Kinect-fusion ICP algorithm
Characteristics and Constraint conditions:
First, the kinectfusion algorithm adopts Frame-to-model, which is more accurate relative to frame-to-frame;
Second, through the first step to build a vector point cloud pyramid, from the highest level to match to the bottom, speed up convergence;
Thirdly, the change of position and posture between two adjacent frames is small;
When the environment is mainly composed of parallel planes, the ICP algorithm is not valid;
compute match points:
The first step, through the ray projection algorithm, calculates the point cloud that the previous position should get from the model.
In the second step, the point cloud of the current frame is converted to the global coordinate system, and then to the camera coordinate system of the previous frame, it is projected into the image coordinate system, and the points of the previous frame are computed.
The third step is to compare the above matching points with the normal vectors to see if they are the matching points.