Novel Approach for Automatic Detection and Classification of Multiple Fruits using Machine Learning and Deep Learning Techniques

Abstract

Fruit detection and classification is a computer vision task that involves automatically identifying and categorizing different types of fruits from images or videos. This task is essential in various industries, including agriculture, food processing, and retail, where accurate and efficient fruit identification can streamline operations and improve quality control ans systems have shown promising results in recent years, providing efficient and reliable solutions for automating fruit sorting, inventory management, and quality assessment in various industries, ultimately contributing to increased productivity and reduced manual labor costs. The process typically involves two main steps: detection and classification. In the detection phase, object detection algorithms are employed to locate the presence and positions of fruits within an image. Once the fruits are detected, the classification phase utilizes machine learning or deep learning techniques to identify the specific type of fruit in each detected region. In our research, we present an innovative automatic detection and classification system for multiple fruits, utilizing the power of machine learning (ML) and deep learning (DL) approaches. The system first employs ML algorithms extract the multiple features including statistical, geometrical, textural, HOG to obtain useful information from fruit images, followed by PCA to selected the desired features and also preprocessing the data to enhance its suitability, Support vector machines(svm) are used to classify the fruits and achieved a 99.46 accuracy with compared to other classification algorithms. Secondly, we have incorporated the hybrid model with combination of SaRF and optimised k means clustering segmentation to detect the defected region, after detecting the defect texture features are extracted from SURF with key point descriptors, at last Bi-LSTM+RAM classifier is to classify the fruits and achieves a 99.89 when compared with RBM is 93.6, LSTM is 91.1, RNN is 86.6.