This paper is published in Volume-10, Issue-2, 2024
Area
Deep Learning
Author
Tejaskumar B. Sheth, Dr. Milind S.Shah
Org/Univ
Gujarat Technological University, Gandhinagar, Gujarat, India
Pub. Date
16 April, 2024
Paper ID
V10I2-1150
Publisher
Keywords
Deep Learning, Convolution Neural Networks, Dense CNN, Hyperspectral Image Classification, Forensic Science, Blood Detection.

Citationsacebook

IEEE
Tejaskumar B. Sheth, Dr. Milind S.Shah. 3D Dense CNN for Hyperspectral Imaging-Based Bloodstain Classification, International Journal of Advance Research, Ideas and Innovations in Technology, www.IJARIIT.com.

APA
Tejaskumar B. Sheth, Dr. Milind S.Shah (2024). 3D Dense CNN for Hyperspectral Imaging-Based Bloodstain Classification. International Journal of Advance Research, Ideas and Innovations in Technology, 10(2) www.IJARIIT.com.

MLA
Tejaskumar B. Sheth, Dr. Milind S.Shah. "3D Dense CNN for Hyperspectral Imaging-Based Bloodstain Classification." International Journal of Advance Research, Ideas and Innovations in Technology 10.2 (2024). www.IJARIIT.com.

Abstract

Blood is a crucial piece of evidence in forensic science for reconstructing and solving crimes. Although numerous chemical procedures are utilized to recognize the blood at a crime scene, these various chemical-based methods might affect DNA analysis. One potential application of bloodstain detection and classification using hyperspectral imaging (HSI) is in forensic science for crime scene investigation. In this paper, we developed a deep learning classifier 2D CNN, 3D CNN and Dense for blood stain detection in the field of forensic science. We conduct experiments using a publicly available Hyperspectral-based Bloodstain dataset for experimental and validation purposes. This dataset contains a variety of chemicals, including blood and blood-like compounds such as ketchup, artificial blood, beetroot juice, poster paint, tomato concentrate, acrylic paint, and questionable blood. With the initial training/testing ratio set to 90/10 of the data samples, we compare the results with state-of-the-art three different CNN architecture with PCA, as preprocessing techniques. The result demonstrates that the 3D Dense CNN can offer improved classification accuracies, smoother classification maps, and more discriminable features for hyperspectral image classification.