A major difficulty in infrastructure management is the identification of defects in road surfaces, therefore compromising the structural integrity of roadways and traveler safety. Conventionally, this problem has been solved with lengthy, costly, ineffectual manual inspection. Automated crack detection systems are absolutely necessary if we are to increase efficiency, save costs, and offer quick road repair actions. This effort attempts to use an automated crack-detecting method to overcome the limits of hand examinations. The proposed approach integrates a two-stage Convolutional Neural Network (CNN) for feature extraction with the Extreme Learning Machine (ELM) algorithm for effective classification, therefore utilizing developments in deep learning. The system is evaluated extensively using a custom dataset as well as the SDNET2018 data to determine its performance. The results show that the proposed model achieves higher accuracy in spotting road fractures than present techniques, so highlighting its ability to improve maintenance practices and hence support road safety via quick intervention. This approach offers a reasonable and cheap solution for a major infrastructure problem.