CN104835124B - A kind of froth images color calibration method and device based on Spectrum Data Fusion - Google Patents

Abstract

The invention discloses a kind of froth images color calibration method based on Spectrum Data Fusion, and froth images camera color data is gathered using industrial camera, and foam spectral information is gathered using visible light detecting system.By analyzing influence of the external environment to color of image, Medium Propagation color change model is established.Original image dark numerical value is calculated, solves the Medium Propagation dissipation flow function of image.Merge spectral information and improve Medium Propagation dissipative function, correct the color dissipation amount of each pixel, the Accurate color numerical value of froth images is reduced using the model.The present invention changes frequently for industrial environment light, air has water smoke, causes that image is fuzzy, cross-color is very targeted, meanwhile, spectroscopy equipment has environment no dependence, is highly suitable for industry spot.

Description

A foam image color correction method and device based on spectral information fusion

technical field

The invention belongs to the fusion field of spectral analysis and computer digital image processing, and in particular relates to an image color correction method and device in the froth flotation process.

Background technique

Among the commonly used beneficiation methods, froth flotation is the most important ore extraction method. The color of the foam is not only closely related to the flotation production conditions, but also reflects the types of minerals contained in the foam and the degree of enrichment of mineral particles, which is an important basis for the operation of the flotation process. Therefore, accurate extraction of flotation froth color features is crucial. Industrial sites collect image color data through cameras, and the process of collecting images by industrial cameras is too dependent on the environment. The environment of the industrial site is harsh, and the light source is easily affected, resulting in unstable light and frequent light changes, resulting in color distortion of the collected images. At the same time, there is fog and dust in the air on site, which affects the imaging effect of the camera. The combination of the two results in inaccurate foam color data collected by the camera, which affects the effect of the flotation control process. Therefore, it is of great significance to study the correction method of the color characteristics of flotation froth.

The current method of foam color correction is limited by the influence of equipment and environment. Most of them compare the foam color data before and after a period of time, analyze the change trend, and study the superimposed influence of the environment on the color data.

It includes the following two categories of methods: The first category is the color correction method based on space mapping, which realizes color space conversion by determining the mapping relationship between the source color space and the target color space. This type of method can achieve relatively high accuracy, but only in the training Only when the samples are sufficient and reasonable can better results be obtained.

The second category is the color correction method based on image analysis, and the representative algorithm is the histogram method and the like. This type of correction algorithm is based on the study of the color distribution of the image itself, and according to the chromaticity distribution characteristics of the image to detect whether the image is color cast and the degree of color cast, and to adjust the overall digital image, generally used in the field of image color cast detection and other fields .

The above algorithms are all designed to perform data modeling on the basis of a large amount of data analysis. They do not analyze the cause of image color changes from the perspective of mechanism, and are not targeted at the color cast phenomenon in industrial sites. At the same time, the imaging process of industrial cameras is greatly affected by the external environment, and the image acquisition process cannot get rid of the dependence on the on-site environment. When the interference conditions increase, the correction method based on the analysis of color data cannot achieve a good correction effect.

Analyzing the on-site environment of the mineral flotation industry, there are two major factors that cause inaccurate colors: the light of the on-site light source changes with time, and the imaging of the industrial camera changes due to the influence of the light source; the fog is diffuse, and the collected images are not clear, resulting in inaccurate colors.

Contents of the invention

The invention provides a foam image color correction method and device which do not require a large amount of color data and have high correction accuracy.

To achieve the above object, the technical scheme of the present invention is as follows:

A foam image color correction method based on spectral information fusion, comprising the following steps:

Step 1: Under the condition of fixed light source and no fog interference, collect the spectral data of different color blocks of the color card and camera color data, and establish a spectrum-camera color space conversion model based on the characteristics of the camera;

Step 2: In the industrial site, the camera collects the original image data I(x) of the foam in the flotation tank and uploads it to the host computer for processing;

Step 3: Use the visible light spectrum system to collect the spectral reflection information of the scattered points of the foam in the flotation cell, and convert the spectral color information of the foam into the RGB color space numerical representation of the camera according to the color space conversion model in step 1;

Step 4: Establish a foam color change model based on the atmospheric scattering principle and the ambient light model;

Step 5: Take the minimum value of the three channels of each pixel of the original image I(x) to form a minimum color value image, perform minimum filtering operations on each pixel of the image matrix, obtain its dark channel image, and calculate the medium propagation Dissipation function;

Step 6: Use the foam scatter point spectral color values collected in step 3 and converted to RGB space to correct the medium propagation dissipation function and solve the accurate color dissipation per pixel of the image;

Step 7: Substituting the accurate color dissipation amount and the original foam image data into the color change model to determine the corrected accurate image color data.

Wherein, the establishment of the color space conversion model in step 1 includes:

Step 11: Under the D65 light source, take the center of each color block of the standard color card as the midpoint, and use 5*10 pixels as the window size, and the industrial camera collects RGB data of the image, and takes the average as the color data value rgb _n of the color block, n=1. .140. Simultaneously collect spectral reflectance k _n of different color blocks, where n=1,2...140;

Step 12: Using CIE1931 as the standard, use the reflectance data of different color blocks to calculate the chromaticity coordinates (x _n , y _n , z _n ) of each color block, and use the position in the cone color space where the chromaticity coordinates are located to obtain Characterizes the color distribution of the foam.

Step 13: With the chromaticity coordinates x, y, z as the smallest unit, establish a decaterm polynomial model of different color blocks, the expression is as follows:

Taking the items of the polynomial as independent variables and taking the RGB color data of the camera as dependent variables, the expressions are as follows:

The multiple linear regression models of the three channels are respectively established, and the expressions are as follows:

E is the coefficient of the multiple regression equation. This model realizes the transformation from the chromaticity coordinate space to the camera color space. Any spectral color information can be represented by the camera color RGB space, that is, rgb=xyz*E.

Wherein, in step 4, the foam color change model is established, and the process includes:

Step 41: According to the analysis of the principle of light propagation and color formation, the influence of the environment on the color is divided into two parts: reflection attenuation loss and ambient light model. Combining the principle of color generation and the atmospheric scattering model, a medium propagation image change model is established. The expression is as follows:

I(x)＝S(x)e ^-βd(x) +A(1-e ^-βd(x) ) (4)

S(x) is the image color when it is not disturbed, that is, the correct image color after correction, x represents each pixel in the picture, A is the atmospheric light value of the scene environment, according to the atmospheric reflection model, S(x) should be the atmospheric The product of light and scene albedo ρ, the scene albedo ρ only depends on the physical reflectivity of the object to light, e ^-βd(x) is the medium propagation dissipation function;

Step 42: Define t(x)=e ^-βd(x) as the medium propagation dissipation function, then the color change model is simplified to an expression:

I(x)=S(x)t(x)+A(1-t(x)). (5)

Among them, the process of correcting the medium propagation dissipation function in step six includes:

Step 61: According to the dark channel prior, solve the media propagation dissipation t ₁ (x) of each point of the original image, the expression is as follows:

J ^dark is the dark channel image of the original image obtained in step five;

Step 62: multiple sets of foam spectrum RGB color data (r _n , g _n , b _n ) measured in step 3 are used as the accurate foam physical color attributes in this case, and step 5 is repeated to obtain the scattered points of each group of foam The exact dissipation at , the expression is as follows:

Step 63: According to the medium propagation function t(x)=e ^-βd(x) , calculate the dissipation t(x) of all pixels in the image

Arranged in ascending order, draw an exponential curve s1; according to this curve, fit and obtain the coefficients of the exponential function, the expression is as follows:

Step 64: Arrange each group of accurate dissipation quantities t ₂ (n) obtained in step 62 in ascending order, and fit the exact dissipation function, the expression is as follows:

Step 65: Compare the two sets of dissipation functions, twist, move, enlarge or reduce the curve s1 to make it infinitely close to the dissipation curve s2, that is, transform the function s1 into the form of the function s2, and correct all s1 The specific process is as follows:

Take the logarithm on both sides of s1 and s2 at the same time, the expression is as follows:

Transform the formulas 10 and 11 to get the following expressions:

Make it consistent with the form of the medium propagation dissipation function of the scattered point, then the value on the right side of the deformed formula is the logarithm of the corrected all accurate dissipation, and take the exponent, then the new total dissipation qt(x) , the expression is as follows:

Step 66: According to the sorting order in step 62, restore the dissipation value of different pixels in qt(x) to the original order, and obtain the accurate dissipation matrix t(x) of all pixels in the foam image.

Wherein, determining the corrected accurate image color data in step seven is specifically: according to the color change model, I(x)=S(x)t(x)+A(1-t(x)), known atmospheric light A The exact color data of the image can be obtained with the dissipation t(x), and the expression of S(x) is as follows:

A foam image color correction device based on spectral information fusion, including the following modules: (17)

The image camera color acquisition module is used to collect the camera RGB color of the foam image. The module uses the industrial camera to collect the RGB color information of the entire image, and transmits the color data to the host computer for processing;

The spectral information collection module is used to collect the scattered point spectral reflectance information at different positions within the range of the image area. The spectral collection module divides the image to be measured into different regions, and uses the visible light micro-spectrometer to collect the scattered point spectral reflectance in each region information;

The data processing module is used to process the camera RGB information and spectral information of the image, use the established model to correct the original image data collected by the camera, and output the corrected color data.

The invention combines the light propagation mechanism with the spectral analysis method, more accurately reproduces the process of color data being affected by the outside world, and at the same time gets rid of the previous requirement for a large amount of color data, and improves the correction accuracy.

Description of drawings

Fig. 1 is a device diagram of the present invention.

Fig. 2 is a flow chart of conversion between industrial camera color space and spectral color space in the implementation of the present invention.

Fig. 3 is a flow chart of color correction for fusing spectral color information in the implementation of the present invention.

Figure 4 is the dark channel image through the original image.

Fig. 5 is a rough full-pixel medium dissipation curve.

Fig. 6 is the accurate medium dissipation curve of the scatter point.

detailed description

The present invention will be further described below in conjunction with the accompanying drawings and examples.

The external environment has two influences on the color of the foam image collected by the camera: one is affected by the industrial site environment, which causes the light to scatter during the propagation of the light to the foam surface, resulting in a loss of part of the reflected light intensity and damage to the image color components. The farther the distance, the deeper the depth of field, the greater the loss, and the intensity of reflected light is inversely proportional to the depth of field; second, due to the ambient light model and the scattering of natural light by atmospheric particles in the process of light propagation, the light source property appears, resulting in an increase in the color component of the image. The longer the propagation distance, the stronger the light source is presented.

In industrial sites, the light source light is unstable, and the scattering loss model and the ambient light model work together, resulting in inaccurate color data of images collected by the camera. Peeling back the superimposed effects of these models in the affected image yields the exact physical color properties of the foam itself.

The present invention aims at camera acquisition color error caused by external environment, fuses spectral information of images to model, and corrects image color data.

The image color correction device used in this example is shown in Figure 1, and the device includes a light source 1, an industrial camera 2, a straight-through optical fiber 3, a visible light micro-spectrometer 4, a color card 5 and a computer 6. The arrangement of the device is as follows: light source 1 selects D65 light source with the same color temperature as the industrial site as the standard light source; industrial camera 2 is connected to the usb interface of computer 6 to upload the camera to collect image data; Detect its light reflection; the other end of the optical fiber 3 is connected to the spectrometer 4 for sending light reflection information; the spectrometer 4 is connected to the usb interface of the computer 6 to transmit spectral reflection data. Color card 5 is a 140-color color card of X-Rite Company, which is used as a modeling and testing object.

From the perspective of functional modules, the image color correction device includes three parts: image camera color acquisition module, spectral information acquisition module and data processing module.

The image camera color acquisition module is used to collect the camera RGB color of the foam image. The module uses the industrial camera to collect the RGB color information of the entire image, and transmits the color data to the host computer for processing, mainly including the industrial camera.

The spectral information collection module is used to collect scattered spectral reflectance information at different positions within the image area. The spectrum acquisition module divides the image to be measured into different regions, and uses a visible light micro-spectrometer to collect the scattered spectral reflectance information in each region. It mainly includes optical fiber and visible light spectrometer.

The data processing module is used to process the camera RGB information and spectral information of the image, use the established model to correct the original image data collected by the camera, and output the corrected color data. The main equipment is a computer.

As shown in FIG. 2 and FIG. 3 , the image color correction method based on spectral information fusion in this embodiment includes the following specific implementation steps.

Step 1: In an environment with a stable light source and no external interference, establish a spectrum-camera color space conversion model P _xyz2rgb based on the spectral characteristics of the camera;

(1) Under the condition of stable irradiation of d65 light source, use a miniature visible light spectrum analyzer to measure the reflectance k _n of all the color blocks of the color card, where n=1,2...140;

(2) Use an industrial camera to collect the image data of the color card. The same color block is centered on the midpoint of the color block, and the window size is 5*10. Take 50 pixel values, and take the average as the final color data value of the color block of the color card rgb _n , n=1..140, transmitted to the host computer for processing;

(3) Based on the provisions of CIE1931, through 140 sets of reflectance data, calculate the spectral chromaticity coordinate data (x _n , y _n , z _n ) corresponding to the measurement point. The data exists in the spectral xyz color space, and the chromaticity coordinates The position in the cone color space represents the color distribution of the foam;

(4) Separate the three channels of the camera color, and use the chromaticity coordinates x, y, and z as the smallest unit to establish a ten-term polynomial regression model. The formula is as follows:

The terms of the polynomial are used as the independent variables of the multiple linear regression, and the three-channel color data of the corresponding color block camera are used as the dependent variables, and the expressions are as follows:

Among the 140 sets of data, 100 sets were randomly selected as the training set to solve the coefficients of the linear regression equation. The expression is as follows:

Use the remaining 40 sets of data to verify the regression accuracy and complete the conversion process between the RGB color space of the industrial camera and the spectral xyz chromaticity coordinate color space, then any chromaticity coordinate can be converted into a camera color space value, that is, rgb=xyz*E.

Step 2: In the industrial field environment, the industrial camera collects the foam image data in the flotation tank and transmits it to the host computer for processing;

Step 3: Due to the limitation of the instrument, the spectrometer is only allowed to measure scattered point data each time. The image area collected by the camera is divided into 10*8 areas, and the optical fiber probe of the spectrum system is used to collect the flotation foam spectrum of a random point in 80 areas. Reflectance information k _n , where n=1,2...80, obtain all chromaticity coordinates (x _n ,y _n ,z _n ) according to the method in step 1 (3);

Step 4: Using the conversion model P _xyz2rgb obtained in step 1, convert the chromaticity coordinates obtained in step 3 into standard RGB color data (r _n , g _n , b _n )n=1,2...80 , representing the physical color values of the 80 measurement scattered points in the flotation cell in this environment;

Step 5: Build a color change model

According to the analysis, the influence of the environment on the color is divided into two parts: the reflection attenuation loss and the ambient light model. Combining the color generation principle and the atmospheric scattering model, the medium propagation image change model is defined as follows:

I(x)＝S(x)e ^-βd(x) +A(1-e ^-βd(x) ) (4)

I(x) is the original image of the camera, and S(x) is the color of the image without interference, that is, the accurate color of the image after correction.

A is the atmospheric light of the scene environment. According to the atmospheric reflection model, S(x) should be the product of atmospheric light and scene albedo ρ. Scene albedo depends only on how physically reflective an object is to light.

e ^-βd(x) is the dissipation function of medium propagation. The farther the light propagation distance is, the greater the scattering loss is, the weaker the reflected light intensity is, and the more obvious the light source characteristics of ambient light are;

Step 6: Solve the dark channel image and atmospheric light A of the original image captured by the camera in step 2. The process includes:

(1) Separate the original color image by R, G, and B three channels, take the minimum value of the three channels for each pixel as the minimum brightness value of the pixel, and form the minimum brightness matrix image J, the expression is as follows:

J＝min _rgb (I ^c (x,y)) (5)

Wherein x, y are the pixel coordinates of each point in the image, and ^Ic is the color image collected by the original camera;

(2) Take x, y as the filter center pixel position, and perform minimum value filtering on the minimum brightness image J in formula 5. The filter window size is 15*15, and the filter radius is 7. The expression of each pixel after the minimum value filter is as follows:

J ^dark (x,y)=min(J(x',y')) (6)

Among them, x', y' are each pixel in the template area centered on x, y. The filtered image J ^dark is the dark channel image of the original image collected by the camera; the solved image is shown in Figure 4.

(3) After the filtered J ^dark image data values are arranged in ascending order, the pixel value positions occupying the first 0.1% are selected, and among these positions in the original image, the largest one among the RGB three-channel data is selected as the atmospheric light value A.

Step 7: Obtain and correct the dissipation of medium propagation, the process includes:

(1) Define t(x)=e ^-βd(x) as the medium propagation function, then the expression of the image propagation model is as follows:

I(x)=S(x)t(x)+A(1-t(x)) (7)

Perform minimum filtering on both sides of the model, and the model becomes:

min(min(I(x)))=min(min(S(x)))t(x)+A(1-t(x)) (8)

According to the dark channel prior, after the value of the three color channels of the fog-free image is minimized, the value of the dark channel image tends to 0. The left side of formula (8) is the dark channel image J ^dark of the original image, and the first item on the right side is the dark channel image of the interference-free image S(x), whose value tends to 0, and the rough medium propagation function t ₁ (x) is obtained , the expression is as follows:

x represents the coordinates of the pixel, and the rough dissipation of an image on all pixels can be calculated;

(2) Spectral color information is an inherent property of the foam and represents an accurate color value. The spectral color (r _n , g _n , b _n ) after 80 sets of scatter point conversion is its exact color value, that _is , (rn , g _n , b _n )=S _n . Substituting the scattered point color data (r _n , g _n , b _n ) into formula (7), and repeating the operation in step 7 (1), the accurate medium propagation dissipation t ₂ of 80 groups of foam scattered points can be calculated (x), the expression is as follows:

These 80 sets of dissipation represent the light loss and superposition caused by the environment in different areas of the flotation cell;

(3) Arrange the dissipation quantities t ₁ (x) of all pixel points in ascending order, and draw a curve s1 , as shown in FIG. 5 . According to the medium propagation function t(x)=e ^-βd(x) , the curve should be approximated as an exponential function related to the light propagation distance, that is, the depth of field. According to this curve, the coefficients of the exponential function can be obtained by fitting, and the expressions are as follows :

(4) Arrange 80 groups of accurate dissipation quantities t ₂ (n) in ascending order, repeat the above process, and draw the curve as shown in Figure 6, and obtain the coefficients of the function by fitting, the expression is as follows:

(5) Comparing the two sets of medium propagation functions, the curve s1 is twisted, moved, enlarged or reduced, and the deformation principle is to make it infinitely similar to the dissipation curve s2. That is, the function s1 is transformed as a whole, and converted into the form of the function s2. The specific process is as follows:

Take the logarithm on both sides of s1 and s2 at the same time, the expression is as follows

Transform both sides of Equation 1 to make it close to Equation 2, the expression is as follows:

Make it consistent with the form of the propagation and dissipation function of the scattered point, then the value on the right side of the deformed formula is the logarithm of the corrected all accurate dissipation, and taking the exponent, the new total dissipation qt(x), The expression is as follows:

By fusing the spectral and color data of foam scattered points, the accurate atmospheric dissipation qt(x) of all pixels in an image is obtained

(6) According to the sorting order in step (3), restore the dissipation value of different pixels in qt(x) to the original order, and obtain the accurate dissipation matrix t(x) of all pixels of the foam image;

Step 8: According to the image propagation influence model, I(x)=S(x)t(x)+A(1-t(x)), the media propagation dissipation t(x) of all pixels is known, and Atmospheric light value A, then the accurate color data S(x) of all pixels in the image can be obtained, the expression is as follows:

In order to measure the accuracy of the model, two sets of images in different industrial environments were collected, and color correction was performed on the two sets of images respectively. Comparing the two sets of corrected data with the original image data without interference, the RGB value of the same position of the image is close to that of the interference-free image. It shows that the calibration process is effective and achieves good results. Comparing the two groups of images with different disturbances, the RGB color values at the same position of the corrected images are similar. It means that under different interference conditions, the correction model can correct the image to a more accurate level, and the robustness and reliability are better.

Compared with the previous data analysis method to infer the modeling process of the color change trend, the present invention simulates the influence process of the industrial site environmental factors on the foam color by using the mechanism modeling method from the perspective of light propagation mode and factory site environmental interference, and establishes a color correction equation. This equation can solve the color change of each pixel affected by the environment, and the correction accuracy is high. At the same time, the device selects spectral equipment to collect image information, utilizes the non-dependence of spectral equipment on the environment, integrates the light propagation mechanism and spectral analysis method, and more accurately solves the parameters of the color change model, getting rid of the need for a large amount of color data in the past. requirements, improving the color correction accuracy and correction effect.

Claims (6)

1. a kind of froth images color calibration method based on Spectrum Data Fusion, it is characterised in that as follows comprising step：

Step 1：Source light is fixed and without under fog disturbed condition, gathers colour atla different color blocks spectroscopic data and camera color Data, establish the spectrum based on the camera characteristics-camera color space conversion model；

Step 2：Industry spot, camera gather foam raw image data I (x) in flotation cell, it is pending to be uploaded to host computer；

Step 3：Using visible light system, the spectral reflectance information of foam scatterplot in flotation cell is gathered, according in step 1 Color space conversion model, the spectral color information of foam is changed into camera RGB color numerical representation method；

Step 4：According to atmospheric scattering theory and environment light model, foam color variation model is established；

Step 5：The minimum value of each pixel triple channels of original image I (x) is taken, minimum color value image is formed, to the image Each pixel of matrix carries out minimum filtering operation processing, obtains its dark channel image, and calculation medium propagates dissipative function；

Step 6：Using being gathered in step 3 and being transformed into the foam scatterplot spectral color value of rgb space, correcting medium propagates consumption Function is dissipated, solves Accurate color dissipation amount of the image per pixel；

Step 7：Accurate color dissipation amount and original foam view data are substituted into color change model, it is determined that after correction Sharp picture color data.

2. image color correction method according to claim 1, it is characterised in that color space conversion model in step 1 Foundation, process includes：

Step 11：Under D65 light sources, using each color lump center of standard color card as midpoint, using 5*10pix as window size, industrial camera Image RGB data is gathered, color lump color data values rgb is used as after average_n, while gather different color blocks spectral reflectivity k_n, its Middle n=1,2...140；

Step 12：Standard is defined as with CIE1931, using different color blocks reflectivity data, calculates the colourity seat for trying to achieve each color lump Mark (x_n,y_n,z_n), with position in cone color space where chromaticity coordinate, to characterize the distribution of color situation of foam；

Step 13：With chromaticity coordinate x, y, z are least unit, establish ten multinomial models of different color blocks, and expression formula is as follows：

It is respectively dependent variable with camera RGB color data, expression formula is as follows with polynomial items for independent variable：

The multiple linear regression model of three passages is established respectively, and expression formula is as follows：

E is each term coefficient of multiple regression equation, model realization conversion of the chromaticity coordinate space to camera color space, times Meaning spectral color information can be characterized by camera color rgb space, i.e. rgb=xyz*E.

3. image color correction method according to claim 1, it is characterised in that foam color change is established in step 4 Model, process include：

Step 41：According to light propagate and color formation basic theory analysis, influence of the environment to color be divided into reflection loss loss and Environment light model two parts, color combining produce principle and atmospherical scattering model, establish Medium Propagation image and change model, expression Formula is as follows：

I (x)=S (x) e^-βd(x)+A(1-e^-βd(x)) (4)

S (x) is the color of image being not affected by during interference, that is, the sharp picture color after correcting, each pixel in x representative graphs, A For site environment air light value, according to atmospheric reflectance model, S (x) should be atmosphere light and scene albedo ρ product, and scene is anti- Physical reflection of the object to light, e are solely dependent upon according to rate ρ^-βd(x)For Medium Propagation dissipative function；

Step 42：Define t (x)=e^-βd(x)For Medium Propagation dissipative function, then color change model simplification is expression formula：

I (x)=S (x) t (x)+A (1-t (x)) (5).

4. image color correction method according to claim 1, it is characterised in that Medium Propagation is corrected in step 6 and is dissipated The process of function includes：

Step 61：According to dark primary priori, every Medium Propagation dissipation amount t of original image is solved₁(x), expression formula is as follows：

J^darkFor the dark channel image of the original image obtained in step 5；

Step 62：The multigroup foam spectrum RGB color data (r measured in step 3_n,g_n,b_n), as accurate in the case of this Foam physics color attribute, repeat step five can seek the accurate dissipation amount at each group foam scatterplot, and expression formula is as follows：

Step 63：According to Medium Propagation function t (x)=e^-βd(x), the dissipation amount t (x) that image all pixels point is calculated is pressed Ascending order arranges, and draws an exponential curve s1；According to this curve, each term coefficient of the exponential function is tried to achieve in fitting, and expression formula is as follows：

Step 64：By the accurate dissipation amount t of each group tried to achieve in step 62₂(n), arranged also according to ascending order, and fit accurate consumption Function is dissipated, expression formula is as follows：

Step 65：Compare two groups of dissipative functions, curve s1 is reversed, move, zoomed in or out, make it infinitely press close to dissipate Curve s2, i.e., function s1 is subjected to overall deformation, translates into function s2 form, correct the dissipation amount in whole s1, specifically Process is as follows：

S1, s2 take the logarithm on both sides simultaneously, and expression formula is as follows：

Formula 10,11 is deformed, it is as follows to obtain expression formula：

It is allowed to consistent with the Medium Propagation dissipative function form of scatterplot, then formula right scale is all accurate after correction after deforming The logarithm of true dissipation amount, to its fetching number, then new whole dissipation amount qt (x), expression formula are as follows：

Step 66：According to the clooating sequence in step 62, different pixels point dissipative value in qt (x) is recovered into original order, obtained To the accurate dissipation moment matrix t (x) of all pixels point of froth images.

5. image color correction method according to claim 1, it is characterised in that it is accurate after correction to be determined in step 7 Image color data process is：According to color change model, I (x)=S (x) t (x)+A (1-t (x)), it is known that atmosphere light A and consumption Amount t (x) is dissipated, obtains the Accurate color data of image, S (x) expression formulas are as follows：

6. a kind of froth images color correction device based on Spectrum Data Fusion, it is characterised in that including with lower module：

Image camera color acquisition module, for gathering the camera RGB color of froth images, module is gathered whole using industrial camera The RGB color information of width image, and it is pending that color data is reached into host computer；

Spectral information acquisition module, for gathering the scatterplot spectral reflectivity information of diverse location in the range of image-region, spectrum Testing image is divided into different zones by acquisition module, using visible ray micro spectrometer, gathers the scatterplot light in each region Compose reflectivity information；

Data processing module, for handling the camera RGB information and spectral information of image, camera is adopted using the model established The raw image data of collection is corrected, the color data after output calibration.