently. By utilizing various features and classifiers, the system achieves promising results for type detection and grading, with SVM showing the most efficient performance.