Methods and systems for crack detection

A system for detecting cracks in a surface, the system comprising:

• a video camera that senses RGB channels to generate overlapping successive frames of color images;
• a light source;
• means for simultaneously scanning the video camera and the light source past the surface while filming with the camera to produce a video of the surface so that the video contains a series of individual frames of the overlapping successive frames of the color images so that surface texture features that appear in a first individual frame of a consecutive frame pair of the overlapping successive frames also appear in at least a second individual frame of the consecutive frame pair;
• a convolutional neural network (CNN) architecture configured to analyze the individual frames of the overlapping successive frames of the video to obtain surface texture feature data for areas of the surface depicted in each of the individual frames and analyze the surface texture feature data to detect surface texture features in the areas of the surface depicted in each of the first and second individual frames of the consecutive frame pair;
• means for tracking the motion of the detected surface texture features in the first and second individual frames of the consecutive frame pair to produce tracking data;
• and
• means for using the tracking data to filter non-crack surface texture features from the detected surface texture features in the first and second individual frames of the consecutive frame pair.

A method for detecting cracks in a surface, the method comprising:

• simultaneously scanning a video camera and a light source past the surface while filming with the camera to produce a video of the surface so that the video contains a series of individual frames of overlapping successive frames of color images of RGB channels so that surface texture features that appear in a first individual frame of a consecutive frame pair of the overlapping successive frames also appear in at least a second individual frame of the consecutive frame pair;
• analyzing the individual frames of the overlapping successive frames of the video to obtain surface texture feature data for areas of the surface depicted in each of the first and second individual frames of the consecutive frame pair;
• analyzing the surface texture feature data to detect the surface texture features in the areas of the surface depicted in each of the first and second individual frames of the consecutive frame pair;
• tracking the motion of the detected surface texture features in the first and second individual frames of the consecutive frame pair to produce tracking data;
• and
• using the tracking data to filter non-crack surface texture features from the detected surface texture features in the first and second individual frames of the consecutive frame pair.