PCL关键点(1)
关键点也称为兴趣点,它是2D图像或是3D点云或者曲面模型上,可以通过定义检测标准来获取的具有稳定性,区别性的点集,从技术上来说,关键点的数量相比于原始点云或图像的数据量减小很多,与局部特征描述子结合在一起,组成关键点描述子常用来形成原始数据的表示,而且不失代表性和描述性,从而加快了后续的识别,追踪等对数据的处理了速度,故而,关键点技术成为在2D和3D 信息处理中非常关键的技术
NARF(Normal Aligned Radial Feature)关键点是为了从深度图像中识别物体而提出的,对NARF关键点的提取过程有以下要求:
a) 提取的过程考虑边缘以及物体表面变化信息在内;
b)在不同视角关键点可以被重复探测;
c)关键点所在位置有足够的支持区域,可以计算描述子和进行唯一的估计法向量。
其对应的探测步骤如下:
(1) 遍历每个深度图像点,通过寻找在近邻区域有深度变化的位置进行边缘检测。
(2) 遍历每个深度图像点,根据近邻区域的表面变化决定一测度表面变化的系数,及变化的主方向。
(3) 根据step(2)找到的主方向计算兴趣点,表征该方向和其他方向的不同,以及该处表面的变化情况,即该点有多稳定。
(4) 对兴趣值进行平滑滤波。
(5) 进行无最大值压缩找到的最终关键点,即为NARF关键点。
关于NARF的更为具体的描述请查看这篇博客www.cnblogs.com/ironstark/p/5051533.html。
PCL中keypoints模块及类的介绍
(1)class pcl::Keypoint<PointInT,PointOutT> 类keypoint是所有关键点检测相关类的基类,定义基本接口,具体实现由子类来完成,其继承关系时下图:

具体介绍:
Public Member Functions |
|
virtual void |
setSearchSurface (const PointCloudInConstPtr &cloud) |
设置搜索时所用搜索点云,cloud为指向点云对象的指针引用 |
|
void |
setSearchMethod (const KdTreePtr &tree) 设置内部算法实现时所用的搜索对象,tree为指向kdtree或者octree对应的指针 |
void |
setKSearch (int k) 设置K近邻搜索时所用的K参数 |
void |
setRadiusSearch (double radius) 设置半径搜索的半径的参数 |
int |
searchForNeighbors (int index, double parameter, std::vector< int > &indices, std::vector< float > &distances) const |
采用setSearchMethod设置搜索对象,以及setSearchSurface设置搜索点云,进行近邻搜索,返回近邻在点云中的索引向量, indices以及对应的距离向量distance其中为查询点的索引,parameter为搜索时所用的参数半径或者K |
(2)class pcl::HarrisKeypoint2D<PointInT,PointOutT,IntensityT>
类HarrisKeypoint2D实现基于点云的强度字段的harris关键点检测子,其中包括多种不同的harris关键点检测算法的变种,其关键函数的说明如下:
Public Member Functions |
|
HarrisKeypoint2D (ResponseMethod method=HARRIS, int window_width=3, int window_height=3, int min_distance=5, float threshold=0.0) |
|
重构函数,method需要设置采样哪种关键点检测方法,有HARRIS,NOBLE,LOWE,WOMASI四种方法,默认为HARRIS,window_width window_height为检测窗口的宽度和高度min_distance 为两个关键点之间 容许的最小距离,threshold为判断是否为关键点的感兴趣程度的阀值,小于该阀值的点忽略,大于则认为是关键点 |
|
void |
setMethod (ResponseMethod type)设置检测方式 |
void |
setWindowWidth (int window_width) 设置检测窗口的宽度 |
void |
setWindowHeight (int window_height) 设置检测窗口的高度 |
void |
setSkippedPixels (int skipped_pixels) 设置在检测时每次跳过的像素的数目 |
void |
setMinimalDistance (int min_distance) 设置候选关键点之间的最小距离 |
void |
setThreshold (float threshold) 设置感兴趣的阀值 |
void |
setNonMaxSupression (bool=false) 设置是否对小于感兴趣阀值的点进行剔除,如果是true则剔除,否则返回这个点 |
void |
setRefine (bool do_refine)设置是否对所得的关键点结果进行优化, |
void |
setNumberOfThreads (unsigned int nr_threads=0) 设置该算法如果采用openMP并行机制,能够创建线程数目 |
(3)pcl::HarrisKeypoint3D< PointInT, PointOutT, NormalT >
类HarrisKeypoint3D和HarrisKeypoint2D类似,但是没有在点云的强度空间检测关键点,而是利用点云的3D空间的信息表面法线向量来进行关键点检测,关于HarrisKeypoint3D的类与HarrisKeypoint2D相似,除了
HarrisKeypoint3D (ResponseMethod method=HARRIS, float radius=0.01f, float threshold=0.0f)
重构函数,method需要设置采样哪种关键点检测方法,有HARRIS,NOBLE,LOWE,WOMASI四种方法,默认为HARRIS,radius为法线估计的搜索半径,threshold为判断是否为关键点的感兴趣程度的阀值,小于该阀值的点忽略,大于则认为是关键点。
(4)pcl::HarrisKeypoint6D< PointInT, PointOutT, NormalT >
类HarrisKeypoint6D和HarrisKeypoint2D类似,只是利用了欧式空间域XYZ或者强度域来候选关键点,或者前两者的交集,即同时满足XYZ域和强度域的关键点为候选关键点,
HarrisKeypoint6D (float radius=0.01, float threshold=0.0) 重构函数,此处并没有方法选择的参数,而是默认采用了Tomsai提出的方法实现关键点的检测,radius为法线估计的搜索半径,threshold为判断是否为关键点的感兴趣程度的阀值,小于该阀值的点忽略,大于则认为是关键点。
(5)pcl::SIFTKeypoint< PointInT, PointOutT >
类SIFTKeypoint是将二维图像中的SIFT算子调整后移植到3D空间的SIFT算子的实现,输入带有XYZ坐标值和强度的点云,输出为点云中的SIFT关键点,其关键函数的说明如下:
void | setScales (float min_scale, int nr_octaves, int nr_scales_per_octave) |
设置搜索时与尺度相关的参数,min_scale在点云体素尺度空间中标准偏差,点云对应的体素栅格中的最小尺寸 | |
int nr_octaves是检测关键点时体素空间尺度的数目,nr_scales_per_octave为在每一个体素空间尺度下计算高斯空间的尺度所需要的参数 | |
void | setMinimumContrast (float min_contrast) 设置候选关键点对应的对比度下限 |
(6)还有很多不再一一介绍
实例分析
实验实现提取NARF关键点,并且用图像和3D显示的方式进行可视化,可以直观的观察关键点的位置和数量 narf_feature_extraction.cpp:
#include <iostream>
#include <boost/thread/thread.hpp>
#include <pcl/range_image/range_image.h>
#include <pcl/io/pcd_io.h>
#include <pcl/visualization/range_image_visualizer.h>
#include <pcl/visualization/pcl_visualizer.h>
#include <pcl/features/range_image_border_extractor.h>
#include <pcl/keypoints/narf_keypoint.h>
#include <pcl/features/narf_descriptor.h>
#include <pcl/console/parse.h>
typedef pcl::PointXYZ PointType;
float angular_resolution = 0.5f; //angular_resolution为模拟的深度传感器的角度分辨率,即深度图像中一个像素对应的角度大小
float support_size = 0.2f; //点云大小的设置
pcl::RangeImage::CoordinateFrame coordinate_frame = pcl::RangeImage::CAMERA_FRAME; //设置坐标系
bool setUnseenToMaxRange = false;
bool rotation_invariant = true;
void printUsage (const char* progName) { std::cout << "\n\nUsage: "<<progName<<" [options] <scene.pcd>\n\n" << "Options:\n" << "-------------------------------------------\n" << "-r <float> angular resolution in degrees (default "<<angular_resolution<<")\n" << "-c <int> coordinate frame (default "<< (int)coordinate_frame<<")\n" << "-m Treat all unseen points to max range\n" << "-s <float> support size for the interest points (diameter of the used sphere - ""default "<<support_size<<")\n" << "-o <0/1> switch rotational invariant version of the feature on/off" << " (default "<< (int)rotation_invariant<<")\n" << "-h this help\n" << "\n\n"; }void setViewerPose (pcl::visualization::PCLVisualizer& viewer, const Eigen::Affine3f& viewer_pose) //设置视口的位姿{ Eigen::Vector3f pos_vector = viewer_pose * Eigen::Vector3f (0, 0, 0);
//视口的原点pos_vector Eigen::Vector3f look_at_vector = viewer_pose.rotation () * Eigen::Vector3f (0, 0, 1) + pos_vector; //旋转+平移look_at_vector Eigen::Vector3f up_vector = viewer_pose.rotation () * Eigen::Vector3f (0, -1, 0); //up_vector viewer.setCameraPosition (pos_vector[0], pos_vector[1], pos_vector[2], look_at_vector[0], look_at_vector[1], look_at_vector[2], up_vector[0], up_vector[1], up_vector[2]); }
int main (int argc, char** argv) { if (pcl::console::find_argument (argc, argv, "-h") >= 0) { printUsage (argv[0]); return 0; } if (pcl::console::find_argument (argc, argv, "-m") >= 0) { setUnseenToMaxRange = true; cout << "Setting unseen values in range image to maximum range readings.\n"; } if (pcl::console::parse (argc, argv, "-o", rotation_invariant) >= 0) cout << "Switching rotation invariant feature version "<< (rotation_invariant ? "on" : "off")<<".\n"; int tmp_coordinate_frame; if (pcl::console::parse (argc, argv, "-c", tmp_coordinate_frame) >= 0) { coordinate_frame = pcl::RangeImage::CoordinateFrame (tmp_coordinate_frame); cout << "Using coordinate frame "<< (int)coordinate_frame<<".\n"; } if (pcl::console::parse (argc, argv, "-s", support_size) >= 0) cout << "Setting support size to "<<support_size<<".\n"; if (pcl::console::parse (argc, argv, "-r", angular_resolution) >= 0) cout << "Setting angular resolution to "<<angular_resolution<<"deg.\n"; angular_resolution = pcl::deg2rad (angular_resolution); // -Read pcd file or create example point cloud if not given-- // pcl::PointCloud<PointType>::Ptr point_cloud_ptr (new pcl::PointCloud<PointType>); pcl::PointCloud<PointType>& point_cloud = *point_cloud_ptr; pcl::PointCloud<pcl::PointWithViewpoint> far_ranges; Eigen::Affine3f scene_sensor_pose (Eigen::Affine3f::Identity ()); std::vector<int> pcd_filename_indices = pcl::console::parse_file_extension_argument (argc, argv, "pcd"); if (!pcd_filename_indices.empty ()) { std::string filename = argv[pcd_filename_indices[0]]; if (pcl::io::loadPCDFile (filename, point_cloud) == -1) { cerr << "Was not able to open file \""<<filename<<"\".\n"; printUsage (argv[0]); return 0; }
scene_sensor_pose = Eigen::Affine3f (Eigen::Translation3f (point_cloud.sensor_origin_[0], point_cloud.sensor_origin_[1],point_cloud.sensor_origin_[2])) * Eigen::Affine3f (point_cloud.sensor_orientation_); std::string far_ranges_filename = pcl::getFilenameWithoutExtension (filename)+"_far_ranges.pcd"; if (pcl::io::loadPCDFile (far_ranges_filename.c_str (), far_ranges) == -1) std::cout << "Far ranges file \""<<far_ranges_filename<<"\" does not exists.\n"; } else { setUnseenToMaxRange = true; cout << "\nNo *.pcd file given => Genarating example point cloud.\n\n"; for (float x=-0.5f; x<=0.5f; x+=0.01f) { for (float y=-0.5f; y<=0.5f; y+=0.01f) { PointType point; point.x = x; point.y = y; point.z = 2.0f - y; point_cloud.points.push_back (point); } } point_cloud.width = (int) point_cloud.points.size (); point_cloud.height = 1; } // -----Create RangeImage from the PointCloud----- // float noise_level = 0.0; float min_range = 0.0f; int border_size = 1; boost::shared_ptr<pcl::RangeImage> range_image_ptr (new pcl::RangeImage); pcl::RangeImage& range_image = *range_image_ptr; range_image.createFromPointCloud (point_cloud, angular_resolution, pcl::deg2rad (360.0f), pcl::deg2rad (180.0f), scene_sensor_pose, coordinate_frame, noise_level, min_range, border_size); range_image.integrateFarRanges (far_ranges); if (setUnseenToMaxRange) range_image.setUnseenToMaxRange (); // -----Open 3D viewer and add point cloud----- // pcl::visualization::PCLVisualizer viewer ("3D Viewer"); viewer.setBackgroundColor (1, 1, 1); pcl::visualization::PointCloudColorHandlerCustom<pcl::PointWithRange> range_image_color_handler (range_image_ptr, 0, 0, 0); viewer.addPointCloud (range_image_ptr, range_image_color_handler, "range image"); viewer.setPointCloudRenderingProperties (pcl::visualization::PCL_VISUALIZER_POINT_SIZE, 1, "range image");
//viewer.addCoordinateSystem (1.0f, "global");
//PointCloudColorHandlerCustom<PointType> point_cloud_color_handler (point_cloud_ptr, 150, 150, 150);
//viewer.addPointCloud (point_cloud_ptr, point_cloud_color_handler, "original point cloud");
viewer.initCameraParameters (); setViewerPose (viewer, range_image.getTransformationToWorldSystem ()); // -----Show range image----- // pcl::visualization::RangeImageVisualizer range_image_widget ("Range image"); range_image_widget.showRangeImage (range_image);
/* 创建RangeImageBorderExtractor对象,它是用来进行边缘提取的,因为NARF的第一步就是需要探测出深度图像的边缘*/ // -----Extract NARF keypoints----- // pcl::RangeImageBorderExtractor range_image_border_extractor; //用来提取边缘 pcl::NarfKeypoint narf_keypoint_detector; //用来检测关键点 narf_keypoint_detector.setRangeImageBorderExtractor (&range_image_border_extractor); // narf_keypoint_detector.setRangeImage (&range_image); narf_keypoint_detector.getParameters ().support_size = support_size; //设置NARF的参数 pcl::PointCloud<int> keypoint_indices; narf_keypoint_detector.compute (keypoint_indices); std::cout << "Found "<<keypoint_indices.points.size ()<<" key points.\n"; // -----Show keypoints in 3D viewer----- // pcl::PointCloud<pcl::PointXYZ>::Ptr keypoints_ptr (new pcl::PointCloud<pcl::PointXYZ>); pcl::PointCloud<pcl::PointXYZ>& keypoints = *keypoints_ptr; keypoints.points.resize (keypoint_indices.points.size ()); for (size_t i=0; i<keypoint_indices.points.size (); ++i) keypoints.points[i].getVector3fMap () = range_image.points[keypoint_indices.points[i]].getVector3fMap (); pcl::visualization::PointCloudColorHandlerCustom<pcl::PointXYZ> keypoints_color_handler (keypoints_ptr, 0, 255, 0); viewer.addPointCloud<pcl::PointXYZ> (keypoints_ptr, keypoints_color_handler, "keypoints"); viewer.setPointCloudRenderingProperties (pcl::visualization::PCL_VISUALIZER_POINT_SIZE, 7, "keypoints"); // -----Extract NARF descriptors for interest points----- // std::vector<int> keypoint_indices2; keypoint_indices2.resize (keypoint_indices.points.size ()); for (unsigned int i=0; i<keypoint_indices.size (); ++i) // This step is necessary to get the right vector type keypoint_indices2[i]=keypoint_indices.points[i]; pcl::NarfDescriptor narf_descriptor (&range_image, &keypoint_indices2); narf_descriptor.getParameters ().support_size = support_size; narf_descriptor.getParameters ().rotation_invariant = rotation_invariant; pcl::PointCloud<pcl::Narf36> narf_descriptors; narf_descriptor.compute (narf_descriptors); cout << "Extracted "<<narf_descriptors.size ()<<" descriptors for " <<keypoint_indices.points.size ()<< " keypoints.\n"; while (!viewer.wasStopped ()) { range_image_widget.spinOnce (); // process GUI events viewer.spinOnce (); pcl_sleep(0.01); } }
运行结果:
