不同的结果与pcl::MomentOfInertiaEstimation和pcl::CropBox？

Question

我试图从点云中得到一个OBB，使用pcl::MomentOfInertiaEstimation方法我得到了一个正确的结果，.I想用CropBox从OBB中提取点，但是我得到的点比原始的点云少得多，也许我的旋转向量有问题:v，我不确定。有谁可以帮我？

struct BoundingBox{ 
        pcl::PointXYZ min_point_OBB; 
        pcl::PointXYZ max_point_OBB; 
        pcl::PointXYZ position_OBB; 
        Eigen::Matrix3f rotational_matrix_OBB; 
        pcl::PointXYZ center; 
        pcl::PointXYZ x_axis; 
        pcl::PointXYZ y_axis; 
        pcl::PointXYZ z_axis; 

}; 
BoundingBox getOBB(pcl::PointCloud<pcl::PointXYZ>::Ptr  cloud,BoundingBox OBB ) 
{ 
  pcl::MomentOfInertiaEstimation <pcl::PointXYZ> feature_extractor; 
  feature_extractor.setInputCloud (cloud); 
  feature_extractor.compute (); 
  feature_extractor.getOBB (OBB.min_point_OBB, OBB.max_point_OBB, OBB.position_OBB, OBB.rotational_matrix_OBB); 
 return OBB; 
}   

int main(void){ 
BoundingBox OBB; 
OBB=getOBB(Npoints,OBB);                   //Npoints is a part of the whole cloud 
Eigen::Quaternionf quat (OBB.rotational_matrix_OBB); 
Eigen::Vector3f position (OBB.position_OBB.x, OBB.position_OBB.y, OBB.position_OBB.z); 
 pcl::visualization::PCLVisualizer viewer; 
pcl::visualization::PointCloudColorHandlerCustom<pcl::PointXYZ> points_color_handler (cloud, 255, 255, 255); 
 viewer.addCube (position, quat, OBB.max_point_OBB.x - OBB.min_point_OBB.x, OBB.max_point_OBB.y - OBB.min_point_OBB.y, OBB.max_point_OBB.z - OBB.min_point_OBB.z,"OBB"); 
viewer.setShapeRenderingProperties(pcl::visualization::PCL_VISUALIZER_REPRESENTATION, pcl::visualization::PCL_VISUALIZER_REPRESENTATION_WIREFRAME,"OBB"); 
////////////////above result is correct//////////////////// 
 Eigen::Vector4f max_point_OBB(OBB.max_point_OBB.x,OBB.max_point_OBB.y,OBB.max_point_OBB.z,1.0); 
Eigen::Vector4f min_point_OBB(OBB.min_point_OBB.x,OBB.min_point_OBB.y,OBB.min_point_OBB.z,1.0); 
pcl::PointCloud<pcl::PointXYZ>::Ptr cloudOut (new pcl::PointCloud<pcl::PointXYZ>); 
pcl::CropBox<pcl::PointXYZ> boxfilter; 
boxfilter.setMax(max_point_OBB);         
boxfilter.setMin(min_point_OBB); 
Eigen::Vector3f v = quat.vec(); 
boxfilter.setTranslation(position); 
boxfilter.setRotation(v); 
boxfilter.setInputCloud(cloud); 
boxfilter.filter (*cloudOut); 
 /////////cloulOut includes much less points than Npoints///////// 

寻找答案！

Answer 1

//boxfilter.setTranslation(position);
//boxfilter.setRotation(v);

Eigen::Translation3f translation(position_OBB.x, position_OBB.y, position_OBB.z);
Eigen::Matrix3f rotationMatrixOOB = quat.toRotationMatrix();
Eigen::Affine3f affine3f = translation * rotational_matrix_OBB;
boxfilter.setTransform(affine3f.inverse());