Universitas Dian Nuswantoro conducted research on the application of Histogram Equalization to improve the visibility of traffic signs in low-lighting conditions. The activity used the ResNet50 and VGG19 deep learning architectures and evaluated the impact of Histogram Equalization on their performance in recognizing traffic signs. Traffic signs are described as important road equipment facilities because they visually inform road users about regulations and instructions.
The research addressed the development and application of automated Traffic Sign Recognition systems for Advanced Driver Assistance Systems. The experimental work compared the reported test accuracy of ResNet50 and VGG19 when Histogram Equalization was applied. The activity therefore focused on image visibility and recognition performance under low illumination, with the stated purpose of evaluating the effect of the image-processing method on traffic sign recognition.
The experimental results showed that Histogram Equalization maintained the ResNet50 test accuracy at around 92%. For VGG19, Histogram Equalization improved the test accuracy from 89.62% to 92.54%. The findings report different effects for the two architectures while providing the stated test-accuracy results. The source further explains that these findings are relevant to Traffic Sign Recognition systems, which are an important component of Advanced Driver Assistance Systems.
By improving traffic sign recognition accuracy under low illumination, Histogram Equalization contributes to improved road safety through more reliable and responsive Advanced Driver Assistance Systems. These systems rely on Traffic Sign Recognition to provide critical information to drivers. Through this research activity, UDINUS addresses low-light traffic sign recognition and its relevance to road safety and the development of sustainable communities.