CN116977593A - Single wood segmentation method based on super-voxel concave-convex segmentation and color region growth - Google Patents
Single wood segmentation method based on super-voxel concave-convex segmentation and color region growth Download PDFInfo
- Publication number
- CN116977593A CN116977593A CN202210429377.3A CN202210429377A CN116977593A CN 116977593 A CN116977593 A CN 116977593A CN 202210429377 A CN202210429377 A CN 202210429377A CN 116977593 A CN116977593 A CN 116977593A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- voxel
- super
- convex
- segmentation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000011218 segmentation Effects 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 title claims abstract description 27
- 239000002023 wood Substances 0.000 title claims abstract description 26
- 238000001914 filtration Methods 0.000 claims abstract description 9
- 238000012545 processing Methods 0.000 claims abstract description 7
- 238000005516 engineering process Methods 0.000 claims abstract description 6
- 238000010606 normalization Methods 0.000 claims abstract description 5
- 239000003086 colorant Substances 0.000 claims 1
- 210000000056 organ Anatomy 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 4
- 239000004744 fabric Substances 0.000 description 6
- 239000002245 particle Substances 0.000 description 6
- 238000011835 investigation Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000005484 gravity Effects 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 241001632422 Radiola linoides Species 0.000 description 1
- 230000001464 adherent effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000031702 trunk segmentation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/20—Finite element generation, e.g. wire-frame surface description, tesselation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention aims to provide a single wood segmentation algorithm for forest land point cloud data of a foundation laser radar based on super-voxel concave-convex segmentation and point cloud color region growth, relates to a method for extracting single wood of a dense forest land, solves the problems that the direct adhesion segmentation effect of adjacent trees of a complex forest land is poor and the super-voxel clustering cannot consider the point cloud color, improves the single wood segmentation precision, and further promotes the wide application of the foundation laser radar in forestry. The specific method comprises the following steps: carrying out accurate multi-station registration on the point cloud data to obtain complete forest land point cloud data; carrying out normalization processing on the point cloud data elevation; separating ground treatment is carried out on the forest land point cloud data by using CSF filtering in the PCL point cloud library; performing super-voxel clustering on the tree point cloud data in a three-dimensional space; clustering different voxel blocks according to the concave-convex relation among different voxel areas and the growing algorithm of the convex areas, and dividing the adhered trees; and carrying out region growth on the clustering result according to the point cloud color rule, clustering the segmented regions, and further extracting complete Shan Mudian cloud data. The method is used in the forestry application field of the laser radar technology.
Description
Technical Field
The invention relates to a single wood segmentation method of forest land point cloud data based on a foundation laser radar, and belongs to application of the foundation laser radar technology in the field of forestry.
Background
The foundation laser radar is used as an active remote sensing technology, can accurately acquire three-dimensional information of a target object, and is gradually applied to forestry research in recent years. For forest resource investigation, the traditional method for acquiring the structural parameter information of the single wood has low efficiency and poor precision, and along with the wide application of the foundation radar technology in forest resource investigation, the forest investigation cost is saved to a great extent and the investigation efficiency is improved.
At present, a plurality of algorithms for extracting single wood based on the point cloud data of the point cloud technology of the foundation laser radar are available, and the algorithms can be divided into 3 kinds of algorithms, namely clustering single wood segmentation, four-time polynomial fitting single wood segmentation and canopy height model-based segmentation, wherein the methods are based on the two-dimensional single wood segmentation of the point cloud data. For dense forests, trees with shielding and adhesion are separated by a single method, which is time-consuming and can be over-segmented. In recent years, the super-voxel clustering method is gradually applied, can be directly divided according to the three-dimensional information of the point cloud data, and only depends on the relation between the spatial characteristics of the point cloud and the normal vector, so that the super-voxel clustering method can reduce the calculation complexity and further improve the subsequent processing speed. Based on the result of super-voxel clustering, researchers have proposed the segmentation of super-voxel concave-convex areas, which can judge the concave-convex relation of each clustered area and perform area growth. This method of segmentation of the super-voxel relief region has a significant effect on crown and trunk segmentation and some adherent object segmentation. Thus, adjacent trees can be segmented out in a forest pattern where canopy density and stand are more complex. However, the LocallyConvexConnectedPatches (LCCP) local lug communication method is used for dividing, only three-dimensional space geometric features are considered, the color features of the point cloud are ignored, and the phenomenon of wrong division exists on the same branches, leaves and the like of a tree.
In order to avoid the phenomenon, the ultra-voxel concave-convex area and the color-based area growth are combined, the area growth is inquired according to the color characteristics of the point cloud, and the same branches She Julei of the tree are subjected to the method, so that complete tree point cloud data are obtained.
Disclosure of Invention
The invention aims to solve the problems, and provides a single-wood segmentation method for super voxel convexity segmentation and color region growth of a forest land point cloud based on a foundation laser radar, so that an effective method is provided for single-wood segmentation extraction of Lin Fendian cloud data with higher canopy density.
The invention discloses a single wood segmentation method for super voxel convexity segmentation and color region growth of a forest land point cloud based on a foundation laser radar, which can effectively separate and extract single trees from a forest stand with high canopy closure, and comprises the following steps:
step one: the method comprises the steps of utilizing a ground three-dimensional laser scanner to acquire forest site cloud data, and utilizing point cloud processing software to accurately register the site cloud data in a multi-station mode to acquire complete forest site cloud data;
step two: extracting tree point clouds from the complete woodland data point clouds obtained in the step one by adopting a point cloud filtering and separating ground algorithm based on terrain normalization processing;
step three: performing super-voxel clustering on the point cloud data in the second step in a three-dimensional space;
step four: clustering the super-voxel cluster blocks obtained in the step three according to the concave-convex relation among different voxel areas and the growth algorithm of the convex areas;
step five: and D, carrying out region growth on the clustering result obtained in the step four according to the point cloud color rule, clustering the segmented regions, and extracting the complete single wood.
The beneficial effects of the invention are as follows:
the invention provides an effective and feasible method for extracting forest site cloud data single wood, which selects a voxel with the minimum average curvature as a seed voxel, carries out region growing to form a super voxel, and overcomes the defect that the information color cannot be identified by the concave-convex segmentation through the combination of the concave-convex segmentation of the super voxel and the color characteristic region growing. Can effectively extract the adhered trees, reduce the treatment time and improve the segmentation effect.
Drawings
For ease of illustration, the invention is described in detail by the following detailed description and the accompanying drawings.
FIG. 1 is a flow chart of the invention;
FIG. 2 is a cloud of tree points after ground separation;
FIG. 3 is a woodland super voxel cluster map;
FIG. 4 is a super-voxel saliency segmentation map;
FIG. 5 is a color region growing cluster map;
fig. 6 is a single-wood dot cloud.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.
Examples:
a single wood segmentation method for super voxel concave-convex segmentation and color region growth of a forest land point cloud based on a foundation laser radar is realized as shown in fig. 1, and comprises the following operations:
acquiring forest land point cloud data by using a ground three-dimensional laser scanner, and importing the acquired forest land point cloud data into point cloud data processing software, and performing accurate multi-station registration on the point cloud data to acquire complete forest land point cloud data;
and (3) scanning five-station cloud data, importing FAROScene into each station data, performing point cloud splicing based on plane target assistance, importing target coordinates, extracting a main station target, splicing main stations by using a target-based splicing mode, and splicing other stations after acquiring main station information. And finally, splicing is completed to obtain complete forest land point cloud data.
Step two: extracting vegetation point clouds by adopting a point cloud filtering separation ground algorithm (CSF) based on terrain normalization processing for the complete woodland data point clouds obtained in the step one;
the point cloud ground point filtering adopts a cloth filtering algorithm, the principle of the filtering algorithm is that the point cloud is turned over firstly, then a piece of cloth is supposed to fall from the upper side under the gravity, and finally the falling cloth can represent the current terrain.
Step 2-1: (1) Firstly, removing isolated points of point cloud by utilizing LIDAR360 software and carrying out normalization processing on the point cloud data; (2) inverting the point cloud data; (3) setting the size of the cloth grid; (4) Projecting all point clouds and grid particles to the same horizontal plane, finding the nearest neighbor point of each particle, and calculating the position of the elevation 'cloth' particle before projection to be influenced by F ext (X, t) and F int Displacement due to the influence of two factors (X, t); (5) The LIDAR point and grid particle distance threshold value is set as a ground point, and the formula is as follows:
wherein X represents the position of the particle in the "cloth" at time t, F ext (X, t) represents an external driving factor (gravity, collision, etc.), F int (X, t) represents an internal driving factor (inter-particle internal linkage).
Step three: performing super-voxel clustering on the point cloud data in the second step in a three-dimensional space;
step 3-1: the method comprises the steps of firstly carrying out downsampling filtration on point clouds by using a PCL (PointCloudLibrary) library, finding out the point clouds falling in a unit voxel box, obtaining the barycenter of the points according to the coordinates of the points, inserting the barycenter points into the unit voxels, and realizing voxelization of the point clouds.
Step 3-2: constructing an adjacency graph through a 26 adjacency relation between two adjacent voxels, realizing rapid search of the adjacency relation in the voxels through a 3-dimensional k-d tree among the voxels, carrying out gridding treatment on the voxel point cloud, and selecting the point closest to the center as a seed point.
Step 3-3: after the seed point is selected, the center of the seed voxel and the connection neighborhood within the two voxels are found in the feature space to initialize the super-voxel feature vector.
Step 3-4: the similarity distance D between adjacent voxels is obtained through weighting the point cloud space position and the geometric relationship, and the formula is as follows:
wherein R is seed Represents setting reasonable seed voxel spacing, D C For the color of the point cloud, D n Is the normal vector, D S Represents distance, w c Is the color impact weight, w s Is the distance influencing weight, w n Is the normal vector impact weight.
Step four: and (3) clustering the different voxel blocks according to the convex-concave relation among the different voxel regions and the growing algorithm of the convex region by the super voxel clustering block obtained in the step (three).
Step 4-1: the relationship between the surface normals of the adjacent super voxels and the vector connecting the centroids thereof is judged, and the concave-convex relationship of the adjacent voxels is judged according to the method of the convexity standard extended by CC (ExtendedConvexityCriterion). Wherein the CC is divided into CC b And CC e 。CC b Defining a basic convexity connection setting a certain curvature threshold beta thresh And comparing the magnitude relation between the angle and the curvature threshold value between the super-voxel surface sheets, and determining a convex connection area. Because the point cloud data contains noise, CCe needs to be introduced as a judgment for expanding convexity in a non-ideal state, and more evidence is introduced to indicate that the connection is a convexity connection.
Step 4-2 SC (Sanitycriterion) is used as another criterion to determine the angle v between the cross product of the vector d connecting the centroid and the normal vector of the two patches, when SC is valid, the two patches are considered to be mostly connected, region growing can be performed, otherwise segmentation is performed.
Step 4-3: judging the convexity and concavity through CC and SC, marking the convexity and concavity area of each super-voxel, selecting any super-voxel as a seed point, marking the seed point, carrying out depth search in a neighborhood graph, allowing the seed point to spread only across the convex edge, carrying out iterative search without the seed point, finally realizing the clustering result of the seed super-voxel and surrounding super-voxels, and ending the concavity and convexity segmentation of the super-voxel.
Step five: and D, carrying out region growth on the clustering result obtained in the step four according to the point cloud color rule, clustering the segmented regions, and extracting the complete single wood.
Step 5-1: firstly, determining a good growth rule, setting the growth rule to be that the color difference between the current seed point and the field point is smaller than a certain threshold value, and adopting the following formula:
wherein CD is the direct color distance threshold value of two seed points, R 1 ,G 1 ,B 1 And R is 2 ,G 2 ,B 2 Respectively are points P 1 And P 2 RGB values of (a).
Step 5-2: obtaining color information of a clustering area, selecting an initial seed point with the lowest RGB value, constructing a neighborhood structure by using an octree, searching for points with similar RGB values by using indexes of a K-D tree, continuing searching if the similarity of the color values is met, updating the seed point if the similarity is not met, and finishing clustering after all seed voxels and adjacent voxels are fused.
Based on the forest land point cloud data single wood segmentation method based on the foundation laser radar, the invention provides an effective and feasible forest land point cloud data single wood extraction method, a voxel with the smallest average curvature is selected as a seed voxel, region growing is carried out to form an ultra-voxel, and the defect that the information color can not be identified by the ultra-voxel concave-convex segmentation is overcome by combining the ultra-voxel concave-convex segmentation with the color characteristic region growing, so that adhered trees can be effectively extracted, the processing time is shortened, and the single wood segmentation effect is improved.
Claims (4)
1. An ultra-voxel concave-convex segmentation and color region growth single wood segmentation algorithm based on forest land point cloud data of a foundation laser radar is characterized in that: the method comprises the following steps:
step one: firstly, acquiring original Lin Fendian cloud data in the field by using a foundation laser radar technology, importing each station of data into point cloud data processing software, and carrying out accurate multi-station registration on the point cloud data to obtain complete forest land point cloud data.
Step two: and (3) for the complete woodland data point cloud obtained in the step one, adopting a point cloud filtering and separating ground algorithm based on terrain normalization processing to separate the ground point Yun Yufei ground point cloud, and then cutting the woodland point cloud to obtain a sample region point cloud data map.
Step three: and (3) performing super-voxel clustering on the point cloud data in the second step in a three-dimensional space.
Step four: and D, obtaining the concave-convex relation among the clustered blocks in the step three. After the convexity is obtained, the region grows across the convex edge for clustering again, and a first clustering result is obtained.
Step five: and (3) carrying out region growth clustering based on the color information again on the clustering result in the step four, and realizing the segmentation of single trees.
2. The single-wood segmentation algorithm based on super-voxel saliency segmentation and color region growing as set forth in claim 1, wherein: in the first step, only the spatial characteristics and normal information of the point cloud are considered, so that the point cloud data can be directly processed in a three-dimensional space, the point cloud is converted into a plurality of small blocks, and the relation among the small blocks is studied. The voxel with the smallest average curvature is selected as a seed voxel, and the specific process of forming the super voxel by region growing is as follows:
step 2-1: the method comprises the steps of firstly carrying out downsampling filtration on point clouds by using a PCL (Point Cloud Library) library, finding out the point clouds falling in a unit voxel box, obtaining the barycenter of the points according to the coordinates of the points, inserting the barycenter points into the unit voxels, and realizing voxelization of the point clouds.
Step 2-2: the clustering between adjacent voxels is realized, an adjacency graph is firstly constructed through a 26 adjacency relation between two adjacent voxels, the rapid search of the adjacency relation in the voxels is realized through a 3-dimensional k-d tree between the voxels, the gridding treatment is carried out on the voxel point cloud, and the point closest to the center is selected as a seed point.
Step 2-3: after the seed point is selected, setting a reasonable seed voxel distance R_seed, searching according to the radius, calculating the adjacent voxels in the searching range according to similarity, obtaining a similarity distance D between the adjacent voxels through weighting the point cloud space position and the geometric relationship, and classifying the similar adjacent voxels into one category according to the following formula:
wherein D is the similarity distance between adjacent voxels, D c Is the color difference, D s Is the distance difference, dn is the normal vector difference, w c Is the color impact weight, w s Is the distance influencing weight, w n Is the normal vector impact weight.
3. The single-wood segmentation algorithm based on super-voxel saliency segmentation and color region growth of claim 2, wherein: the method for connecting local lugs by Locally Convex Connected Patches (LCCP) does not consider color information of point cloud, only uses space information and normal line information to traverse each super-voxel, judges the concave-convex relation of adjacent super-voxels through the information between the surface normal line of the super-voxels and the vector connecting the centroids of the super-voxels, and clusters the convex connection of voxel areas, and comprises the following specific steps:
step 3-1: firstly, judging the concave-convex relation of adjacent voxels, wherein the concave-convex relation has two judging criteria, namely a CC (Extended Convexity Criterion) extended convexity condition and SC (Sanity criterion) rational standard.
Step 3-2: after judging the concave-convex relation of the voxel region, clustering the convex regions according to the growth of the cross-convex regions, and marking the clustered regions.
4. A mono-wood segmentation algorithm based on super-voxel saliency segmentation and color region growing as claimed in claim 3, characterized in that: the super voxel clustering and the concave-convex segmentation only consider the space information and the normal characteristic of the point cloud, and the color information of the point cloud is not considered, so that the same organ of a tree is easy to be segmented in error. Therefore, the areas of the same class can be clustered again by combining the area growth of the colors, so that a complete tree is obtained.
Step 4-1: in the algorithm, input data are super voxel cluster data segmented according to concave-convex areas, each cluster area is marked, a cluster center is searched for in each cluster area according to indexes of K-D trees, an initial seed point with the lowest RGB value is selected, a growth criterion is set to be that the color difference between the current seed point and a field point is smaller than a certain threshold value, and the current seed point and the field point are classified as an area, wherein the formula is as follows:
wherein CD is the direct color distance threshold value of two seed points, R 1 ，G 1 ，B 1 And R is 2 ，G 2 ，B 2 Respectively are points P 1 And P 2 RGB values of (a).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210429377.3A CN116977593A (en) | 2022-04-22 | 2022-04-22 | Single wood segmentation method based on super-voxel concave-convex segmentation and color region growth |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210429377.3A CN116977593A (en) | 2022-04-22 | 2022-04-22 | Single wood segmentation method based on super-voxel concave-convex segmentation and color region growth |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116977593A true CN116977593A (en) | 2023-10-31 |
Family
ID=88483670
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210429377.3A Pending CN116977593A (en) | 2022-04-22 | 2022-04-22 | Single wood segmentation method based on super-voxel concave-convex segmentation and color region growth |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116977593A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117710717A (en) * | 2024-02-05 | 2024-03-15 | 法奥意威(苏州)机器人系统有限公司 | Super-body clustering point cloud segmentation method, device, equipment and storage medium |
-
2022
- 2022-04-22 CN CN202210429377.3A patent/CN116977593A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117710717A (en) * | 2024-02-05 | 2024-03-15 | 法奥意威(苏州)机器人系统有限公司 | Super-body clustering point cloud segmentation method, device, equipment and storage medium |
| CN117710717B (en) * | 2024-02-05 | 2024-05-28 | 法奥意威(苏州)机器人系统有限公司 | Super-body clustering point cloud segmentation method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110363861B (en) | Laser radar point cloud-based field crop three-dimensional reconstruction method | |
| CN110781827B (en) | Road edge detection system and method based on laser radar and fan-shaped space division | |
| CN106199557B (en) | A kind of airborne laser radar data vegetation extracting method | |
| CN109166145B (en) | Fruit tree leaf growth parameter extraction method and system based on cluster segmentation | |
| CN112070769B (en) | Layered point cloud segmentation method based on DBSCAN | |
| CN107369161A (en) | A kind of workpiece point cloud segmentation method at random based on the European cluster of improvement | |
| CN110992341A (en) | Segmentation-based airborne LiDAR point cloud building extraction method | |
| CN111340723B (en) | Terrain-adaptive airborne LiDAR point cloud regularization thin plate spline interpolation filtering method | |
| CN115205690B (en) | Method and device for extracting street tree in monomer mode based on MLS point cloud data | |
| CN109215112B (en) | Method for marking single-side point cloud model | |
| CN111145129A (en) | Point cloud denoising method based on hyper-voxels | |
| CN114119863A (en) | Method for automatically extracting street tree target and forest attribute thereof based on vehicle-mounted laser radar data | |
| CN114387288A (en) | Single standing tree three-dimensional information extraction method based on vehicle-mounted laser radar point cloud data | |
| CN116310849B (en) | Tree point cloud monomerization extraction method based on three-dimensional morphological characteristics | |
| CN115063555A (en) | Method for extracting vehicle-mounted LiDAR point cloud street tree growing in Gaussian distribution area | |
| CN115049925A (en) | Method for extracting field ridge, electronic device and storage medium | |
| CN110348478B (en) | Method for extracting trees in outdoor point cloud scene based on shape classification and combination | |
| CN116258857A (en) | Outdoor tree-oriented laser point cloud segmentation and extraction method | |
| CN116523898A (en) | Tobacco phenotype character extraction method based on three-dimensional point cloud | |
| CN116977593A (en) | Single wood segmentation method based on super-voxel concave-convex segmentation and color region growth | |
| CN106683105A (en) | Image segmentation method and image segmentation device | |
| CN112070787B (en) | Aviation three-dimensional point cloud plane segmentation method based on opponent reasoning theory | |
| CN113724400A (en) | Oblique photography-oriented multi-attribute fusion building point cloud extraction method | |
| CN116824379A (en) | Laser point cloud building contour progressive optimization method based on multidimensional features | |
| CN111860359A (en) | Point cloud classification method based on improved random forest algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |