Crack detection, assessment and visualization using deep learning with 3d mesh model

A method comprising

• acquiring a set of two-dimensional (2D) images of infrastructure;
• identifying and segmenting, by a deep learning processes executing on one or more electronic devices, surface cracks in one or more of the 2D images of the set of 2D images to produce a set of segmented 2D images that are each divided into crack pixels and non-crack pixels;
• generating, by a photogrammetry process executing on one or more electronic devices, a textured three-dimensional (3D) mesh model of the infrastructure from the segmented 2D images, the 3D mesh model including a polygon mesh and a texture cover that includes the identified surface cracks;
• and
• displaying the textured 3D mesh model to provide a visualization of identified surface cracks.

A non-transitory electronic device-readable medium having executable instructions stored thereon, the instructions when executed by one or more processors being operable to:

• acquire a set of two-dimensional (2D) images of infrastructure from a handheld camera and/or a camera of an unmanned aerial vehicle (UAV);
• identify and segment surface cracks in one or more of the 2D images of the set of 2D images using a Faster region-based convolutional neural network (Faster-RCNN) to identify surface cracks and a structured random forest edge detection (SFRED) technique to segment surface cracks, or using a Mask region-based convolutional neural network (Mask-RCNN) to both identify and segment surface cracks;
• generate a textured three-dimensional (3D) mesh model of the infrastructure, the 3D mesh model including a polygon mesh and a texture cover that indicates the identified surface cracks;
• and
• store the textured 3D mesh model.