描述子提取 / Descriptor Extraction

Functions
template<typename Signature >
std::vector< std::pair< size_t, size_t > >	toolbox::pcl::match_descriptors (const std::vector< Signature > &source_descriptors, const std::vector< Signature > &target_descriptors, typename Signature::data_type max_distance)
	描述子类型及其特点 / Descriptor types and their characteristics
template<typename Signature >
std::vector< std::pair< size_t, size_t > >	toolbox::pcl::match_descriptors_ratio_test (const std::vector< Signature > &source_descriptors, const std::vector< Signature > &target_descriptors, float ratio_threshold=0.8f)
	使用比率测试的描述子匹配 / Descriptor matching with ratio test
template<typename Signature >
Signature	toolbox::pcl::compute_descriptor_centroid (const std::vector< Signature > &descriptors)
	计算描述子集的统计信息 / Compute statistics of descriptor set
template<typename Signature >
std::pair< float, float >	toolbox::pcl::evaluate_descriptor_matching (const std::vector< Signature > &descriptors1, const std::vector< Signature > &descriptors2, const std::vector< std::pair< size_t, size_t > > &ground_truth_matches, typename Signature::data_type max_distance)
	描述子性能评估辅助函数 / Descriptor performance evaluation helper

Detailed Description

Function Documentation

compute_descriptor_centroid()

template<typename Signature >

Signature toolbox::pcl::compute_descriptor_centroid

(

const std::vector< Signature > &

descriptors

)

计算描述子集的统计信息 / Compute statistics of descriptor set

Template Parameters

Signature

描述子签名类型 / Descriptor signature type

Parameters

descriptors

描述子集 / Descriptor set

Returns

平均描述子（用于聚类等） / Average descriptor (for clustering etc.)

// 计算描述子集的中心 / Compute centroid of descriptor set
auto centroid = compute_descriptor_centroid(descriptors);

evaluate_descriptor_matching()

template<typename Signature >

std::pair< float, float > toolbox::pcl::evaluate_descriptor_matching	(	const std::vector< Signature > &	descriptors1,
		const std::vector< Signature > &	descriptors2,
		const std::vector< std::pair< size_t, size_t > > &	ground_truth_matches,
		typename Signature::data_type	max_distance
	)

描述子性能评估辅助函数 / Descriptor performance evaluation helper

Template Parameters

Signature

描述子签名类型 / Descriptor signature type

Parameters

descriptors1	第一组描述子 / First descriptor set
descriptors2	第二组描述子 / Second descriptor set
ground_truth_matches	真实匹配对 / Ground truth matches

Returns

准确率和召回率 / Precision and recall

// 评估描述子匹配性能 / Evaluate descriptor matching performance
auto [precision, recall] = evaluate_descriptor_matching(
    desc1, desc2, ground_truth, 0.25f);
std::cout << "准确率 / Precision: " << precision << std::endl;
std::cout << "召回率 / Recall: " << recall << std::endl;

match_descriptors()

template<typename Signature >

std::vector< std::pair< size_t, size_t > > toolbox::pcl::match_descriptors	(	const std::vector< Signature > &	source_descriptors,
		const std::vector< Signature > &	target_descriptors,
		typename Signature::data_type	max_distance
	)

描述子类型及其特点 / Descriptor types and their characteristics

局部描述子 / Local Descriptors:

• PFH (Point Feature Histogram): 125维，精确但计算较慢 / 125D, accurate but slow
• FPFH (Fast PFH): 33维，PFH的快速近似版本 / 33D, fast approximation of PFH
• SHOT (Signature of Histograms of OrienTations): 352维，包含颜色和形状信息 / 352D, includes color and shape
• 3DSC (3D Shape Context): 1980维，基于球形网格的形状描述 / 1980D, spherical grid-based shape description
• ROPS (Rotational Projection Statistics): 135维，旋转不变性 / 135D, rotation invariant

全局描述子 / Global Descriptors:

• VFH (Viewpoint Feature Histogram): 308维，整个物体的描述 / 308D, describes entire object
• CVFH (Clustered VFH): 308维×聚类数，处理遮挡情况 / 308D×clusters, handles occlusion

// 选择合适的描述子 / Choose appropriate descriptor
if (need_fast_matching) {
    // FPFH: 快速且效果良好 / FPFH: fast and effective
    fpfh_extractor_t<float> extractor;
} else if (need_high_accuracy) {
    // SHOT: 更高的区分度 / SHOT: higher discriminative power
    shot_extractor_t<float> extractor;
} else if (need_global_descriptor) {
    // VFH: 全局物体描述 / VFH: global object description
    vfh_extractor_t<float> extractor;
}

描述子匹配辅助函数 / Descriptor matching helper function

Template Parameters

Signature

描述子签名类型 / Descriptor signature type

Parameters

source_descriptors	源描述子集 / Source descriptor set
target_descriptors	目标描述子集 / Target descriptor set
max_distance	最大匹配距离 / Maximum matching distance

Returns

匹配对的索引 / Indices of matched pairs

// 匹配两个点云的描述子 / Match descriptors between two point clouds
auto matches = match_descriptors(source_desc, target_desc, 0.25f);
 
for (const auto& [src_idx, tgt_idx] : matches) {
    std::cout << "匹配 / Match: " << src_idx << " -> " << tgt_idx << std::endl;
}

match_descriptors_ratio_test()

template<typename Signature >

std::vector< std::pair< size_t, size_t > > toolbox::pcl::match_descriptors_ratio_test	(	const std::vector< Signature > &	source_descriptors,
		const std::vector< Signature > &	target_descriptors,
		float	ratio_threshold = 0.8f
	)

使用比率测试的描述子匹配 / Descriptor matching with ratio test

Template Parameters

Signature

描述子签名类型 / Descriptor signature type

Parameters

source_descriptors	源描述子集 / Source descriptor set
target_descriptors	目标描述子集 / Target descriptor set
ratio_threshold	比率阈值（通常0.7-0.8） / Ratio threshold (typically 0.7-0.8)

Returns

匹配对的索引 / Indices of matched pairs

// 使用Lowe's ratio test进行鲁棒匹配 / Robust matching using Lowe's ratio test
auto matches = match_descriptors_ratio_test(source_desc, target_desc, 0.8f);