Simulated Annealing-Based Hyperspectral Data Optimization for Fish Species Classification: Can the Number of Measured Wavelengths Be Reduced?

Abstract

Relative to standard red/green/blue (RGB) imaging systems, hyperspectral imaging systems offer superior capabilities but tend to be expensive and complex, requiring either a mechanically complex push-broom line scanning method, a tunable filter, or a large set of light emitting diodes (LEDs) to collect images in multiple wavelengths. This paper proposes a new methodology to support the design of a hypothesized system that uses three imaging modes—fluorescence, visible/near-infrared (VNIR) reflectance, and shortwave infrared (SWIR) reflectance—to capture narrow-band spectral data at only three to seven narrow wavelengths. Simulated annealing is applied to identify the optimal wavelengths for sparse spectral measurement with a cost function based on the accuracy provided by a weighted k-nearest neighbors (WKNN) classifier, a common and relatively robust machine learning classifier. Two separate classification approaches are presented, the first using a multi-layer perceptron (MLP) artificial neural network trained on sparse data from the three individual spectra and the second using a fusion of the data from all three spectra. The results are compared with those from four alternative classifiers based on common machine learning algorithms. To validate the proposed methodology, reflectance and fluorescence spectra in these three spectroscopic modes were collected from fish fillets and used to classify the fillets by species. Accuracies determined from the two classification approaches are compared with benchmark values derived by training the classifiers with the full resolution spectral data. The results of the single-layer classification study show accuracies ranging from ~68% for SWIR reflectance to ~90% for fluorescence with just seven wavelengths. The results of the fusion classification study show accuracies of about 95% with seven wavelengths and more than 90% even with just three wavelengths. Reducing the number of required wavelengths facilitates the creation of rapid and cost-effective spectral imaging systems that can be used for widespread analysis in food monitoring/food fraud, agricultural, and biomedical applications.

A Novel Machine-Learning Framework Based on a Hierarchy of Dispute Models for the Identification of Fish Species Using Multi-Mode Spectroscopy

“…Simpler ML algorithms such as LDA that were used in their study have proven their efficacy on datasets with comparatively lower numbers of classes, but our dispute model framework provides a solution for the confusion that ensues when dealing with many classes/species, as is the case when dealing with many different fish within a market. Chen and Chauvin [14,15] both provided solutions with the ability to be applied to handheld systems, offering rapid and portable measurements as was carried out in our study, but again, their sample size was comparatively small, and does not classify as wide a range of species as seen within this research. De Graeve et al [10] analyzed many samples and species, but this more comprehensive dataset does not have the means to be applied to a portable system that does not harm the samples in the process.…”

“…This algorithm is stated to be suitable for small mobile devices, and would allow for the implementation of edge computing for small NIR spectrometers. Chauvin et al also utilized the VNIR region to obtain reflectance spectra, as well as to collect fluorescence spectra with UV illumination [15]. These data were analyzed from fourteen fish representing six species.…”

“…The spectral reconstruction algorithm with only three spectral bands garnered 99.98% and 99.94% with the linear discriminant and subspace discriminant classifiers, respectively. The same authors continued their work and expanded the dataset to 133 samples comprised of 25 different species [15]. Fluor, SWIR reflectance, and VNIR reflectance were used to capture narrow-band spectral data at three to seven wavelengths.…”

Unifying Variable Importance Scores from Different Machine Learning Models Using Simulated Annealing

“…The positions of the changed entries were randomly chosen. The process continues by replacing w by w' whenever the performance of w' is better than w [12,13]. In this paper, the performance of the solution is Z, the minimum correlation between w and Ri.…”

Perspective Chapter: Hyperspectral Imaging for the Analysis of Seafood

“…Additionally, the use of imaging technologies able to identify different fish species is attractive both for the consumer and for the industry, since they can help to mitigate fraud in fish mislabeling [61]. In a research study performed by Chauvin et al, the authors evaluated the potential of the spectral information of fillets from different species in order to correctly classify them [62,63]. A total of 22 fish species were recorded using diffuse reflectance illumination (VNIR and SWIR spectral ranges) and fluorescence excitation (VIS).…”