Southern University College

Facial analysis in the thermal infrared spectrum enables low-light operation, medical diagnostics, and privacy-preserving biometrics, but scarce annotated datasets limit progress. We introduce TFD68: 28,496 thermal images of 137 individuals, each paired with a visual image matched for head poses, occlusions, and expression. Every thermal image includes 68 facial landmarks and per-pixel thermal maps. The dataset contains systematic yaw and pitch variations, accessory occlusions, and a wide range of expressions to support pose-invariant, occlusion-resilient evaluation. We demonstrate TFD68’s utility on thermal facial landmark detection and facial expression recognition, and show that it enables cross-modal and physiological analyses. To our knowledge, TFD68 is the first thermal facial dataset combining detailed landmark annotations, visual-thermal pairs, and temperature maps, establishing a foundation for advanced thermal facial research.

The primary objective of this study was to assess whether biomass from empty fruit bunch (EFB), Rice Husk (RH), and Rice Straw (RS) can be effectively used for gasification. This study conducted a Computational Fluid Dynamics (CFD)simulation using ANSYS FLUENT and compare the results with published experimental data. First, Thermogravimetric and differential thermal analyses (TG-DTA)were performed to verify the usability of the feedstocks in gasification. The results show that EFB has a higher Carbon content and Higher Heating Value (HHV) than RH and RS, indicating a higher gasification potential. CFD simulations supported these findings, showing that EFB outperforms RH and RS in gasification potential by approximately 40% and 35%,respectively. The results demonstrated that EFB, with its higher efficiency, can offer significant environmental benefits compared to RH and RS

Abstract The bipolar pulse current can effectively mimic the external time-varying stimulus of neurons, and its effect of neuronal dynamics has rarely been reported. To this end, this paper reports the effects of bipolar pulses on a two-dimensional single inertial neuron model, showcasing the chaotic dynamics of hidden attractors and coexisting symmetric attractors, which is of significant importance for understanding the complex behaviors of neuron dynamics under time-varying external stimuli and its application. Firstly, the mathematical model of the single intertial neuron model with forced bipolar pulse is presented, and then the equilibrium states behaving as unstable saddle point (USP), stable node-focus (SNF), and stable node point (SNP) are analyzed. Additionally, by using multiple dynamical methods including bifurcation plots, basins of attraction, and phase plots, complex dynamics of interesting bifurcation behaviors and coexisting attractors are revealed, which are induced by the forced bipolar pulse current as well as initial values, both. In addition, such effets are well valideted via a simple multiplerless electronic neuron circuit. The implementation circuit of presented model is constructed on the analog level and executed using PSIM circuit platform. The measurement results verified the double-scroll chaotic attractors and the coexisting period/chaos behaviors. Finally, the chaotic sequences of the model are applied to color image encryption for the benefit of requirements on modern security field. The encryption effectiveness is demonstrated through various evaluation indexes, including histogram analysis, information entropy, correlation coefficient, plaintext sensitivity, and resistance to noise attacks.

A six-port reflectometry (SPR) system was developed to predict the dielectric properties of both tumor and normal breast tissue, intended for medical diagnostic applications. Ensuring precise measurements, the SPR underwent calibration using a well-established four-step procedure, which will be briefly outlined. Afterward, the investigated coaxial probe was connected to the SPR through the calibrated measurement port. Subsequently, the exposed end of the probe aperture was immersed into synthetic samples representing both healthy and cancerous breast tissue to assess the dielectric constant, εrʹ and loss factor, εrʺ across frequencies ranging from 1.5 GHz to 3.3 GHz. The dielectric constant, εrʹ and loss factor, εrʺ were derived from the measured reflection coefficient using a closed-form equation associated with the coaxial probe. An examination was undertaken to compare the performance of a commercially available vector network analyzer (VNA) outfitted with a Keysight 85070E dielectric probe against an SPR-probe system. The comparison was based on analyzing the reflection coefficient magnitude, phase shift, dielectric constant, and loss factor of synthetic breast tissue samples. The study revealed maximum absolute errors of 0.01, 1.07°, 1.12, and 0.75 for the measured reflection coefficient magnitude, phase shift, dielectric constant, and loss factor, respectively. The calibrated reflection coefficient and predicted relative permittivity, εr can be effectively utilized to distinguish between normal (εrʹ < 50) and tumor (εrʹ > 50) breast tissue.

Abstract: Brain computer interface (BCI) system empowers command over external device by retrieving brain waves and interpreting them into machine instructions. The system utilizes electroencephalogram (EEG) for receiving, processing and classifying signals to control by means of brain generated signals. The paper focused on mental task designs for BCI by acquiring the signals generated by mental activity using EEG comb electrodes, placed over three-dimensional (3D) printed headset. The experiment involved the blinking of left and right eyes for the forward and backward movements of the prototype wheelchair. The experimental measurement was performed using a Cyton board where the information was transmitted through Bluetooth which were later processed and translated to the wheelchair to perform activities. The system has successfully achieved the real time control of an assistive device by using signals from the brain.

Like most plant diseases, tomato leaf disease has physical symptoms. The currently accepted technique is for a trained plant pathologist to identify the condition through visual inspection of affected plant leaves and stems. But due to the manual process, the disease identification time and accuracy have always been questionable facts, making the problem a great application area for Computer-aided diagnostic techniques. Due to the durability and cutting-edge performance, the convolutional neural network (CNN) has been proposed in this study for the real-time classification of tomato leaf disease. The picture data used in this study for tomato leaves came from Plant Village databases. The models employed in this study were trained and tested using 16011 images from original and augmented datasets. The results are then applied to create a real-time system for mobile applications. The initial adjustment with the CNN architecture is proposed to get higher accuracy. A step-by-step systematic approach to parameter tuning is also proposed to improve the system’s performance. The proposed methods show maximum prediction accuracies of 98.00%, and 99.04% are achieved.

This study is focused on improving the behaviour of the "armoured vehicle" in terms of handling responses during firing by enhancing the performance of yaw disturbance rejection control (YDRC). A YDRC is designed to overcome external disturbance using deep learning-based Neuro-PI controller to optimise the variables of the neural network. Moreover, cost-effective approaches are required to evaluate the capability of the controller to enhance the lateral dynamic response of the armoured vehicle. Thus, hardware-in-the-loop (HIL) simulation testing has been adopted in this study to analyse the response of the YDRC. The HIL simulation testing was performed using Cronos Compact data acquisition box developed by integrated measurement and control and integrated with Matlab Simulink. The percentage of error between HIL and software-in-the-loop (SIL) simulation testing using deep learning-based based neuro PI of YDRC is less than 7% for overall simulation testing.

Abstract In this fast-changing era of globalization, the speed of communication, the civilization of material comfort and the advent of artificial intelligence have improved human lives tremendously. However, the distance between people seems to be even further; the hatred among races has increased, and the fighting among nations has never ended.