In pavement structures, the occurrence and propagation of cracks are critical elements that affect road performance, health, and pose a potential threat to the safety of vehicles. The pavement condition information usually collected by manual inspection is laborious, time-consuming, inspector dependent, and easily vulnerable to the perspicacity of the inspector. Therefore, this paper aims to present a computer vision-based crack detection system for pavement structures. In the proposed work, a Convolutional Neural Network (CNN) is used for feature extraction and an AdaBoost classifier is used for classification. The combined architecture is named the Convolutional AdaBoost architecture. A comparative study was conducted and the performance of various classifiers, such as Random Forest (RF), Logistic Regression (LR), Multi-Layer Perceptron (MLP), Support Vector Machine (SVM), Decision Tree (DT), Ada-Boost, and Softmax, was evaluated using different evaluation metrics. The dataset was created based on various pavement structures in the United Arab Emirates and consists of 5600 crack and non-crack images. From the experimental results, it is concluded that the combined CNN features and Ada-Boost classifier yield the best performance with an accuracy of 98%. The performance of other classifiers with CNN features is also comparable with each other. Overall, the integrated Convolutional AdaBoost improves pavement crack detection performance and is more accurate than using CNN alone.