The 2D image obtained by traditional infrared thermal imager lacks target position and the data is chaotic in the detection of thermal defects of building wall. In view of this, a method by fusion of image information and point-cloud model is presented in this paper. This method obtains a 3D building model with target information through infrared image processing, histogram area calculating, edge contour extracting and model fusing. The mosaic error of the obtained model is also analyzed using the identification points. Research and experimental results demonstrate the feasibility of the present method and the high precision of the splicing model. The position coordination and area of thermal defects can be acquired directly by the integrated 3D building model to guide the repair work. The research finding is of guiding significance and application value for integrated information management and intelligent building detection.