Research On The Application Of Real-World 3D Model And Infrared Image Fusion In Bridge Disease Detection

Concrete bridges are usually not plastered,and surface cracks are one of the key focuses during routine inspections.Currently,bridge inspections are primarily conducted through manual close-range visual inspection or photography,especially during restricted "window periods" when daylight inspections are not possible.This approach results in several challenges,including low efficiency,numerous blind spots,and limited coverage.To address these issues,this paper proposes a non-contact method for bridge crack detection by fusing realistic 3D models with infrared images.By fusing heterogeneous data from various sources,a detailed realistic 3D model of the bridge is constructed.The model is then spatially matched and fused with infrared 2D crack images,enabling rapid identification and precise positioning of cracks within the realistic 3D model.The main research work is summarized as follows:(1)Establishment of high-precision realistic 3D models of multi-strand girder bridges by integrating heterogeneous data from airborne and ground-based sources.This approach involves the conversion and registration of data from drone imagery and laser scanning point clouds to achieve integration of heterogeneous data from different sources.By doing so,a complete high-precision realistic 3D model of multi-strand girder bridges is established.This method solves the problem of missing textures,data gaps,positional and dimensional deviations in realistic 3D models caused by overlapping components when using a single drone to capture multi-strand girder bridges.The geometric accuracy of the realistic 3D model established by this method is high,with an absolute accuracy error of 11 mm for planar precision and 18 mm for elevation precision,and an average difference of 3mm for relative precision.(2)Spatial registration of 2D images and realistic 3D models based on texture mapping for cracks recognition.Firstly,the infrared image containing cracks is processed by grayscale conversion,filtering,enhancement and segmentation to obtain the feature image of cracks.Secondly,the texture mapping method is used to spatially register and fuse the feature 2D image of cracks and the realistic 3D model,thereby establishing a realistic 3D model with identified information of the area containing cracks.Finally,a crack area index database is created to achieve rapid identification and accurate positioning of the crack area.Validated and analyzed results show that the entire crack information is well presented,and the average error of the crack stitching position is 1.09 mm.(3)Inspection of a concrete arch bridge on a high-speed railway in Guilin.During the inspection,18 cracks with a maximum width of 0.36 mm were found in the structural components of the concrete arch bridge.A high-definition realistic 3D model was created,which includes information about the location of the cracks,fused crack images,and an index of the crack area.This model was provided to the railway maintenance department for repair and maintenance testing.Compared with manual inspection,the efficiency of this project was improved by approximately 19%,and the number of personnel required was reduced by approximately 33%.