Method of extracting number of stem and tillering for wheat under field condition

Patent metadata

Field	Value
Publication number	US-12154290-B2
Application number	US-202017441877-A
Country	US
Kind code	B2
Filing date	Mar 20, 2020
Priority date	Mar 22, 2019
Publication date	Nov 26, 2024
Grant date	Nov 26, 2024

Abstract

A field wheat stem tillering number extraction method, including: acquiring field wheat point clouds by means of a LiDAR, and extracting any row of wheat point clouds in a research area; projecting a Y axis to a plane, and retaining an X and Z axis; applying adaptive layering to obtain number of clusters of the wheat row; applying hierarchical clustering analysis to obtain tillering number of each wheat cluster; and further obtaining stem tillering number of the whole wheat row, so as to extract a field wheat stem tillering number. The feasibility of an algorithm is verified by comparing the wheat stem tillering number extracted by means of the method with an actually measured field stem tillering number, and the method realizes rapid, accurate and nondestructive extraction of a large-field crop stem tillering number and provides theoretical basis and technical support for extraction of the field wheat stem tillering number.

First claim

What is claimed is:

1. A field wheat stem tillering number extraction method, comprising the following steps:
step (1): acquiring a wheat stem tillering number sample and point clouds;
step (2): randomly acquiring a row of wheat point clouds from a wheat point cloud obtained in step (1), preprocessing the row of wheat point clouds, projecting a Y axis to a plane, and retaining an X axis and a Z axis to obtain a wheat two-dimensional side view;
step (3): applying adaptive layering to obtain a number of clusters of a row of wheat; and
step (4): applying hierarchical clustering analysis to obtain a tillering number of each cluster of wheat, and further obtaining a stem tillering number of the whole row of wheat.

2. The field wheat stem tillering number extraction method according to claim 1 , wherein in step (1), the wheat point clouds are acquired by using an RIEGL VZ-1000 three-dimensional laser scanner, and the wheat stem tillering number sample is synchronously acquired.

3. The field wheat stem tillering number extraction method according to claim 1 ,
wherein in step (2), the preprocessing the wheat point clouds comprises: after point cloud registration and denoising processing, randomly acquiring a row of wheat at first from a wheat point cloud of an area; expressing the row of wheat into a cuboid, defining a row length as X, a row width as Y and a plant height as Z; and performing three-dimensional sorting of point clouds in space by using an X coordinate axis as a main axis and a Z axis as an auxiliary axis, and projecting the three-dimensional sorted point clouds onto an X-Z plane to obtain the wheat two-dimensional side view.

4. The field wheat stem tillering number extraction method according to claim 1 ,
wherein in step (3), the adaptive layering comprises: distinguishing wheat point clouds of a leaf layer from wheat point clouds of a stem layer in a Z direction;
laminating the stem layer onto the leaf layer to integrate the stem layer and the leaf layer into one layer; sorting the X axis according to X coordinate values; determining that a front is a cluster of wheat when a mixture of the leaf layer and the stem layer is met, continuous point clouds of the leaf layer are met later, but no point cloud of the stem layer is met; and
determining the number of clusters of the row of wheat one by one by parity of reasoning.

5. The field wheat stem tillering number extraction method according to claim 4 , wherein the adaptive layering comprises:
a) dividing a row of wheat into segments at intervals in a direction of the X axis, calculating a plant height of each segment of wheat after segment division, and obtaining an average plant height H of the row of wheat;
b) performing layering according to the Z axis, wherein a number of layers n=average plant height H/tangent sheaf height h, and numbering the layers in a sequence of Arabic numerals from top to bottom in the direction of the Z axis;
c) marking the leaf layer and the stem layer: removing a lowest layer and a highest layer, and respectively marking point clouds of a top layer and a bottom layer of the remaining layers as the leaf layer and the stem layer;
d) laminating and integrating the stem layer and the leaf layer determined in step c) into one layer; sorting the X axis according to X coordinate values; determining that the front is a cluster of wheat when a mixture of the leaf layer and the stem layer is met, continuous point clouds of the leaf layer are met later, but no point cloud of the stem layer is met; and determining the number of clusters of the row of wheat one by one by parity of reasoning;
e) for remaining layers, sequentially repeating step c) and step d) for analysis in the direction of the Z axis by taking adjacent 2 layers as a combination from top to bottom, so as to obtain the number of clusters of the row of wheat, wherein if the number of clusters of the row of wheat obtained from a certain combination is greater than that of a previous combination, it shows that the layer with a greater number in the combination is the stem layer, the layer with a smaller number is the leaf layer, the combination is a stem and leaf connection combination; and marking layers of all leaf layers higher than the stem and leaf connection combination in the direction of the Z axis as the leaf layers, and marking layers of all stem layers lower than the stem and leaf connection combination as the stem layers; and
f) selecting one layer from the marked leaf layers, selecting one layer from the marked stem layers, and repeating step c) and step d) again to obtain the number of clusters of the row of wheat T n , wherein n=1, 2, 3, . . . , n.

6. The field wheat stem tillering number extraction method according to claim 5 , wherein the wheat is divided into segments at intervals according to a step length of 0.1 m in the direction of the X axis, and a tangent sheaf height h is 4 cm.

7. The field wheat stem tillering number extraction method according to claim 5 ,
wherein in step (4), the hierarchical clustering analysis comprises: in point clouds of each cluster of wheat, by taking each point as a variable and selecting a Euclidean distance as a similarity measurement, respectively calculating the similarity measurement between the variables, classifying points with a high similarity into one class, i.e., one tillering so as to calculate a number of classes to obtain a wheat tillering number, and calculating the tillering number T nk of each cluster of wheat, wherein k=1, 2, 3, . . . , k, and k represents the tillering number in each cluster of wheat; obtaining a stem tillering number of one row of wheat according to T sum =T 1 +T 2 +T 3 + . . . +T n +T 1k +T 2k +T 3k + . . . +T nk ; and finally, converting the stem tillering number of one row of wheat into a stem tillering number of the whole area.

8. The field wheat stem tillering number extraction method according to claim 7 ,
wherein a similarity measurement formula is:
d ij =√{square root over (( x i −x j ) 2 +( y i −y j ) 2 +( z i −z j ) 2 )}.

wherein i and j represent 2 arbitrary points, dij is a Euclidean distance between variables i and j, coordinate information of the variable i is (x i , y i , z i ), and coordinate information of the variable j is (x j , y j , z j ).

9. The field wheat stem tillering number extraction method according to claim 7 , wherein the hierarchical clustering analysis comprises: in point clouds of each cluster of wheat, by taking each point as a variable and selecting a Euclidean distance as a similarity measurement, respectively calculating the similarity measurement between the variables, finding clusters 1 and 2 with a shortest Euclidean distance, merging the clusters 1 and 2 to be used as a new class, calculating a space from the new class to other clusters in the current cluster, setting an iteration condition to be that the Euclidean distance d=2 cm, and stopping iteration when the condition that d>2 cm is met, so as to obtain the tillering number of each cluster of wheat.

10. The field wheat stem tillering number extraction method according to claim 1 , further comprising: verifying feasibility of an algorithm: verifying feasibility and accuracy of a test algorithm by using independent year stem tillering number sample data, and performing comprehensive evaluation on a result by using a determination coefficient R 2 , a root-mean-square error (RMSE) and a relative root-mean-square error (RRMSE).

Assignees

• Univ Nanjing Agricultural

Inventors

• Yao Xia
• Guo Tai
• Zhang Xiaohu
• Zhu Yan
• Zheng Hengbiao
• Cheng Tao
• Tian Yongchao
• Cao Weixin
• Guo Caili
• Zhang Yu
• Ma Jifeng
• Huang Rui
• Zhu Jie
• Ai Hongxu
• Jiang Chongya
• Zhou Dong

Classifications

• G06T5/70

  Denoising; Smoothing · CPC title

• G06F18/2321 Primary

  using statistics or function optimisation, e.g. modelling of probability density functions · CPC title

• G06T2207/30188

  Vegetation; Agriculture · CPC title

• G06T2207/10028

  Range image; Depth image; 3D point clouds · CPC title

• G06T7/0002

  Inspection of images, e.g. flaw detection · CPC title