Fully automated barcode less checkouts in supermarkets utilising machine vision and advanced image recognition algorithms are gaining popularity. A main step in the barcode less automated checkout machines is to recognise objects such that they can be added to the inventory list before the billing and payment by the consumers. For pre-packaged products, the task of image recognition is relatively well addressed in scientific literature and industrial applications. A key challenge arises when the object detection and recognition must be carried out for fresh produce, inside semi-transparent plastic packaging, which is often the case when consumer selects fresh fruits and uses plastic packaging to collect them. To address this, the study proposes a novel solution based on near-infrared hyperspectral imaging and spectral orthogonalization to remove the plastic contribution from the imaged scene. The aim is to use the independent pure spectra of plastics to define a detrimental sub-space on which the imaged hyperspectral scene is orthogonally projected. By doing this, it is demonstrated that, the contribution of the plastic from the imaged scene can be removed and an enhanced insight of the fruits inside the plastic bags is gained. Such technique can facilitate the automated checkout process, improving the recognition performance of fresh products.
|Number of pages||6|
|Publication status||Published - Aug 2023|
- Automated checkout systems
- Fresh-fruit chains