Domestic Plant Phenotype Analysis System

PhenoWAtch high-throughput plant phenotype analysis system

based onSensor to PlantA mobile high-precision plant phenotype imaging system designed with concept

The PhnoWatch high-throughput plant phenotype analysis system deeply integrates laser radar, hyperspectral imaging, infrared thermal imaging, multispectral imaging, and RGB imaging units. It can intelligently move to different measurement areas, automatically scan crops according to preset values, generate three-dimensional images containing multispectral information, identify individual plants in a population, separate stems and leaves of individual plants, accurately obtain phenotype parameters such as plant height, plant width, leaf length, leaf width, leaf inclination angle, leaf area, canopy closure, and canopy transmittance. Combined with spectral characteristics, it calculates vegetation indices and statistically analyzes canopy temperature and biomass.

1. PhenoWatchhardware system

The hardware of PhenoWatch high-throughput plant phenotype analysis system is mainly divided into 3DImaging unit, greenhouse or field mobile platform.

3D imaging unit (Sensor Box)

At the bottomPoint cloud module: LiDAR, equipped with dual axis compensator and height sensor, as well as angle measurement function, performs horizontal calibration for each scan, on-site automatic equipment compensation, higher accuracy and farther distance scanning, obtains plant spatial point cloud, 3D modeling, and ultimately extracts plant population parameters and individual plant morphological phenotype parameters.

At the bottomInfrared thermal imaging module: an external thermal imager used for machine vision, equipped with an uncooled vanadium oxide infrared detector, high-precision infrared thermal imaging CCD, capable of generating 640x480 pixel thermal images, withhigherVisual display of linear ROI temperature values and temperature profiles based on image quality.

At the bottomHyperspectral imaging module: Each pixel in the image records the spectral characteristics of the chemical composition, quality, color, and other information of its corresponding sample point, which is used for qualitative and quantitative analysis of plant biomass.

At the bottomMultispectral imaging module: Using five channel (Blue, Green, Red, NIR, RedEdge) spectral images as data sources, spectral information is assigned to three-dimensional point clouds through matching and fusion of images and point clouds, ultimately achieving three-dimensional vegetation index calculation.

At the bottomRGB imaging module: a high-resolution RGB camera that matches and fuses color images with point clouds to ensure the restoration of the true colors of plants while obtaining high-precision 3D images.

mobile platform

According to the actual on-site environment, the size and structure of the mobile platform can be flexibly designed:

lMulti scenario applicability, can be designed in combination with existing greenhouse structures and plant cultivation racks, can be used in greenhouses or in the wild;

lThe size customization degree of the mobile platform is high, and it can be flexibly designed according to installation conditions. The span can be greater than 10 meters, the height can be greater than 5 meters, and the length of the guide rail can be greater than 1000 meters;

lHigh degree of automation, defining the movement direction of the large vehicle on the guide rail as the X-axis, the movement axis of the Sensor Box on the crossbeam as the Y-axis, and the adjustment axis of the Sensor Box in the vertical direction as the Z-axis. It can achieve automatic control of the X, Y, and Z axes, and can move according to spatial coordinates and set distances;

lThe coverage area is wide, and it can be designed with a large span or a multi span design, that is, a gantry can be used to scan across multiple ridges of land;

lThere are various ways, including portal structure design, overhead crane structure design, or trolley structure design;

lHigh positioning accuracy, multiple limit protection, and high safety.

PhenoWatch GF Field Longmen style Plant Phenotype Imaging System

PhenoWatch GH Greenhouse Crane type Plant Phenotype Imaging System

PhenoWatch MB greenhouse gantry style plant phenotype imaging system

PhenoWatch MB Field Mobile Plant Phenotype Imaging System

PhenoWatch Plant Culture Shelf Version Plant Phenotype Imaging System

2. PhenoWatch software system

The 3D point cloud data and image information collected by the PhenoWatch system can be fused and modeled using PhenoWatch software to extract data from crop populations, obtaining population parameters such as canopy closure, canopy transmittance, and vegetation index; Based on the deep learning algorithm Faster RCNN model, the root position of individual crops is identified, and the traditional growth method is used to divide the population into individual plants. Then, the stem and leaf segmentation of plants is performed. There are two embedded stem and leaf segmentation methods in the software: region growth based stem and leaf segmentation and voxel based stem and leaf segmentation.

Based on the growth method, stem and leaf segmentation is based on the region growth method to identify and segment the stems and different leaves of individual plants. Based on the voxel method, stem and leaf segmentation is based on deep learning to identify and segment the stems and different leaves of individual plants. Then, independent point cloud files are generated for different leaves and stems, and surface and skeleton lines are fitted to the leaves and stems to calculate phenotypic parameters such as plant height, plant width, leaf length, leaf width, leaf inclination angle, and leaf area.

PhenoWatch software is a specialized software system for extracting three-dimensional phenotype parameters of crops,newThe PhenoWatch software version integrates neural network technology and deep learning algorithms, greatly improving the accuracy of crop single plant segmentation and stem leaf segmentation processing. This software uses parallel processing and GPU acceleration to further improve the speed of processing massive point cloud data. The software's single plant segmentation and stem leaf segmentation algorithms for crops meet the needs of crop genotype phenotype researchers for extracting crop morphological parameters at different scales. In addition, we also provide development services for customized data processing modules.

PhenoWatch analysis software processes point cloud data:

3D to 2D Projection and Curve Fitting Algorithm for Plant Skeleton

Analysis and Extraction of Plant Skeleton Lines

3. Measurement parameters

At the bottomGroup parameters:

lDigital Elevation Model (DEM): The digital representation of terrain surface morphology.

lDigital Surface Model (DSM): refers to a ground elevation model that includes the heights of surface buildings, bridges, and vegetation.

lThe canopy height model CHM: Subtracting the digital elevation model from the digital surface model yields the canopy height model.

lCanopy Cover: The vertical projection of crops as a percentage of the field area.

lCanopy transmittance: refers to the proportion of incident sunlight that can be received on different height layers of crops.

l3D point cloud vegetation index:

3D NDVI、3D TVI、3D RVI、3D DVI 等。

l2D vegetation index:

2D NDVI、2D TVI、2D RVI、2D DVI 等。

lVegetation index statistics: statistics of multispectral vegetation indices, such as meanbigValue, minimum value, etc

lGreen degree of plants: Custom threshold to reflect the green degree of plants

lThermal infrared analysis: temperature of each pixel in the image, plant canopy temperature (equipped with thermal infrared imaging unit)

lReport on population phenotype parameters, including height, crown width, projected area, crown height ratio, and volume of each plant, as well as data statistics:

At the bottomPhenotypic parameters of individual plant:

lHeightlcrown spread lprojected area lCrown height ratio

lVolume lNumber of leaveslTotal leaf area lLeaf Length

lYe Kuan lLeaf inclination angle lLeaf area lMain trunk width

lExport single plant phenotype parameter report:

4. Powerful and scalable features

The system can expand and select imaging modules according to different research needs to achieve more functions

Optional module1Plant multispectral camera

The frequency range of 400-900nm, covering 5 bands including red, green, blue, near-infrared, and red edge, can be used to calculate multiple vegetation indices such as NDVI, meeting the application of crop multispectral research.

The multispectral module can obtain chlorophyll distribution NDVI、 Digital surface model, RGB image

Multi spectral vegetation index NDVI imaging analysis Multi spectral vegetation index DVI imaging analysis Multi spectral vegetation index SR imaging analysis

Optional module2: Infrared thermal imaging module

The thermal infrared imaging module is equipped with an uncooled vanadium oxide (VoX) infrared detector, which can generate 640 x 480 pixel thermal images, making thermal imaging more accurate; Equipped with high-speed infrared window option; Clearly display a temperature difference of 50 mk; Built in 25 ° lens with electric focus and autofocus. The analysis software has the functions of point temperature measurement, line temperature measurement, elliptical area temperature measurement, rectangular area temperature measurement, as well as temperature maximum value, average value, Delta function statistics. At the same time, it records information such as emissivity, atmospheric temperature, external optical temperature, relative humidity, etc., in order to obtain useful information from thermal imaging images.

Infrared thermography fixed-point analysis Infrared thermography isothermal analysis

Analysis of Infrared Thermal Imaging Data

Optional module3: Hyperspectral imaging module

Configure a hyperspectral imaging module, which combines visible near-infrared (VNIR or NIR) spectroscopy with high-resolution imaging. The hyperspectral imaging system uses push room imaging technology to collect line by line full band spectra of moving or stationary samples and synchronously generate images to obtain quantitative data on the chemical composition and spatial distribution of the samples. Each pixel in the image records the spectral characteristics of its corresponding sample point, including chemical composition, mass, color, etc., for qualitative and quantitative analysis of the samples.

Hyperspectral imaging and hyperspectral curve analysis

Hyperspectral monitoring of wheat leaf blight disease

5. Main technical parameters:

6. Application Cases

Lidar measurement results of maize plant population

Identification of individual maize plants within a group

Identification and measurement of individual leaves within a plant

Crop canopy coverage measurement and distribution imaging

Two dimensional imaging and data display of crop vegetation index

3D false color view of crop vegetation index

Origin: OLAN Corporation, Israel