Deep Learning Applications in Thermal IR Image Processing

IR image recognition has been a promising field for the past few years. However, it is difficult to identify facial emotions when it is dark, the lighting is poor, or there are other elements present. Thermal pictures are therefore recommended as a fix for these issues and for a number of additional advantages. Additionally, focusing on important areas of a face rather than the complete face is sufficient to reduce processing while also enhancing accuracy. This study provides brand-new infrared thermal image-based methods for identifying the images. The face’s whole image is first divided into four parts. Then, to create training and testing datasets, we only allowed four active regions (ARs). An approach to machine learning is called active neural network (ANN). We also used a parallelism strategy to speed up processing of training and testing datasets. We have observed a 46% reduction in processing time as a result. Finally, to increase the recognition accomplishes a recognition accuracy of 92.38%. The achieved accuracy validates the stability of our suggested strategy.