Why the PCL OrganizedMultiPlaneSegmentation are not useful.It segs few plans

Question

This is the normals estimation of every point using covariance matrix

Here is my code:

    pcl::PointCloud<pcl::Normal>::Ptr normals(new pcl::PointCloud<pcl::Normal>);
    pcl::NormalEstimation<pcl::PointXYZRGB, pcl::Normal> ne;
    pcl::search::KdTree<PointT>::Ptr tree(new pcl::search::KdTree<PointT>);
    pcl::OrganizedMultiPlaneSegmentation<PointT, pcl::Normal, pcl::Label> seg;
    pcl::IntegralImageNormalEstimation<PointT, pcl::Normal> ne_normals;
    std::vector<pcl::PlanarRegion<PointT>, Eigen::aligned_allocator<pcl::PlanarRegion<PointT>>> regions;

    cout << "start segmentation and display" << endl;

    start = clock();
    ne_normals.setNormalEstimationMethod(ne_normals.COVARIANCE_MATRIX);
    ne_normals.setMaxDepthChangeFactor(0.1f);
    ne_normals.setNormalSmoothingSize(20.0f);
    ne_normals.setInputCloud(this->cloud);
    ne_normals.compute(*normals);
    finish = clock();

    total_time = (double)(finish - start) / CLOCKS_PER_SEC;
    cout << "normal estimation TIME=" << total_time << endl;

    start = clock();
    seg.setInputCloud(this->cloud);

    seg.setInputNormals(normals);
    seg.setMinInliers(100);   //the minimum points a plane contain
    seg.setAngularThreshold(0.12);  
    seg.setDistanceThreshold(2);
    seg.segment(coefficients_vector, inliers_vector);
    finish = clock();
    total_time = (double)(finish - start) / CLOCKS_PER_SEC;

Here is the segment result:

  $:Start segmentation and display
  $:normal setimation TIME=9.563
  $:seg TIME=11.138
  $:segmented planes number:2
  $: Plane 1 contains:170 points
  $: Plane 2 contains:169 points

Question: From the normal estimation result,the estimation is not bad, why the multiplaneSeg are very few and bad? Anyone could help me?

this is the original picture(640*480) got from kinect,and plot seg result in it