registration accuracy

[LUOmengce]

I used Stanford's bunny point cloud data as the original point cloud, which has more than 30,000 points. Then, it is rotationally translated as a registration point cloud.
The first method I use ISS (Extracting 26 Keys) + FPFH + SAC_IA.
The second method I use RSCS + FPFH + SAC_IA.
The two method parameters are set as follows:
 FPFH has a radius of 50, numSP = 25, spSize = 100, csRad = 60
Use pcl library getFitnessScore () for evaluation, the first method is much higher than the second method of registration accuracy is high, I ask this reason, is my parameter settings is not right? Actually, I want to use the method in your dissertation, but my ability to understand and program is limited.
Thank you for your help, but also forgive me for my poor English proficiency.
Here is how I use the RSCS method Calculate the center points of the RSCSsuperpoints.
numSP = 25
spSize = 100
csRad = 60
pntCloud = loadDataSet1 ()
superPntList = createRandomSphereCoverSetFixedNum (pntCloud, superPointNum = numSP, pointsInSP = spSize, coverSphereRad = csRad)

def centerPoint (numSP): # Calculate the center points of the RSCSsuperpoints.
    superPointscenters = []
    for sp in superPntList:
        splen = len (sp [3])
        if splen> 0:
            for i in range (splen):
                if sp [3] [i] == numSP:
                    superPointscenters.append (sp [0: 3])

    superMat = np.mat (superPointscenters)
    superMat2 = np.mean (superMat, axis = 0)
    return superMat2

for i in range (numSP):
    superPntcenter = centerPoint (i)
    print (superPntcenter)

[gilbaz]

Ok great. I am traveling now, I will be back on Thursday. Then I can check out the parameter settings. We will make it work 😌 Thanks, Gil

…

On 27 Nov 2017, at 3:04, LUOmengce ***@***.***> wrote: I used Stanford's bunny point cloud data as the original point cloud, which has more than 30,000 points. Then, it is rotationally translated as a registration point cloud. The first method I use ISS (Extracting 26 Keys) + FPFH + SAC_IA. The second method I use RSCS + FPFH + SAC_IA. The two method parameters are set as follows: FPFH has a radius of 50, numSP = 25, spSize = 100, csRad = 60 Use pcl library getFitnessScore () for evaluation, the first method is much higher than the second method of registration accuracy is high, I ask this reason, is my parameter settings is not right? Actually, I want to use the method in your dissertation, but my ability to understand and program is limited. Thank you for your help, but also forgive me for my poor English proficiency. Here is how I use the RSCS method Calculate the center points of the RSCSsuperpoints. numSP = 25 spSize = 100 csRad = 60 pntCloud = loadDataSet1 () superPntList = createRandomSphereCoverSetFixedNum (pntCloud, superPointNum = numSP, pointsInSP = spSize, coverSphereRad = csRad) def centerPoint (numSP): # Calculate the center points of the RSCSsuperpoints. superPointscenters = [] for sp in superPntList: splen = len (sp [3]) if splen> 0: for i in range (splen): if sp [3] [i] == numSP: superPointscenters.append (sp [0: 3]) superMat = np.mat (superPointscenters) superMat2 = np.mean (superMat, axis = 0) return superMat2 for i in range (numSP): superPntcenter = centerPoint (i) print (superPntcenter) — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.