In its operation, pipelines encounter a variety of damages, from improper application and unfavorable environmental conditions, which causes defects like metal loss, corrosion, cracks, among others. Along with the growing use of mobile robotic systems for pipelines inspection, we proposed a stereo camera-based monitoring system that can scan, detect, locate, and measure internal defects, particularly on cracks and leakage. To achieve autonomy, the system utilizes a stereo camera to extract 3D information, while a deep learning algorithm, namely Convolutional Neural Network (CNN), is used to identify the defect classes. The result demonstrates the generation of 3D point clouds, classification, and defect quantification. This paper aims to cover the device specification, control solution, system performance, as well as current drawbacks and enhancement approaches.