Narrow space and poor light in urban cable drainage pipe shafts and substation basement, etc. lead to inconvenience and potential safety hazards in the acceptance process. In this paper, an inspection robot equipped with Azure Kinect depth camera is designed, which can automatically collect depth image data and color image data of dark environment in underground cable drainage pipe shafts. Meanwhile, a data processing software is developed. A 3D virtual model of underground cable drainage pipe shafts is constructed through point cloud generation, splicing and fusion processing of original data after which is de-noised, sampled and smoothed to obtain a more accurate data model. Through measuring the relevant structural parameters of the 3D model and comparing with the engineering data, it is managed to determine the acceptance of the results, which helps inspect the quality problems of engineering projects and complete the acceptance work.