Long-term traffic signal recognition and identification are critical in comprehending automated truck technology (TSDR). Because this task is so important, a lot of research is currently done and many intriguing approaches are currently proposed in the present literature. Nevertheless, the majority of these techniques have only been tried with clean, problem-free datasets. The speed degradation brought on by different CCs those obscure real-world traffic sign images has not been considered. In this work, we study the TSDR problem for CCs and focus on the accompanying productivity decrease. Because of this, we recommend previous augmentation utilizing a TSDR architecture based on a Convolutional Semantic Network (CNN). Their modular solution consists of two different CNN styles, sign-discovery and category, an encoder-decoder CNN style named Enhance-Net, and a convolutional neural network (CNN) challenge classifier. We present an innovative training procedure for Enhance-Net that focuses on improving the online traffic indicator regions—rather than the entire image—in challenging photographs that are correctly recognized. Our method's effectiveness was assessed using the CURE-TSD dataset, consisting of traffic video clips shot with various consent consents.