CN104634943A - Method for measuring salt content of saline and alkaline soil on line - Google Patents

Abstract

The invention relates to a method for measuring the salt content of saline and alkaline soil on line and aims at solving the problems that the sampling operation is time/labor-consuming and the online measurement precision of the salt content is low in a conventional method for measuring the salt content of saline and alkaline soil on line. The online measuring method comprises the following steps: step 1, acquiring soil cracking sample photos by virtue of a soil cracking experiment, extracting textural features, measuring the salt content in soil, and establishing a soil salt content inversion model; 2, measuring the salt content of field soil on line, calculating a normalized grayscale co-occurrence matrix of the field soil and the textural feature value with the best corresponding fitting relationship, thereby obtaining the preliminary online measurement result of the soil salt content; and step 3, correcting the measurement result of the salt content of the soil according to the maximum value and minimum value of the salt content of the soil sample. According to the method disclosed by the invention, based on the influence of the salt content of the saline and alkaline soil on the soil dry-shrinkage cracking degree, a quantitative model for the textural feature quantity of the cracking sample and the sample salt content is established, and the measurement accuracy is improved.

Description

A kind of On-line Measuring Method of salinized soil salt content

Technical field

The present invention relates to a kind of On-line Measuring Method of salinized soil salt content, particularly relate to the foundation of indoor salinized soil salt content inverse model and realized the On-line Measuring Method of soil salt content by on-line measurement saline-alkali soil textural characteristics.

Background technology

Salting of soil is the land deterioration process that becomes international, its havoc soil property and reduce the output of crops to a great extent, also has larger harm to ecologic environment simultaneously.Therefore, soil salt content level is measured fast and accurately for determining salting of soil degree, instructing the improvement of salinized soil to have important practical significance.At present, traditional salt content measuring method is determined mainly for the total content of each salt ion in field sampling, Indoor measurement soil, the measuring method of this soil salt content wastes time and energy and can not the field real conditions of real time reaction salt content, although remote sensing has the real-time feature of large area, but spatial resolution and the spectral resolution of remotely-sensed data itself are limited, make its determination being confined to large-area salt content degree and drawing application, and the single-point salt content measuring accuracy of small scale is very poor.The soda type saline-alkali soil of song-Nen plain is a kind of typical salinized soil, and its soil clay particle content is higher has very strong dilation characteristic.Therefore, after precipitation, the dry shrinkage and cracking phenomenon that soil surface occurs with the evaporation of moisture is very general.So far, existing a lot of research pays close attention to the identification of soil crack and measurement and soil physical chemistry parameter, environmental baseline etc. to aspects such as the impacts of cracking degree.Salt content, as the main chemical characteristic of salinized soil, also has the dehiscence process of soil and crack and affects significantly.

Summary of the invention

The measuring method sampling that the object of the invention is to solve existing salinized soil salt content is wasted time and energy, the problem that the on-line measurement precision of salt content is lower, and provides a kind of On-line Measuring Method of salinized soil salt content.

The On-line Measuring Method of salinized soil salt content of the present invention follows these steps to realize:

One, soil salt content inverse model is set up

1, soil cracking experiment

Select many groups soil sample of different salt content, carry out air-dry to all soil samples, milled processed also crosses the sieve in 2mm aperture, then the soil sample after every group being sieved is divided into soil sample A and soil sample B, soil sample A is wherein carried out dry shrinkage and cracking experiment, described dry shrinkage and cracking experiment first all soil sample A is configured to the identical saturated mud of water cut respectively, saturated mud good for every group configuration is poured into respectively in the wooden sample box of 50 cm x 50 cm x, 3 centimetres of sizes, place after floating mud and carry out natural drying in the lab to produce cup shake, until when in sample box, the quality of sample no longer reduces, obtain the soil sample A of season cracking,

Digital camera is arranged on fixing experiment porch, digital camera distance floor level is 1m, with the digital camera lens center that is projected as on the ground determine the square area of 50 cm x 50 centimetres with ensure the soil sample A of all season crackings take pictures process geometric distortion impact identical, the black and white grid calibration plate of 50 cm x 50 centimetres is covered the soil sample A surface of season cracking, then use the soil sample A of digital camera to season cracking to take pictures, obtain cracking soil sample photo;

2, texture feature extraction

Split soil sample photo carry out geometric distortion correction to often opening, the colored slit region image obtaining sample is sheared to the sample image after geometric distortion correction, then gray-scale map is converted into by often opening colored slit region image, again according to grey level histogram selected threshold and then carry out binary conversion treatment, then the crack image of each cracking soil sample is divided into uncracked piece of district (representing with black picture dot) and constriction zone (comprise the constriction of surrounding in fragmented parts and sample box, represent with white picture dot) after inversion operation;

Calculate the gray level co-occurrence matrixes of each cracking soil sample binary image, 40 will be set to apart from step-length, 4 directions are respectively 0 °, 45 °, 90 ° and 135 °, then the Normalized Grey Level co-occurrence matrix obtained by four direction carries out arithmetic mean thus with the impact of removing direction, calculate each Statistic Texture value according to each cracking soil sample gray level co-occurrence matrixes, the computing formula of each Statistic Texture value is as follows:

1) contrast:

$\mathrm{contra}=\underset{n}{\overset{\mathrm{Ng}-1}{\mathrm{\Σ}}}{n}^2\left\{\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\right\},|i-j|=n,---\left(1\right)$

2) angle second moment:

$\mathrm{ASM}=\underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}{\left\{p(i,j)\right\}}^2,---\left(2\right)$

3) entropy:

$\mathrm{Entropy}=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left(p(i,j)\right),---\left(3\right)$

4) consistance:

$\mathrm{Homo}=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\frac{1}{1+{(i-j)}^2}p(i,j),---\left(4\right)$

5) correlativity:

$\mathrm{Cerrelation}=\frac{\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\left(\mathrm{ij}\right)p(i,j)-{\mathrm{\μ}}_x{\mathrm{\μ}}_y}{{\mathrm{\σ}}_x{\mathrm{\σ}}_y},---\left(5\right)$

6) cluster is shady:

$\mathrm{Clustershade}=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}{((i-{\mathrm{\μ}}_i)+(j-{\mathrm{\μ}}_j))}^{3}p(i,j),---\left(6\right)$

7) maximum probability:

Maxprobability＝max{p(i,j)}， (7)

8) and average:

$\mathrm{SumAverage}=\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{\mathrm{ip}}_{x+y}\left(i\right),---\left(8\right)$

P _x+yi () represents that the probable value of gray level i appears in x capable and y row simultaneously;

9) and entropy:

$\mathrm{SumEntropy}=-\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{p}_{x+y}\left(i\right)\log \left\{p_{x+y}\left(i\right)\right\},---\left(9\right)$

10) and variance:

$\mathrm{SumVariance}=\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{(i-\mathrm{SumEntropy})}^2{p}_{x+y}\left(i\right),---\left(10\right)$

11) relevant information feature 1:

$\mathrm{Inforofcorrelation}1=\frac{\mathrm{HXY}-\mathrm{HXY}1}{\max (\mathrm{HX},\mathrm{HY})},---\left(11\right)$

12) relevant information feature 2:

Inforofcorrelation2＝{1-exp[-2*(HXY2-HXY)]} ^1/2， (12)

Wherein:

P (i, j) is Normalized Grey Level co-occurrence matrix (i, j) position element value, i.e. the probable value that simultaneously occurs of image gray levels i, j, and Ng is co-occurrence matrix number of greyscale levels, and i, j are positive integer,

$p_x\left(i\right)=\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),---\left(13\right)$

$p_y\left(j\right)=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),---\left(14\right)$

$p_{x-y}\left(k\right)=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),i+j=k,k=\mathrm{2,3},....2\mathrm{Ng},---\left(15\right)$

$\mathrm{HXY}=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left\{p(i,j)\right\},---\left(16\right)$

$\mathrm{HXY}1=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left\{p_x\left(i\right)p_y\left(j\right)\right\},---\left(17\right)$

$\mathrm{HXY}2=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}{p}_x\left(i\right)p_y\left(j\right)\log \left\{p_x\left(i\right)p_y\left(j\right)\right\},---\left(18\right)$

μ _x, μ _yfor p _x(i), p _y(j) average, σ _x, σ _yfor p _x(i), p _y(j) variance, n is gray level;

3, soil salt is measured

Each soil sample B is configured to soil extraction, measures the Ca in soil sample B respectively ²⁺, Mg ²⁺, Cl ^-, CO ^2-and HCO ₃ ^-content, calculates Ca in each soil sample B ²⁺, Mg ²⁺, Cl ^-, CO ^2-and HCO ₃ ^-the summation of content, is the salt content of soil sample B;

4, inverse model is set up

The salt content of soil sample B and each Statistic Texture value of calculating is utilized to set up linear model, horizontal ordinate (independent variable) wherein in linear model is texture characteristic amount, ordinate (dependent variable) is salt content, select the texture characteristic amount best with soil sample B salt content fit correlation, set up the normal linearity inverse model of soil salt content field on-line measurement;

Two, on-line measurement field soil salt content

For the on-line measurement of field to be measured salinized soil, digital camera is placed on a mobile platform, adjustment podium level makes digital camera height perpendicular to the ground be 1m, it is identical that digital camera parameter and soil parameter when testing that ftractures is set, then the cracking salinized soil in field is taken pictures, geometric distortion correction is carried out to the cracking salinized soil photograph in field, to take pictures corresponding 50cm*50cm area size and shearing field, position photo according to soil sample A, thus obtain standard crack photo, subsequently gray-scale map conversion and binary conversion treatment are carried out to standard crack photo, calculate Normalized Grey Level co-occurrence matrix and the best textural characteristics value of corresponding fit correlation, the result of calculation of textural characteristics value best for fit correlation is brought in normal linearity inverse model, thus realize the preliminary on-line measurement of salinized soil salt content,

Three, soil salt content measurement result is revised

In all on-line measurement field soil samples, select at soil sample min corresponding to the textural characteristics minimum value of line computation with at soil sample max corresponding to the textural characteristics maximal value of line computation, accurately measure the salt content of soil sample min and soil sample max respectively, if soil sample min salt content on-line measurement result is C1, soil sample max salt content on-line measurement result is C2, the accurate results of soil sample min salt content is D1, the accurate results of soil sample max salt content is D2, utilize (C1, D1), (C2, D2) linear scaled model is set up, then according to linear scaled model, all on-line measurement results are revised, thus complete the on-line measurement of soil salt content.

First soil salt content On-line Measuring Method of the present invention uses digital camera to obtain the crack image of soil sample, calculate the gray level co-occurrence matrixes data of sample crack image, the textural characteristics value of sample crack image is calculated according to gray level co-occurrence matrixes, then the salt content of soil sample is measured respectively, thus foundation is horizontal ordinate with eigenwert, take salt content as the linear relationship of ordinate, then obtain its textural characteristics value by the soil sample image in field, obtain the salt content of soil to be measured according to linear relationship.

The On-line Measuring Method of salinized soil salt content of the present invention is based on the impact of saline-alkali soil salt content on soil drying shrinkage cracking degree, first soil split test is carried out to the soil of different salt content, set up the cracking texture characteristic amount of sample and the quantitative model of sample salt content, and then the on-line measurement realized soil salt content under the natural conditions of field, measurement result precision R ²reach more than 0.85, show that measurement result can provide comparatively in real time for the improvement in alkaline land, Data support accurately.

Embodiment

Embodiment one: the On-line Measuring Method of present embodiment salinized soil salt content follows these steps to implement:

One, soil salt content inverse model is set up

1, soil cracking experiment

Select many groups soil sample of different salt content, carry out air-dry to all soil samples, milled processed also crosses the sieve in 2mm aperture, then the soil sample after every group being sieved is divided into soil sample A and soil sample B, soil sample A is wherein carried out dry shrinkage and cracking experiment, described dry shrinkage and cracking experiment first all soil sample A is configured to the identical saturated mud of water cut respectively, saturated mud good for every group configuration is poured into respectively in the wooden sample box of 50 cm x 50 cm x, 3 centimetres of sizes, place after floating mud and carry out natural drying in the lab to produce cup shake, until when in sample box, the quality of sample no longer reduces, obtain the soil sample A of season cracking,

Digital camera is arranged on fixing experiment porch, digital camera distance floor level is 1m, with the digital camera lens center that is projected as on the ground determine the square area of 50 cm x 50 centimetres with ensure the soil sample A of all season crackings take pictures process geometric distortion impact identical, the black and white grid calibration plate of 50 cm x 50 centimetres is covered the soil sample A surface of season cracking, then use the soil sample A of digital camera to season cracking to take pictures, obtain cracking soil sample photo;

2, texture feature extraction

Split soil sample photo carry out geometric distortion correction to often opening, the colored slit region image obtaining sample is sheared to the sample image after geometric distortion correction, then gray-scale map is converted into by often opening colored slit region image, again according to grey level histogram selected threshold and then carry out binary conversion treatment, then the crack image of each cracking soil sample is divided into uncracked piece of district (representing with black picture dot) and constriction zone (comprise the constriction of surrounding in fragmented parts and sample box, represent with white picture dot) after inversion operation;

Calculate the gray level co-occurrence matrixes of each cracking soil sample binary image, 40 will be set to apart from step-length, 4 directions are respectively 0 °, 45 °, 90 ° and 135 °, then the Normalized Grey Level co-occurrence matrix obtained by four direction carries out arithmetic mean thus with the impact of removing direction, calculate each Statistic Texture value according to each cracking soil sample gray level co-occurrence matrixes, the computing formula of each Statistic Texture value is as follows:

1) contrast:

$\mathrm{contra}=\underset{n}{\overset{\mathrm{Ng}-1}{\mathrm{\Σ}}}{n}^2\left\{\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\right\},|i-j|=n,---\left(1\right)$

2) angle second moment:

$\mathrm{ASM}=\underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}{\left\{p(i,j)\right\}}^2,---\left(2\right)$

3) entropy:

$\mathrm{Entropy}=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left(p(i,j)\right),---\left(3\right)$

4) consistance:

$\mathrm{Homo}=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\frac{1}{1+{(i-j)}^2}p(i,j),---\left(4\right)$

5) correlativity:

$\mathrm{Cerrelation}=\frac{\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\left(\mathrm{ij}\right)p(i,j)-{\mathrm{\μ}}_x{\mathrm{\μ}}_y}{{\mathrm{\σ}}_x{\mathrm{\σ}}_y},---\left(5\right)$

6) cluster is shady:

$\mathrm{Clustershade}=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}{((i-{\mathrm{\μ}}_i)+(j-{\mathrm{\μ}}_j))}^{3}p(i,j),---\left(6\right)$

7) maximum probability:

Maxprobability＝max{p(i,j)}， (7)

8) and average:

$\mathrm{SumAverage}=\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{\mathrm{ip}}_{x+y}\left(i\right),---\left(8\right)$

P _x+yi () represents that the probable value of gray level i appears in x capable and y row simultaneously;

9) and entropy:

$\mathrm{SumEntropy}=-\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{p}_{x+y}\left(i\right)\log \left\{p_{x+y}\left(i\right)\right\},---\left(9\right)$

10) and variance:

$\mathrm{SumVariance}=\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{(i-\mathrm{SumEntropy})}^2{p}_{x+y}\left(i\right),---\left(10\right)$

11) relevant information feature 1:

$\mathrm{Inforofcorrelation}1=\frac{\mathrm{HXY}-\mathrm{HXY}1}{\max (\mathrm{HX},\mathrm{HY})},---\left(11\right)$

12) relevant information feature 2:

Inforofcorrelation2＝{1-exp[-2*(HXY2-HXY)]} ^1/2， (12)

Wherein:

P (i, j) is Normalized Grey Level co-occurrence matrix (i, j) position element value, i.e. the probable value that simultaneously occurs of image gray levels i, j, and Ng is co-occurrence matrix number of greyscale levels, and i, j are positive integer,

$p_x\left(i\right)=\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),---\left(13\right)$

$p_y\left(j\right)=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),---\left(14\right)$

$p_{x-y}\left(k\right)=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),i+j=k,k=\mathrm{2,3},....2\mathrm{Ng},---\left(15\right)$

$\mathrm{HXY}=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left\{p(i,j)\right\},---\left(16\right)$

$\mathrm{HXY}1=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left\{p_x\left(i\right)p_y\left(j\right)\right\},---\left(17\right)$

$\mathrm{HXY}2=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}{p}_x\left(i\right)p_y\left(j\right)\log \left\{p_x\left(i\right)p_y\left(j\right)\right\},---\left(18\right)$

μ _x, μ _yfor p _x(i), p _y(j) average, σ _x, σ _yfor p _x(i), p _y(j) variance, n is gray level;

3, soil salt is measured

Each soil sample B is configured to soil extraction, measures the Ca in soil sample B respectively ²⁺, Mg ²⁺, Cl ^-, CO ^2-and HCO ₃ ^-content, calculates Ca in each soil sample B ²⁺, Mg ²⁺, Cl ^-, CO ^2-and HCO ₃ ^-the summation of content, is the salt content of soil sample B;

4, inverse model is set up

The salt content of soil sample B and each Statistic Texture value of calculating is utilized to set up linear model, horizontal ordinate (independent variable) wherein in linear model is texture characteristic amount, ordinate (dependent variable) is salt content, select the texture characteristic amount best with soil sample B salt content fit correlation, set up the normal linearity inverse model of soil salt content field on-line measurement;

Two, on-line measurement field soil salt content

For the on-line measurement of field to be measured salinized soil, digital camera is placed on a mobile platform, adjustment podium level makes digital camera height perpendicular to the ground be 1m, it is identical that digital camera parameter and soil parameter when testing that ftractures is set, then the cracking salinized soil in field is taken pictures, geometric distortion correction is carried out to the cracking salinized soil photograph in field, to take pictures corresponding 50cm*50cm area size and shearing field, position photo according to soil sample A, thus obtain standard crack photo, subsequently gray-scale map conversion and binary conversion treatment are carried out to standard crack photo, calculate Normalized Grey Level co-occurrence matrix and the best textural characteristics value of corresponding fit correlation, the result of calculation of textural characteristics value best for fit correlation is brought in normal linearity inverse model, thus realize the preliminary on-line measurement of salinized soil salt content,

Three, soil salt content measurement result is revised

In all on-line measurement field soil samples, select at soil sample min corresponding to the textural characteristics minimum value of line computation with at soil sample max corresponding to the textural characteristics maximal value of line computation, accurately measure the salt content of soil sample min and soil sample max respectively, if soil sample min salt content on-line measurement result is C1, soil sample max salt content on-line measurement result is C2, the accurate results of soil sample min salt content is D1, the accurate results of soil sample max salt content is D2, utilize (C1, D1), (C2, D2) linear scaled model is set up, then according to linear scaled model, all on-line measurement results are revised, thus complete the on-line measurement of soil salt content.

Present embodiment step one is in order to take into full account conforming factor, according to the objective scope of soil salt content in the soil salt content classification chart in salt content region and region, many groups soil sample of the different salt content of uniform design in the scope of mulching soil salt content.In the experiment of step one soil cracking when using the pedotheque of digital camera shooting cracking, in order to reduce the impact of bad border change comparison film quality of taking pictures, need the white balance calibration carrying out camera with the white color board of standard, and the light sensitivity of camera is set according to the intensity of illumination that digital photometer is measured, the parameter such as aperture size and shutter speed.

In present embodiment step one, texture feature extraction part is split soil sample photo carry out geometric distortion correction to often being opened, geometric distortion correction be center distortion in order to remove the photo that camera lens in the process of taking pictures causes and take pictures time the angle lens deviation distortion of photo on some directions that cause, to obtain all identical true soil sample image of all directions ratio.According to grey level histogram selected threshold and then carry out binary conversion treatment, its process is to each soil sample image line statistics of histogram, notable difference is there is because crackle is in soil surface gray scale, therefore grey level histogram presents obvious bimodal morphology, select the lowest point place gray level between two gray scale peak values as threshold value, and then binary conversion treatment is carried out to gray level image.

N in contrast formula in present embodiment textural characteristics value represents the difference between two gray levels.Calculate the gray level co-occurrence matrixes of each cracking soil sample binary image, the element of gray level co-occurrence matrixes is the second order hybrid conditional probability density between image picture elements gray scale, it reflects in image the Gray Correlation between the pixel that has on certain distance step-length and direction.For a known gray level image f (x, y), if the gray level of image is defined as L level, then gray level co-occurrence matrixes is the matrix of L × L size, and the computing formula of its second order hybrid conditional probability density is as follows:

p(i,j)＝g{(x1,y1),(x2,y2)∈m×n|f(x1,y1)＝i,f(x2,y2)＝j}，

Wherein p (i, j) represents the element value of gray level co-occurrence matrixes in (i, j) position, and wherein i, j represent the gray level of gray level image f (x, y) at (x1, y1) and (x2, y2) coordinate position place respectively.

Embodiment two: present embodiment and embodiment one are split soil sample photo unlike adopting polynomial method in step one texture feature extraction part carry out geometric distortion correction to often being opened.Other step and parameter identical with embodiment one.

Embodiment three: the mass ratio that present embodiment and embodiment one or two measure soil quality and moisture in soil extraction in soil salt part unlike step one is 1:5.Other step and parameter identical with embodiment one or two.

Embodiment four: one of present embodiment and embodiment one to three are measured in soil salt part unlike step one and adopted flare photometer to measure Na in soil sample B ⁺and K ⁺content.Other step and parameter identical with one of embodiment one to three.

Embodiment five: one of present embodiment and embodiment one to four are measured in soil salt part unlike step one and adopted EDTA compleximetry to measure Ca in soil sample B ²⁺and Mg ²⁺content.Other step and parameter identical with one of embodiment one to four.

Embodiment six: one of present embodiment and embodiment one to five are measured in soil salt part unlike step one and adopted AgNO ₃solution titrimetry measures Cl in soil sample B ^-content.Other step and parameter identical with one of embodiment one to five.

Embodiment seven: one of present embodiment and embodiment one to six are measured in soil salt part unlike step one and adopted Dual-indicator neutralisation to measure CO in soil sample B ^2-and HCO ₃ ^-content.Other step and parameter identical with one of embodiment one to six.

Embodiment: the On-line Measuring Method of the present embodiment salinized soil salt content follows these steps to implement:

One, soil salt content inverse model is set up

1, soil cracking experiment

Select 56 groups of soil samples of different salt content according to space distribution and salinization and alkalization in the most serious Daan City of song-Nen plain soda soil salinization and alkalization, carry out air-dry to all soil samples, milled processed also crosses the sieve in 2mm aperture, then the soil sample after every group being sieved is divided into soil sample A and soil sample B, soil sample A is wherein carried out dry shrinkage and cracking experiment, described dry shrinkage and cracking experiment first all soil sample A is configured to the identical saturated mud of water cut respectively, saturated mud good for every group configuration is poured into respectively in the wooden sample box of 50 cm x 50 cm x, 3 centimetres of sizes, place after floating mud and carry out natural drying in the lab to produce cup shake, until when in sample box, the quality of sample no longer reduces, obtain the soil sample A of season cracking,

Digital camera is arranged on fixing experiment porch, digital camera distance floor level is 1m, with the digital camera lens center that is projected as on the ground determine the square area of 50 cm x 50 centimetres with ensure the soil sample A of all season crackings take pictures process geometric distortion impact identical, the black and white grid calibration plate of 50 cm x 50 centimetres is covered the soil sample A surface of season cracking, then use the soil sample A of digital camera to season cracking to take pictures, obtain cracking soil sample photo;

2, texture feature extraction

Polynomial method is adopted to split soil sample photo carry out geometric distortion correction to often opening, the colored slit region image obtaining sample is sheared to the sample image after geometric distortion correction, then gray-scale map is converted into by often opening colored slit region image, again according to grey level histogram selected threshold and then carry out binary conversion treatment, then the crack image of each cracking soil sample is divided into uncracked piece of district (representing with black picture dot) and constriction zone (comprise the constriction of surrounding in fragmented parts and sample box, represent with white picture dot) after inversion operation;

Calculate the gray level co-occurrence matrixes of each cracking soil sample binary image, gray level Ng is set to 2, distance step-length is set to 40,4 directions are respectively 0 °, 45 °, 90 ° and 135 °, and the Normalized Grey Level co-occurrence matrix then obtained by four direction carries out arithmetic mean thus with the impact of removing direction, calculates each Statistic Texture value according to each cracking soil sample gray level co-occurrence matrixes, the computing formula of each Statistic Texture value is as follows, wherein n=i-j, 0≤i≤Ng, 0≤j≤Ng:

1) contrast:

$\mathrm{contra}=\underset{n}{\overset{\mathrm{Ng}-1}{\mathrm{\Σ}}}{n}^2\left\{\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\right\},|i-j|=n,---\left(1\right)$

2) angle second moment:

$\mathrm{ASM}=\underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}{\left\{p(i,j)\right\}}^2,---\left(2\right)$

3) entropy:

$\mathrm{Entropy}=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left(p(i,j)\right),---\left(3\right)$

4) consistance:

$\mathrm{Homo}=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\frac{1}{1+{(i-j)}^2}p(i,j),---\left(4\right)$

5) correlativity:

$\mathrm{Cerrelation}=\frac{\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\left(\mathrm{ij}\right)p(i,j)-{\mathrm{\μ}}_x{\mathrm{\μ}}_y}{{\mathrm{\σ}}_x{\mathrm{\σ}}_y},---\left(5\right)$

6) cluster is shady:

$\mathrm{Clustershade}=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}{((i-{\mathrm{\μ}}_i)+(j-{\mathrm{\μ}}_j))}^{3}p(i,j),---\left(6\right)$

7) maximum probability:

Maxprobability＝max{p(i,j)}， (7)

8) and average:

$\mathrm{SumAverage}=\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{\mathrm{ip}}_{x+y}\left(i\right),---\left(8\right)$

P _x+yi () represents that the probable value of gray level i appears in x capable and y row simultaneously;

9) and entropy:

$\mathrm{SumEntropy}=-\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{p}_{x+y}\left(i\right)\log \left\{p_{x+y}\left(i\right)\right\},---\left(9\right)$

10) and variance:

$\mathrm{SumVariance}=\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{(i-\mathrm{SumEntropy})}^2{p}_{x+y}\left(i\right),---\left(10\right)$

11) relevant information feature 1:

$\mathrm{Inforofcorrelation}1=\frac{\mathrm{HXY}-\mathrm{HXY}1}{\max (\mathrm{HX},\mathrm{HY})},---\left(11\right)$

12) relevant information feature 2:

Inforofcorrelation2＝{1-exp[-2*(HXY2-HXY)]} ^1/2， (12)

Wherein:

P (i, j) is Normalized Grey Level co-occurrence matrix (i, j) position element value, i.e. the probable value that simultaneously occurs of image gray levels i, j, and Ng is co-occurrence matrix number of greyscale levels, and i, j are positive integer,

$p_x\left(i\right)=\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),---\left(13\right)$

$p_y\left(j\right)=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),---\left(14\right)$

$p_{x-y}\left(k\right)=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),i+j=k,k=\mathrm{2,3},....2\mathrm{Ng},---\left(15\right)$

$\mathrm{HXY}=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left\{p(i,j)\right\},---\left(16\right)$

$\mathrm{HXY}1=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left\{p_x\left(i\right)p_y\left(j\right)\right\},---\left(17\right)$

$\mathrm{HXY}2=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}{p}_x\left(i\right)p_y\left(j\right)\log \left\{p_x\left(i\right)p_y\left(j\right)\right\},---\left(18\right)$

μ _x, μ _yfor p _x(i), p _y(j) average, σ _x, σ _yfor p _x(i), p _y(j) variance, Ng is gray level;

3, soil salt is measured

Each soil sample B is configured to soil quality-biodiversity than the soil extraction for 1:5, measures the Ca in soil sample B respectively ²⁺, Mg ²⁺, Cl ^-, CO ^2-and HCO ₃ ^-content, calculates Ca in each soil sample B ²⁺, Mg ²⁺, Cl ^-, CO ^2-and HCO ₃ ^-the summation of content, be the salt content of soil sample B, measurement result is as table 1;

Cracking sample salt content tested by table 1

4, inverse model is set up

The salt content of soil sample B and each Statistic Texture value of calculating is utilized to set up linear model, horizontal ordinate (independent variable) wherein in linear model is texture characteristic amount, ordinate (dependent variable) is salt content, comparative selection degree contra (CON) is as texture characteristic amount, the measurement result of contrast contra is as table 2, set up the normal linearity inverse model of soil salt content field on-line measurement, contrast and salt content linear model y=79.41x – 0.7144;

Cracking sample contrast C ON tested by table 2

Two, on-line measurement field soil salt content

Under 48 natural conditions are selected in Daan City, Jilin Province, the soil sample of season cracking carries out the on-line measurement of soil salt content, for the on-line measurement of field to be measured salinized soil, digital camera is placed on a mobile platform, adjustment podium level makes digital camera height perpendicular to the ground be 1m, it is identical that digital camera parameter and soil parameter when testing that ftractures is set, then the cracking salinized soil in field is taken pictures, geometric distortion correction is carried out to the cracking salinized soil photograph in field, to take pictures corresponding 50cm*50cm area size and shearing field, position photo according to soil sample A, thus obtain standard crack photo, subsequently gray-scale map conversion and binary conversion treatment are carried out to standard crack photo, calculate Normalized Grey Level co-occurrence matrix and contrast C ON textural characteristics value corresponding to standard inversion model, it is as shown in table 3 that contrast C ON textural characteristics value extracts result, the result of calculation of contrast C ON textural characteristics value is brought in normal linearity inverse model, thus realize the preliminary on-line measurement of salinized soil salt content, preliminary on-line measurement result is as shown in table 4,

Table 3 field crackle sample contrast C ON

Table 4 on-line measurement PRELIMINARY RESULTS

Three, soil salt content measurement result is revised

In all on-line measurement field soil samples, select at soil sample min corresponding to the textural characteristics minimum value of line computation with at soil sample max corresponding to the textural characteristics maximal value of line computation, accurately measure the salt content of soil sample min and soil sample max respectively, if soil sample min salt content on-line measurement result C1 is 1.45mg/g, soil sample max salt content on-line measurement result C2 is 44.31mg/g, the accurate results D1 of soil sample min salt content is 2.87mg/g, the accurate results D2 of soil sample max salt content is 21.84mg/g, utilize (C1, D1), (C2, D2) (wherein x represents the first on-line measurement value of salt content to set up linear scaled model y=0.46x+0.78, y represents salt content calibration value), then according to linear scaled model, all on-line measurement results are revised, thus complete the on-line measurement of soil salt content, measurement result is as shown in table 5, precision R ²=0.94.

Table 5 salt content on-line measurement calibration result

Claims (7)

1. an On-line Measuring Method for salinized soil salt content, is characterized in that the On-line Measuring Method of salinized soil salt content follows these steps to realize:

One, soil salt content inverse model is set up

1, soil cracking experiment

Select many groups soil sample of different salt content, carry out air-dry to all soil samples, milled processed also crosses the sieve in 2mm aperture, then the soil sample after every group being sieved is divided into soil sample A and soil sample B, soil sample A is wherein carried out dry shrinkage and cracking experiment, described dry shrinkage and cracking experiment first all soil sample A is configured to the identical saturated mud of water cut respectively, saturated mud good for every group configuration is poured into respectively in the wooden sample box of 50 cm x 50 cm x, 3 centimetres of sizes, place after floating mud and carry out natural drying in the lab to produce cup shake, until when in sample box, the quality of sample no longer reduces, obtain the soil sample A of season cracking,

Digital camera is arranged on fixing experiment porch, digital camera distance floor level is 1m, with the digital camera lens center that is projected as on the ground determine the square area of 50 cm x 50 centimetres with ensure the soil sample A of all season crackings take pictures process geometric distortion impact identical, the black and white grid calibration plate of 50 cm x 50 centimetres is covered the soil sample A surface of season cracking, then use the soil sample A of digital camera to season cracking to take pictures, obtain cracking soil sample photo;

2, texture feature extraction

Split soil sample photo carry out geometric distortion correction to often opening, the colored slit region image obtaining sample is sheared to the sample image after geometric distortion correction, then gray-scale map is converted into by often opening colored slit region image, again according to grey level histogram selected threshold and then carry out binary conversion treatment, after inversion operation, then the crack image of each cracking soil sample is divided into uncracked piece of district and constriction zone;

Calculate the gray level co-occurrence matrixes of each cracking soil sample binary image, 40 will be set to apart from step-length, 4 directions are respectively 0 °, 45 °, 90 ° and 135 °, then the Normalized Grey Level co-occurrence matrix obtained by four direction carries out arithmetic mean thus with the impact of removing direction, calculate each Statistic Texture value according to each cracking soil sample gray level co-occurrence matrixes, the computing formula of each Statistic Texture value is as follows:

1) contrast:

$\mathrm{contra}=\underset{n}{\overset{\mathrm{Ng}-1}{\mathrm{\Σ}}}{n}^2\left\{\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\right\},|i-j|=n---\left(1\right)$

2) angle second moment:

$\mathrm{ASM}=\underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}{\left\{p(i,j)\right\}}^2,---\left(2\right)$

3) entropy:

$\mathrm{Entropy}=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left(p(i,j)\right),---\left(3\right)$

4) consistance:

$\mathrm{Homo}=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\frac{1}{1+{(i-j)}^2}p(i,j),---\left(4\right)$

5) correlativity:

$\mathrm{Correlation}=\frac{\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\left(\mathrm{ij}\right)p(i,j)-{\mathrm{\μ}}_x{\mathrm{\μ}}_y}{{\mathrm{\σ}}_x{\mathrm{\σ}}_y},---\left(5\right)$

6) cluster is shady:

$\mathrm{Clustershade}=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}{((i-{\mathrm{\μ}}_i)+(j-{\mathrm{\μ}}_j))}^{3}p(i,j),---\left(6\right)$

7) maximum probability:

Maxprobability＝max{p(i,j)}， (7)

8) and average:

$\mathrm{SumAverage}=\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{\mathrm{ip}}_{x+y}\left(i\right),---\left(8\right)$

P _x+yi () represents that the probable value of gray level i appears in x capable and y row simultaneously;

9) and entropy:

$\mathrm{SumEntropy}=-\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{p}_{x+y}\left(i\right)\log \left\{p_{x+y}\left(i\right)\right\},---\left(9\right)$

10) and variance:

$\mathrm{SumVariance}=\underset{i=2}{\overset{2\mathrm{Ng}}{\mathrm{\Σ}}}{(i-\mathrm{SumEntropy})}^2{p}_{x+y}\left(i\right),---\left(10\right)$

11) relevant information feature 1:

$\mathrm{Inforofcorrelation}1=\frac{\mathrm{HXY}-\mathrm{HXY}1}{\max (\mathrm{HX},\mathrm{HY})},---\left(11\right)$

12) relevant information feature 2:

Inforofcorrelation2＝{1-exp[-2*(HXY2-HXY)]} ^1/2， (12)

Wherein:

P (i, j) is Normalized Grey Level co-occurrence matrix (i, j) position element value, i.e. the probable value that simultaneously occurs of image gray levels i, j, and Ng is co-occurrence matrix number of greyscale levels, and i, j are positive integer,

$p_x\left(i\right)=\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),---\left(13\right)$

$p_y\left(j\right)=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),---\left(14\right)$

$p_{x-y}\left(k\right)=\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j),i+j=k,k=\mathrm{2,3},...2\mathrm{Ng},---\left(15\right)$

$\mathrm{HXY}=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left\{p(i,j)\right\},---\left(16\right)$

$\mathrm{HXY}1=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}p(i,j)\log \left\{p_x\left(i\right)p_y\left(j\right)\right\},---\left(17\right)$

$\mathrm{HXY}2=-\underset{i=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}\underset{j=1}{\overset{\mathrm{Ng}}{\mathrm{\Σ}}}{p}_x\left(i\right)p_y\left(j\right)\log \left\{p_x\left(i\right)p_y\left(j\right)\right\},---\left(18\right)$

μ _x, μ _yfor p _x(i), p _y(j) average, σ _x, σ _yfor p _x(i), p _y(j) variance, n is gray level;

3, soil salt is measured

Each soil sample B is configured to soil extraction, measures the Ca in soil sample B respectively ²⁺, Mg ²⁺, Cl ^-, CO ^2-and HCO ₃ ^-content, calculates Ca in each soil sample B ²⁺, Mg ²⁺, Cl ^-, CO ^2-and HCO ₃ ^-the summation of content, is the salt content of soil sample B;

4, inverse model is set up

The salt content of soil sample B and each Statistic Texture value of calculating is utilized to set up linear model, horizontal ordinate wherein in linear model is texture characteristic amount, ordinate is salt content, select the texture characteristic amount best with soil sample B salt content fit correlation, set up the normal linearity inverse model of soil salt content field on-line measurement;

Two, on-line measurement field soil salt content

For the on-line measurement of field to be measured salinized soil, digital camera is placed on a mobile platform, adjustment podium level makes digital camera height perpendicular to the ground be 1m, it is identical that digital camera parameter and soil parameter when testing that ftractures is set, then the cracking salinized soil in field is taken pictures, geometric distortion correction is carried out to the cracking salinized soil photograph in field, to take pictures corresponding 50cm*50cm area size and shearing field, position photo according to soil sample A, thus obtain standard crack photo, subsequently gray-scale map conversion and binary conversion treatment are carried out to standard crack photo, calculate Normalized Grey Level co-occurrence matrix and the best textural characteristics value of corresponding fit correlation, the result of calculation of textural characteristics value best for fit correlation is brought in normal linearity inverse model, thus realize the preliminary on-line measurement of salinized soil salt content,

Three, soil salt content measurement result is revised

In all on-line measurement field soil samples, select at soil sample min corresponding to the textural characteristics minimum value of line computation with at soil sample max corresponding to the textural characteristics maximal value of line computation, accurately measure the salt content of soil sample min and soil sample max respectively, if soil sample min salt content on-line measurement result is C1, soil sample max salt content on-line measurement result is C2, the accurate results of soil sample min salt content is D1, the accurate results of soil sample max salt content is D2, utilize (C1, D1), (C2, D2) linear scaled model is set up, then according to linear scaled model, all on-line measurement results are revised, thus complete the on-line measurement of soil salt content.

2. the On-line Measuring Method of a kind of salinized soil salt content according to claim 1, is characterized in that adopting in step one texture feature extraction part polynomial method to split soil sample photo carry out geometric distortion correction to often opening.

3. the On-line Measuring Method of a kind of salinized soil salt content according to claim 1, is characterized in that the mass ratio that step one measures soil quality and moisture in soil extraction in soil salt part is 1:5.

4. the On-line Measuring Method of a kind of salinized soil salt content according to claim 1, is characterized in that step one is measured in soil salt part and adopts flare photometer to measure Na in soil sample B ⁺and K ⁺content.

5. the On-line Measuring Method of a kind of salinized soil salt content according to claim 1, is characterized in that step one is measured in soil salt part and adopts EDTA compleximetry to measure Ca in soil sample B ²⁺and Mg ²⁺content.

6. the On-line Measuring Method of a kind of salinized soil salt content according to claim 1, is characterized in that step one is measured in soil salt part and adopts AgNO ₃solution titrimetry measures Cl in soil sample B ^-content.

7. the On-line Measuring Method of a kind of salinized soil salt content according to claim 1, is characterized in that step one is measured in soil salt part and adopts Dual-indicator neutralisation to measure CO in soil sample B ^2-and HCO ₃ ^-content.