如何在PCL配准中使用kd-Tree以外的搜索方法？

针对有序点云场景下，想替换PCL配准模块默认的kd-Tree为OrganizedNeighbor搜索方法的需求，有两种可行实现方案：

方案一：自定义适配OrganizedNeighbor的CorrespondenceEstimation子类 #

PCL的CorrespondenceEstimation类默认依赖kd-Tree，但可以通过继承并重写核心方法来适配OrganizedNeighbor。核心思路是替换内部的搜索对象，同时处理setPointRepresentation这个仅kd-Tree支持的方法：

1. 定义子类继承自pcl::CorrespondenceEstimation<PointT, PointT>
2. 内部维护pcl::OrganizedNeighbor<PointT>对象替代原有的kd-Tree
3. 重写setInputTarget、computeCorrespondences等方法，调用OrganizedNeighbor的接口
4. 重写setPointRepresentation方法，直接抛出异常或忽略（因为OrganizedNeighbor不支持该功能）

示例代码片段：

#include <pcl/registration/correspondence_estimation.h>
#include <pcl/search/organized.h>

template <typename PointT>
class CorrespondenceEstimationOrganized : public pcl::CorrespondenceEstimation<PointT, PointT> {
public:
    using Ptr = boost::shared_ptr<CorrespondenceEstimationOrganized<PointT>>;
    using ConstPtr = boost::shared_ptr<const CorrespondenceEstimationOrganized<PointT>>;

    void setInputTarget(const typename pcl::PointCloud<PointT>::ConstPtr &target) override {
        target_ = target;
        tree_->setInputCloud(target);
    }

    void computeCorrespondences(pcl::Correspondences &correspondences, double max_distance = std::numeric_limits<double>::max()) override {
        correspondences.resize(source_->size());
        for (size_t i = 0; i < source_->size(); ++i) {
            std::vector<int> indices(1);
            std::vector<float> distances(1);
            tree_->nearestKSearch((*source_)[i], 1, indices, distances);
            if (distances[0] <= max_distance) {
                correspondences[i].index_query = static_cast<int>(i);
                correspondences[i].index_match = indices[0];
                correspondences[i].distance = distances[0];
            } else {
                correspondences[i].index_query = static_cast<int>(i);
                correspondences[i].index_match = -1;
            }
        }
    }

    // 重写setPointRepresentation，因为OrganizedNeighbor不支持该功能
    void setPointRepresentation(const typename pcl::PointRepresentation<PointT>::ConstPtr &) override {
        throw std::runtime_error("OrganizedNeighbor does not support PointRepresentation");
    }

private:
    typename pcl::OrganizedNeighbor<PointT>::Ptr tree_{new pcl::OrganizedNeighbor<PointT>()};
    using pcl::CorrespondenceEstimation<PointT, PointT>::source_;
    using pcl::CorrespondenceEstimation<PointT, PointT>::target_;
};

使用时，将配准类（如ICP）的对应估计器替换为自定义子类即可：

pcl::IterativeClosestPoint<PointXYZ, PointXYZ> icp;
auto corr_est = boost::make_shared<CorrespondenceEstimationOrganized<PointXYZ>>();
icp.setCorrespondenceEstimation(corr_est);

方案二：手动计算对应关系后传入配准类 #

如果不想自定义子类，更直接的方式是手动用OrganizedNeighbor计算所有源点到目标点的最近邻对应，再将结果传入配准类：

1. 初始化OrganizedNeighbor并设置目标点云
2. 遍历源点云每个点，调用nearestKSearch获取最近邻
3. 生成pcl::Correspondences对象存储有效对应（过滤距离过大的匹配）
4. 调用配准类的setCorrespondences方法，跳过内部对应估计步骤

示例代码片段：

#include <pcl/search/organized.h>
#include <pcl/registration/icp.h>

// 源点云和目标点云
pcl::PointCloud<pcl::PointXYZ>::ConstPtr source, target;

// 初始化OrganizedNeighbor
pcl::OrganizedNeighbor<pcl::PointXYZ> organized_neighbor;
organized_neighbor.setInputCloud(target);

pcl::Correspondences correspondences;
const double max_dist = 0.1; // 根据场景设置距离阈值

for (size_t i = 0; i < source->size(); ++i) {
    std::vector<int> indices(1);
    std::vector<float> distances(1);
    organized_neighbor.nearestKSearch((*source)[i], 1, indices, distances);
    if (distances[0] <= max_dist) {
        pcl::Correspondence corr;
        corr.index_query = static_cast<int>(i);
        corr.index_match = indices[0];
        corr.distance = distances[0];
        correspondences.push_back(corr);
    }
}

// 传入配准类
pcl::IterativeClosestPoint<pcl::PointXYZ, pcl::PointXYZ> icp;
icp.setInputSource(source);
icp.setInputTarget(target);
icp.setCorrespondences(correspondences);
// 执行配准
pcl::PointCloud<pcl::PointXYZ> aligned;
icp.align(aligned);

这种方法无需修改PCL原有类，灵活性更高，适合快速验证需求。

内容的提问来源于stack exchange，提问作者oarfish