Abstract
The objective of the proposed human-machine cooperation (HMC) workstation is to both rapidly detect calcium-based fish bones in masses of minced fish floss and visually guide operators in approaching and removing the detected fish bones by hand based on the detection of fingernails or plastic-based gloves. Because vibration is a separation mechanism that can prevent absorption or scattering in thick fish floss for UV fluorescence detection, the design of the HMC workstation included a vibration unit together with an optical box and display screens. The system was tested with commonly used fish (swordfish, salmon, tuna, and cod) representing various cooking conditions (raw meat, steam-cooked meat, and fish floss), their bones, and contaminating materials such as derived from gloves made of various types of plastic (polyvinylchloride, emulsion, and rubber) commonly used in the removal of fish bones. These aspects were each investigated using the spectrum analyzer and the optical box to obtain and analyze the fluorescence spectra and images. The filter was mounted on a charge-coupled device, and its transmission-wavelength window was based on the characteristic band for fish bones observed in the spectra. Gray-level AI algorithm was utilized to generate white marker rectangles. The vibration unit supports two mechanisms of air and downstream separation to improve the imaging screening of fish bones inside the considerable flow of fish floss. Notably, under 310 nm ultraviolet B (UVB) excitation, the fluorescence peaks of the raw fillets, steam-cooked meat, and fish floss were observed at for bands at longer wavelengths (500-600 nm), whereas those of the calcium and plastic materials occurred in shorter wavelength bands (400-500 nm). Perfect accuracy of 100% was achieved with the detection of 20 fish bones in 2 kg of fish floss, and the long test time of around 10-12 min results from the manual removal of these fish bones.