Hybrid approach for fruits quality prediction using image processing and sensors technique

The ability to recognise freshness of fruits and vegetables will be an incredible help for agriculturists to optimise the harvesting stage which abstains from reaping either under-developed or over-developed natural products. Techniques like image processing, near infrared spectroscopy (NIRS), sensors are used to predict the quality of fruits. Most of these mentioned techniques are costly and focused on external features of the fruits. Therefore, this paper presents a hybrid framework to predict the quality of fruits by combining image processing techniques and sensor setup. The proposed hybrid framework improves prediction accuracy rate by considering external as well as internal features of the fruit. Applying image preprocessing, 36 features are extracted from all banana samples. Among them, the most five influential features selected with the help of correlation matrix and linear regression. Simultaneously, with the help of multi-sensor setup, four influential gaseous features are extracted. K-means unsupervised learning algorithm applied on effective features of banana samples and their quality classified into three categories, such as Eatable, MaybeEatable, NotEatable. The implemented framework could classify banana fruit samples with approximately 95% accuracy.