The major issues of fruit quality loss post-harvest and export rejection persist in developing countries due to the non-availability of a cost-effective and non-destructive fruit quality testing system. The current fruit quality testing methods, such as manual grading analysis and lab analysis, are of poor accuracy and lack scalability. On the other hand, the other state-of-the-art methods using X-ray or CT scanning technology are exclusive and expensive for mid-scale fruit export companies. But as a solution to that, a new fruit quality testing system based on a low-cost multi-modal analysis approach with the support of decision-making algorithms will be designed. The proposed scheme integrates optical im-aging and secondary sensing modalities to extract the structural, textural, and quality attributes of the produce non-destructively. The data from the sensors is processed using a lean ML scheme optimised for real-time prediction with an industrial throughput. The proposed scheme was implemented and tested with different images of the produce and was found to be optimal for defect detection and produce grading. The experimental results reveal that data fusion yields a remarkable improvement in detection rates compared to single-sensor solutions. The results verify that the proposed system is capable of ensuring quality evaluation at a lower price compared to that of the conventional inspection systems and is applicable for use in the export-oriented packhouse. The proposed work has made significant contributions in terms of providing a realistic application to the post-harvest quality control and has pointed out the ability to apply the mul-timodality approach using machine learning to mitigate the economic loss in the fruit export chain.

Cost-effective, X-ray, CT Scanning, Multimodality, Algorithms, Single-sensor