An Improved 3D Registration Method of Mobile Augmented Reality for Urban Built Environment

Table 1

Pseudocode of 3D registration algorithms.

Input: video frame
Output: registration matrix
Step 1:	Function CaculateRegistrationMatrix(video frame)
Step 2:	If exist a matching tag then
Step 9:	target tracking calculation using KLT with preImg, prePts and current frame curImg;
Step 9:	obtain the feature point nextPts of the tracking target;
Step 10:	if feature points are greater than the set threshold then
Step 11:	update the tracking parameters: PreImg=curImg, prePts=nextPts;
Step 13:	calculate the 3D registration matrix based on PNP algorithm of target feature points;
	else
Step 12:	delete matching tag;
	end if
	return;
Step 3:	else
Step 4:	SUFREAK algorithm extraction of feature points description of feature points;
Step 5:	carry out feature matching with offline target images;
Step 6:	if matched feature points are greater than the set threshold then
Step 7:	store the current frame image with feature points of preImg and prePts;
Step 7:	add matching tags;
Step 13:	calculate the 3D registration matrix based on PNP algorithm of garget feature points;
Step 8:	else
Step 8:	delete matching tag;
	end if
	return;
	end if
	End function