CN110398465A - A method for measuring the biomass of cultured laver based on spectral remote sensing images - Google Patents

Abstract

The invention provides a method for measuring the biomass of cultured laver based on spectral remote sensing images. Based on the correlation between the reflectance spectrum of cultured seaweed and its biomass, the characteristic spectrum band and characteristic parameters are selected, and a multi-spectral remote sensing monitoring model of cultured laver is constructed. . According to the model, the real-time non-destructive monitoring of Porphyra biomass in a large area is obtained through the inversion of multispectral remote sensing images. The invention provides a new biomass measurement method for cultured economical seaweed, which has the advantages of real-time non-destructive, high-throughput, effective reduction of a large amount of field investigation workload, and the like. This method can be applied to the growth monitoring of cultured economical seaweed, and can provide a theoretical and experimental basis for the digitization of marine farming. Compared with the traditional ground sampling measurement method, the method of the present invention saves time and effort, can achieve non-destructive large-area measurement of the target area, and at the same time has higher precision than satellite remote sensing measurement, timely acquisition, and good timeliness of measurement.

Description

A method for measuring the biomass of cultured laver based on spectral remote sensing images

technical field

The invention belongs to the technical field of biomass remote sensing monitoring, in particular to a method for measuring the biomass of cultivated laver based on near-earth multispectral remote sensing images.

Background technique

Porphyra cultivation is concentrated in the intertidal zone and near-shore waters, and operations on the ground or sea are extremely susceptible to the complex and changeable marine environment. So far, the monitoring of its growth and growth is still lacking in high-precision, high-throughput and digitalization, which seriously affects The healthy and sustainable development of the aquaculture industry.

At present, the traditional sampling method is mostly used for the biomass measurement of cultured seaweed, which requires a large-scale collection of samples, which consumes a lot of labor and time. Due to the limitation of the marine environment, it is difficult to be timely and representative, and the sampling is lossy and direct. Affects the growth and yield of economic algae.

The ever-growing near-ground remote sensing technology has the advantages of fast, accurate, non-destructive and long-term, dynamic and continuous macroscopic monitoring. Now focus on two aspects:

One is the identification and division of breeding areas, the comparative characteristics and shapes of the application objects are obvious and the changes are small with time and space;

The second is the determination of chlorophyll, water and organic matter content or salinity of water or bottom substrate. The measurement objects are relatively uniform in spatial distribution and the variation range is relatively small.

However, most economic seaweeds in coastal waters are cultivated in water bodies, and they will be disturbed by sea water and currents in terms of remote sensing information acquisition and sea sample collection, which is not conducive to model construction.

Contents of the invention

The purpose of the present invention is to provide a fast, accurate, non-destructive and high-throughput method for measuring the biomass of cultured laver, making it suitable for long-term dynamic macroscopic field monitoring, and promoting the establishment and development of a new model of mariculture industry.

The biomass determination method of cultured laver of the present invention comprises the steps:

1) Obtain high-resolution multispectral near-earth remote sensing images of the target area:

In the target area to be measured, around noon on a sunny or cloudy day, when the cultured seaweed is completely exposed to the water surface, use a drone equipped with a high-resolution multispectral instrument and a motion camera to take round-trip shots at a height of 40m above the target area. 2m, to acquire multi-spectral images;

The high-resolution multispectral instrument is a high-resolution multispectral instrument RedEdge-M sensor; the motion camera is a Firefly 8s motion camera;

Acquisition of spectral reflectance calibration data: place the spectral reflectance calibration plate on the ground, use a drone equipped with a high-resolution multispectral instrument with a pixel size of 1280*960 pixels, a viewing angle of 47.2°, and a motion camera to take off and hover 2-5 meters above the ground. The spectral camera shoots 2-3 sets of 5-channel multi-spectral data on the calibration plate to obtain reflection calibration plate images for multi-spectral reflectance calibration;

The five channels are blue light (B, center wavelength 475nm, 20nm width), green light (G, center wavelength 560nm, 20nm width), red light (R, center wavelength 668nm, 10nm width), red edge (RE , center wavelength 717nm, 10nm width) and near infrared (NIR, center wavelength 840nm, 40nm width);

2) Remote sensing image data preprocessing:

Use the automatic aerial triangulation function of the Pix4D software to perform automatic aerial triangulation on the acquired high-resolution multispectral images, load the data that needs stitching and orthorectification, and load the reflection calibration panel image obtained in step 1) band by band, and input the reflection rate and automatically process the acquired multispectral images.

3) Execute the band spectrum calculation applet in the ENVI5.1 software on the processed multispectral image, select the near-infrared band ρ _NIR and the red light band ρ _R as the characteristic bands related to biomass, and use the band calculation applet to extract the Vegetation index; where ratio vegetation index RVI=ρ _NIR /ρ _R ; difference vegetation index DVI=ρ _NIR -ρ _R ; normalized difference vegetation index NDVI=(ρ _NIR -ρ _R )/(ρ _NIR +ρ _R );

4) The spectral vegetation index is substituted into the model to estimate the biomass. The regression model formula is as follows: Biomass(g/m ² )=235.30DVI+12.91RVI+8.90NDVI+8.52, to calculate the biomass of seaweed cultured in the target area.

Compared with the traditional ground sampling measurement method, the method of the present invention saves time and effort, can achieve non-destructive large-area measurement of the target area, and at the same time has higher precision than satellite remote sensing measurement, timely acquisition, and good timeliness of measurement.

Detailed ways

According to the spectral characteristics of seaweed, the present invention constructs a biomass spectral data model of cultured seaweed, and provides new technical support for further expanding the application of near-ground remote sensing in the seawater culture industry and establishing an ecological health assessment system for seaweed culture.

The present invention will be described in detail below in conjunction with the examples.

Example 1: Construction process of regression model

1) Acquisition of high-resolution multispectral images in the target area:

The target area is the laver breeding sea area near Fuxin Fishing Port, Lanshan District, Rizhao City, Shandong Province. The latitude and longitude are 119°26′1′E to 119°26′6′E and 35°15′23′N to 35°15N '25', the flight time was around noon on a clear day.

Before the flight, place the spectral reflectance calibration plate on the ground, and use the EcoDrone UAS-8 drone to carry the high-resolution multispectral RedEdge-M sensor with a pixel size of 1280*960 pixels and a viewing angle of 47.2° and the Firefly 8s motion camera to take off off the ground 2 -Hovering at 5 meters, shooting 5-channel multi-spectral data of 2-3 sets of spectral reflectance calibration plates for different channel reflectance calibration of multi-spectrum.

Among them, the five channels are blue light (B, center wavelength 475nm, 20nm width), green light (G, center wavelength 560nm, 20nm width), red light (R, center wavelength 668nm, 10nm width), red edge ( RE, center wavelength 717nm, 10nm width) and near infrared (NIR, center wavelength 840nm, 40nm width).

Using the EcoDrone UAS-8 unmanned aerial vehicle equipped with a high-resolution multispectral sensor RedEdge-M sensor and Firefly8s motion camera, within 1 hour when the cultivated laver is completely exposed to the water surface, the round-trip shooting is carried out at a height of 40m above the target area, and the distance between the routes is 2m. high-resolution multispectral imagery.

2) Remote sensing image data preprocessing:

Process the acquired high-resolution multispectral images of the target area with Pix4D software, use the software’s automatic aerial triangulation function, load the data that needs stitching and orthorectification, and load the reflection calibration panel images obtained in step 1) band by band, input Reflectance, automatic processing of acquired multispectral imagery.

3) Execute the band spectrum calculation applet in the ENVI5.1 software on the processed multispectral image, and select the special reflectance spectrum bands of laver according to the reflection spectrum characteristics of laver, including the near-infrared band ρ _NIR and the red light band ρ _R as biological Quantity-related characteristic bands, use the band operation applet to extract the vegetation index; among them, the ratio vegetation index RVI = ρ _NIR / ρ _R ; the difference vegetation index DVI = ρ _NIR - ρ _R ; the normalized difference vegetation index NDVI = (ρ _NIR -ρ _R )/(ρ _NIR +ρ _R ).

4) Acquisition of measured sea area biomass data:

Synchronized with UAV spectral remote sensing imaging flight measurement, a total of 27 aquaculture seaweed net curtains were randomly selected in the target aquaculture sea area, and 3 sample plots were taken for each net curtain, and each sample plot was 0.60m*0.60m. All sample quadrats were positioned by high-precision GPS. The biomass of each quadrat was obtained by removing all the seaweeds in the quadrat and drying them at 80°C to constant weight, and weighing them with a balance with a precision of 0.01g.

5) Biomass estimation model construction:

According to the combination of biomass data of 81 quadrats sampled in the field and one or more vegetation indices, a linear or nonlinear regression model of one or more vegetation indices was established. like:

Biomass=280.58DVI+21.92

Biomass=65.48RVI+104.28NDVI-45.11

Biomass=268.43DVI+17.14NDVI+21.18

Biomass＝420.75NDVI2 ⁺ 178.50NDVI+17.03

Biomass＝235.30DVI+12.91RVI+8.90NDVI+8.52

6) Model accuracy evaluation:

The accuracy of each model is evaluated by the estimated biomass and the measured biomass on the ground, such as root mean square (RMSE), precision (Ac) and degree of fit (R ² ) to evaluate the accuracy of the model and the relationship between estimated and measured values. similarity.

in Where X and Y are measured biomass and estimated biomass respectively, i is the corresponding sample plot, n is the total number of sample plots, is the average biomass of all sample plots.

Finally, the optimal model was determined as Biomass (g/m ² )=235.30DVI+12.91RVI+8.90NDVI+8.52, R ² was 0.93, RMSE was 5.67, and Ac was 82.36%.

Application of the model established in embodiment 2

Determination of Porphyra Biomass in Laver Breeding Area in Lanshan District, Rizhao City, Shandong Province

1) Acquisition of high-resolution multispectral near-Earth remote sensing images in the target area

The target area is the area surrounded by east longitude 119°25′12.5′ to east longitude 119°25′17.5′ and north latitude 35°15′12′ to north latitude 35°15′15.5′ in the laver breeding area of Lanshan District, Rizhao City, Shandong Province.

Around noon on a clear day, the EcoDrone UAS-8 UAV is equipped with a high-resolution multispectral RedEdge-M sensor with a pixel size of 1280*960 pixels and a viewing angle of 47.2° and a Firefly 8s action camera, hovering 2-5 meters above the ground, and shooting 5-channel multispectral data of 2-3 sets of spectral reflectance calibration boards; within 1 hour when the cultured seaweed is completely exposed to the water surface, round-trip shooting is carried out at a height of 40m above the target area, and the flight distance is 2m to obtain multispectral images.

2) Remote sensing image data processing

The acquired high-resolution multispectral images will be automatically triangulated using Pix4D software to load the data that needs to be stitched and orthorectified, and the reflection calibration panel images will be loaded band by band, and the reflectance will be input to automatically process the acquired multispectral images.

Use ENVI5.1 software to run the small program of spectral calculation in different bands to extract the vegetation index. Among them, ratio vegetation index RVI=ρ _NIR /ρ _R ; difference vegetation index DVI=ρ _NIR -ρ _R ; normalized difference vegetation index NDVI=(ρ _NIR -ρ _R )/(ρ _NIR +ρ _R ).

3) Acquisition of measured sea area biomass data:

Synchronized with UAV spectral imaging flight measurement, a total of 28 quadrats on the net curtain of cultured seaweed were randomly selected in the target area, each quadrat was 0.60m*0.60m, with high-precision GPS positioning. The biomass of each quadrat was obtained by taking all the seaweeds in the quadrat and drying them at 80°C to constant weight, and weighing them with a balance with a precision of 0.01g.

4) Comparison between the measured value and the estimated biomass of the spectral model:

In the spliced multi-spectral images, according to the coordinates of the GPS positioning points in the field sampling, the different vegetation indices corresponding to the 28 quadrats are obtained, and the above-mentioned optimal biomass estimation model Biomass (g/m ² )=235.30DVI+ 12.91RVI+8.90NDVI+8.52, estimated biomass per square meter: 27.16g, 72.87g, 66.82g, 159.63g, 96.70g, 51.57g, 212.28g, 7.39g, 81.01g, 128.96g, 60.48g , 3.35g, 26.80g, 207.22g, 76.03g, 51.57g, 50.09g, 129.96g, 107.51g, 80.20g, 68.22g, 55.32g, 89.69g, 39.11g, 24.612g, 17.492g, 23.90g, and 141.814 g;

The measured values are: 34.31g, 72.06g, 63.28g, 141.94g, 94.94g, 60.33g, 219.64g, 6.53g, 77.611g, 129.89g, 51.44g, 8.472g, 14.22g, 205.75g, 78.81g, 60.33g, 41.19g, 130.89g, 102.78g, 69.89g, 66.11g, 45.89g, 94.17g, 31.00g, 11.97g, 12.03g, 26.92g, and 151.44g.

The root mean square of the estimated value of the model and the measured value is 7.46, and the accuracy is 90.07%, which shows that the model established by the method of the present invention is practical and can be used for accurate determination of the biomass of cultured laver in the target area.

Claims (6)

1. a kind of cultivation of porphyra biomass estimation method based on spectral remote sensing image, which is characterized in that the method includes Following step:

1) target area high-resolution multi-spectral near-earth remote sensing images are obtained:

In target area to be determined, in it is sunny or it is partly cloudy around noon, be completely exposed the water surface to cultivating seaweed, utilize nothing Man-machine carrying high-resolution multi-spectral instrument and the moving camera height at the height 40m of target area are shot back and forth, are navigated Line spacing is 2m, obtains multispectral image；

Spectral reflectance calibration data obtains: spectral reflectance calibration plate being placed in ground, is 1280* using UAV flight's pixel 960pixels, the high-resolution multi-spectral instrument and moving camera at 47.2 ° of visual angle depart 2-5 meters of hoverings, multispectral camera pair Calibration plate shoots 5 channel multispectral data of 2-3 group, obtains reflection calibration plate image to calibrate for multispectral reflectivity；

2) remote sensing image data pre-processes:

The high-resolution multi-spectral image that will acquire utilizes the automatic triangulation function of Pix4D software, carries out automatic empty Intermediate cam measurement, load need to splice and just penetrate the data of correction, and by wave band load step 1) reflection that obtains calibrates panel Image data, input reflection rate automatically process the multispectral image of acquisition；

3) by treated, multispectral image executes band spectrum operation small routine in ENVI5.1 software, chooses near-infrared wave Section ρ_NIR, red side wave section ρ_REWith red spectral band ρ_RFor the relevant characteristic wave bands of biomass, extracted wherein using band math small routine Vegetation index；Wherein ratio vegetation index RVI=ρ_NIR/ρ_R；Difference vegetation index DVI=ρ_NIR-ρ_R；Normalized differential vegetation index NDVI=(ρ_NIR-ρ_R)/(ρ_NIR+ρ_R)；

4) spectral vegetation indexes substitute into model, estimate biomass, wherein regression model formula is as follows: Biomass (g/m²)= 235.30DVI+12.91RVI+8.90NDVI+8.52 calculates the biomass of target area cultivating seaweed.

2. the method as described in claim 1, which is characterized in that described 1) in high-resolution multi-spectral instrument be high-resolution Multispectral instrument RedEdge-M sensor.

3. the method as described in claim 1, which is characterized in that described 1) in moving camera be Firefly 8s move phase Machine.

4. the method as described in claim 1, which is characterized in that described 1) in Five-channel multispectral data be respectively indigo plant Light, central wavelength 475nm, 20nm wave are wide；Green light, central wavelength 560nm, 20nm wave are wide；Feux rouges, central wavelength 668nm, 10nm Wave is wide；Red side, central wavelength 717nm, 10nm wave is wide and near-infrared, central wavelength 840nm, 40nm wave are wide.

5. a kind of for calculating the model of cultivation of porphyra biomass, which is characterized in that the model is described in claim 1 Method establish.

6. model as claimed in claim 5, which is characterized in that the formula of the model is as follows: Biomass (g/m²)= 235.30DVI+12.91RVI+8.90NDVI+8.52；

Wherein ratio vegetation index RVI=ρ_NIR/ρ_R；Difference vegetation index DVI=ρ_NIR-ρ_R；Normalized differential vegetation index NDVI= (ρ_NIR-ρ_R)/(ρ_NIR+ρ_R)；

Infrared band ρ_NIR, red side wave section ρ_REWith red spectral band ρ_RFor the relevant characteristic wave bands of biomass.