Hyperspectral Classification of Blood-Like Substances Using Machine Learning Methods Combined with Genetic Algorithms in Transductive and Inductive Scenarios

Abstract

This study is focused on applying genetic algorithms (GAs) to model and band selection in hyperspectral image classification. We use a forensic-inspired data set of seven hyperspectral images with blood and five visually similar substances to test GA-optimised classifiers in two scenarios: when the training and test data come from the same image and when they come from different images, which is a more challenging task due to significant spectral differences. In our experiments, we compare GA with a classic model optimisation through a grid search. Our results show that GA-based model optimisation can reduce the number of bands and create an accurate classifier that outperforms the GS-based reference models, provided that, during model optimisation, it has access to examples similar to test data. We illustrate this with experiments highlighting the importance of a validation set.

Correction of Substrate Spectral Distortion in Hyper-Spectral Imaging by Neural Network for Blood Stain Characterization

“…A stratified k-fold cross-validation [ 32 ] with 10 fold is performed, obtaining a mean classification accuracy of 98.9% with a standard deviation of on the test dataset. This seems consistent with results in [ 8 , 9 ]. The parameters selected for the optimization are: Batch size .…”

“…Both networks were trained exploiting a hyper-parameter Bayesian optimization technique and evaluated with a k-fold cross-validation method with 10 folds. It has been shown how, through the use of simple, lightweight MLP models, the performance of more complex DNN architectures can be achieved, consistent with results in [ 8 , 9 ]. The results of this study showed that this new approach can achieve high performance for all substrates under investigation, being able to obtain a reflectance spectra traceable to the reference one regardless of the surfaces involved in a hypothetical crime scene under investigation.…”

“…Then, the high performance of those methods are compared with the baseline results of SVM. In Palka et al [ 9 ], Genetic Algorithms (GA) are used for pattern and band selection in the classification of hyper-spectral images characterized by both blood stain and non-blood stains. In this trial, they demonstrated that the GA-based model succeeds in decreasing the number of bands and creates an accurate classifier.…”

NCA‐GA‐SVM: A new two‐level feature selection method based on neighborhood component analysis and genetic algorithm in hepatocellular carcinoma fatality prognosis

“…GA is considered one of the most important classes of algorithms based on biological processes. Indeed, GA is widely applied, for instance, in machine learning problems to optimize parameters and select features 31,32,47–52 . It is worth noting that other biology‐inspired algorithms are also used to solve this problem, such as elephant search optimization, artificial bee colony and gray wolf optimizer 53–55 .…”

“…Indeed, GA is widely applied, for instance, in machine learning problems to optimize parameters and select features. 31,32,[47][48][49][50][51][52] It is worth noting that other biology-inspired algorithms are also used to solve this problem, such as elephant search optimization, artificial bee colony and gray wolf optimizer. [53][54][55] Elephant search optimization was applied in the problem of selecting features for microarray data analysis.…”