Self position estimation device, self position estimation method, program, and image processing device

A self position estimation method that estimates a self position of a camera-equipped mobile robot performing inspection of a structure by subjecting the structure to split imaging, the self position estimation method comprising:

• acquiring a plurality of structure images including at least a first structure image obtained by subjecting the structure to split imaging at a first position with the camera-equipped mobile robot and a second structure image obtained by subjecting the structure to split imaging at a second position with the camera-equipped mobile robot after the camera-equipped mobile robot moves along an imaging plan from the first position;
• generating a panorama composite image by subjecting the plurality of structure images including the first structure image and the second structure image to panorama composition;
• storing, by a storage unit, the panorama composite image;
• acquiring second image coordinates as coordinates of a specific position of the second structure image on the panorama composite image based on a moving distance of the camera-equipped mobile robot on the panorama composite image with first image coordinates corresponding to the first position in the panorama composite image which is read from the storage unit as an origin of a coordinate system of the panorama composite image;
• calculating a normal vector of a plane of the structure using a stereo image captured by the camera-equipped mobile robot;
• correcting the second image coordinates using a misregistration amount which is an amount from a center of the second structure image which is estimated as a camera position;
• and
• calculating relative position coordinates as relative actual coordinates of the second position as a relative self position using a transformation coefficient for transforming the second image coordinates from an image coordinate system to an actual coordinate system, the transformation coefficient being calculated from an imaging range divided by a number of pixels of an image captured by the camera, and the imaging range being calculated from a distance between the camera and the structure, a focal length of the camera, and a size of an imaging element of the camera.

A non-transitory computer-readable tangible medium comprising computer executable instructions stored thereon, which, when executed by one or more processors, cause the one or more processors to perform a self position estimation method that estimates a self position of a camera-equipped mobile robot performing inspection of a structure by subjecting the structure to split imaging, the self position estimation method comprising:

• acquiring a plurality of structure images including at least a first structure image obtained by subjecting the structure to split imaging at a first position with the camera-equipped mobile robot and a second structure image obtained by subjecting the structure to split imaging at a second position with the camera-equipped mobile robot after the camera-equipped mobile robot moves along an imaging plan from the first position;
• generating a panorama composite image by subjecting the plurality of structure images including the first structure image and the second structure image to panorama composition;
• storing, by a storage unit, the panorama composite image;
• acquiring second image coordinates as coordinates of a specific position of the second structure image on the panorama composite image based on a moving distance of the camera-equipped mobile robot on the panorama composite image with first image coordinates corresponding to the first position in the panorama composite image which is read from the storage unit as an origin of a coordinate system of the panorama composite image;
• calculating a normal vector of a plane of the structure using a stereo image captured by the camera-equipped mobile robot;
• correcting the second image coordinates using a misregistration amount which is an amount from a center of the second structure image which is estimated as a camera position;
• and
• calculating relative position coordinates as relative actual coordinates of the second position as a relative self position using a transformation coefficient for transforming the second image coordinates from an image coordinate system to an actual coordinate system, the transformation coefficient being calculated from an imaging range divided by a number of pixels of an image captured by the camera, and the imaging range being calculated from a distance between the camera and the structure, a focal length of the camera, and a size of an imaging element of the camera.