Underwater hyperspectral imaging technology has potential to differentiate and monitor scallop populations

Abstract

Accurate and low-impact monitoring of scallop abundance is critical for stock assessment, especially in sensitive habitats. The possibility of using low-impact hyperspectral imaging (HSI) for differentiating scallop species in the marine environment was investigated. Live saucer (Ylistrum balloti) and mud (Ylistrum pleuronectes) scallops (N =  31) were scanned inside a sea simulator using a visible to near infrared (400–1000 nm) line-scanner HSI camera. Partial least square discriminant analysis (PLS-DA) was trained to distinguish between the species using their spectral signatures. Important wavelengths were identified and new models were developed using these wavelengths to reduce the model complexity and potentially increase the imaging speed when applied under at-sea conditions. The PLS-DA model distinguished between saucer and mud scallops using any area of the left valve that was exposed above the sediments, with 90.73% accuracy when all 462 available wavelengths were used. Using the subset of important wavelengths (N = 13) reduced the classification accuracy to 84%. Overall, our results showed that HSI has potential for detecting, distinguishing and counting commercially important saucer scallops for low-impact monitoring and resource management, and to complement RGB imaging that relies solely on morphological properties.