| CPC G06V 10/245 (2022.01) [G06T 7/246 (2017.01); G06T 7/60 (2013.01); G06T 7/73 (2017.01); G06T 2207/20081 (2013.01); G06V 2201/07 (2022.01)] | 8 Claims |
|
1. A key point correction system comprising:
at least one memory storing instructions; and
at least one processor that executes the instructions to:
acquire a plurality of captured images;
detect a plurality of key points of a moving object from each of the captured images;
generate one or more orthoimages by performing orthorectification on the plurality of captured images, the one or more orthoimages include a plurality of key point correspondence points that correspond to the plurality of key points projected onto the one or more orthoimages;
decide whether or not the plurality of key points of the moving object detected from the plurality of captured images satisfy a predetermined appropriacy condition;
determine, when the plurality of key points of the moving object satisfy the appropriacy condition, a correction parameter to be used to correct the plurality of key points detected for the moving object using the plurality of key point correspondence points on the one or more orthoimages, a rectangular area on the one or more orthoimages that represents the moving object, and determining one or more dimensions of the rectangular area as the correction parameter, the one or more dimensions including a combination of width and height of the rectangular area; and
correct, when the correction parameter has been determined for the moving object regarding which the plurality of key points are detected from the plurality of captured images, the plurality of key points by using the correction parameter,
wherein the appropriacy condition indicates a decision range, which is a range that a position, a size, or a posture of the moving object, or a combination of two or more of them is required to satisfy, the position or the posture of the moving object including a plurality of surfaces of the moving object in a three-dimensional manner, and
the at least one processor further executes the instructions to:
determine the position, the size, or the posture of the moving object using the plurality of key points of the moving object; and
decide that the appropriacy condition is satisfied when the determined position, size, or posture, or a combination of two or more of them is included in the decision range;
compute the plurality of key point correspondence points on the one or more orthoimages that correspond to the plurality of respective key points by performing orthorectification on the plurality of key points of the moving object detected from the plurality of captured images;
compute the position, the size, or the posture of the moving object on the one or more orthoimages that is determined based on the positions of the plurality of the computed key point correspondence points as the position, the size, or the posture of the moving object;
train an object detection model to output the plurality of key points of the moving object included in the plurality of captured images in response to inputting the plurality of captured images into the object detection model;
track the moving object by determining whether the moving object is the same that has already been detected from another captured image from the plurality of captured images; and
allocate, in response to a result of the determination that the moving object is the same that has already been detected from the another captured image from the plurality of captured images, a first identifier that is different from a second identifier that is associated with determining that the moving object has already been detected.
|