CN110108362A - The adaptive online test method of color difference and device based on SLIC super-pixel segmentation - Google Patents

Abstract

The present invention provides a kind of adaptive online test method of color difference based on SLIC super-pixel segmentation and devices, and wherein method includes: to carry out Image Acquisition to sample to be detected by online linear array CCD camera；The image collected is subjected to gaussian filtering process and is transformed into and device-independent uniform color space；Image is pre-processed using super-pixel segmentation method；Colorimetry is carried out using CIELDE2000 colour difference formula, obtains calculated result；Checkout result is compared with preset color difference threshold, exports comparison result.To plain color cloth acetes chinensis low efficiency existing for solving in current industry, the not high situation of the degree of automation saves many operands, greatly improves detection efficiency.

Description

Chromatic difference adaptive online detection method and device based on SLIC superpixel segmentation

technical field

The present invention relates to the technical field of color difference detection based on machine vision, in particular to a color difference adaptive online detection method and device based on SLIC superpixel segmentation.

Background technique

For textiles, color is an important feature that enriches its appearance. The quality of fabric dyeing is also a key factor in evaluating the quality of textile products. Therefore, the online detection of textile color difference is of great significance in the textile industry. The color difference reflects the comprehensive color perception difference of two dyed fabric samples in the three color values of chroma, lightness and hue, and is one of the important parameters for testing products. In the actual production process, the enterprise will use the color palette provided by the customer as the standard to produce textile products that meet the customer's needs. When customers check whether the dyeing quality of textile products meets the standard, they will compare the color difference and gloss of the product with the original sample. If the two are completely consistent, it is considered that the color of the textile product meets the requirements.

Although some advanced technical equipment has been introduced through the development of industrial technology, there is still a large gap in the detection level of the color difference of printed and dyed fabrics compared with foreign developed countries. There are still many enterprises that cannot fully realize the online detection of textile samples. Therefore, it is of great significance and practical value to study a new automatic detection system for textile and other related enterprises to adapt to the increasingly competitive modern market.

Contents of the invention

The present invention aims to provide a color difference adaptive online detection method and device based on SLIC superpixel segmentation that overcomes one of the above problems or at least partially solves any of the above problems.

In order to achieve the above object, the technical solution of the present invention is specifically realized in the following way:

One aspect of the present invention provides a color difference adaptive online detection method and device based on SLIC superpixel segmentation, including: collecting images of the samples to be detected through an online line array CCD camera; performing Gaussian filtering on the collected images and converting them to To a uniform color space that has nothing to do with the device; preprocess the image with the superpixel segmentation method; use the CIELDE2000 color difference formula to calculate the color difference, and get the calculation result; compare the settlement result with the preset color difference threshold, and output the comparison result.

Among them, performing Gaussian filtering on the collected image and converting it to a uniform color space that has nothing to do with the device includes: cutting the collected image to obtain the image to be processed; setting conversion parameters according to different lighting objects and elevation angles; the image to be processed Perform Gaussian filter processing and gamma correction; perform color space conversion on the corrected image according to the conversion parameters.

Among them, using the superpixel segmentation method to preprocess the image includes: dividing the image into a preset number of superpixels, setting seed points in each superpixel; calculating Gradient value, move the seed point to the position with the smallest gradient value and record it as a new seed point; search for similar pixels in the 2S×2S area centered on the new seed point, perform clustering, and assign class labels to similar pixels ;Measure the distance of each searched similar pixel point, determine the color distance and spatial distance from each similar pixel point to the seed point; determine the color feature value of each superpixel in the image, and the color feature value of each ordinary pixel and all superpixel color feature values of the image.

Among them, the CIELDE2000 color difference formula is used to calculate the color difference, and the calculation results include: calculating the lightness L, chroma a, chroma b, and heart chroma C _ab in the L*a*b* color space; calculating L', a'b ', the hue h' _ab and the adjustment factor G of the a* axis in the CIEL*a*b* color space; calculate the lightness difference ΔL, the chroma difference ΔC _ab , the hue difference ΔH'_ab; calculate the weight function S _L , S _C , S _H and rotation functions R _T , R _C ; determine the correction coefficients K _L , K _C , K _H ; through the formula compute to obtain the computed result.

Another aspect of the present invention provides a chromatic aberration adaptive online detection device based on SLIC superpixel segmentation, including: an acquisition module, which is used to collect images of samples to be detected through an online linear array CCD camera; a conversion module, which is used to collect The obtained image is processed by Gaussian filtering and converted to a device-independent uniform color space; the preprocessing module is used to preprocess the image using the superpixel segmentation method; the calculation module is used to calculate the color difference using the CIELDE2000 color difference formula to obtain the calculated Result; a comparison module, configured to compare the settlement result with a preset color difference threshold, and output the comparison result.

Among them, the conversion module specifically performs Gaussian filter processing on the collected images and converts them to a uniform color space that has nothing to do with the device through the following methods: the conversion module is specifically used to crop the collected images to obtain images to be processed; Different settings of conversion parameters and elevation angles; Gaussian filtering and gamma correction are performed on the image to be processed; color space conversion is performed on the corrected image according to the conversion parameters.

Wherein, the preprocessing module specifically uses the superpixel segmentation method to preprocess the image in the following manner: the preprocessing module is specifically used to divide the image into a preset number of superpixels, and set a seed point in each superpixel; calculate The gradient values of all pixels in the 3×3 neighborhood of each seed point, move the seed point to the position with the smallest gradient value and record it as a new seed point; search for similar pixels in the 2S×2S area centered on the new seed point , perform clustering, and assign class labels to similar pixels; perform distance measurement on each searched similar pixel, and determine the color distance and spatial distance from each similar pixel to the seed point; determine each superpixel in the image The color eigenvalues of , the color eigenvalues of each ordinary pixel, and the color eigenvalues of all superpixels of the image.

Among them, the calculation module specifically uses the CIELDE2000 color difference formula to calculate the color difference in the following manner, and obtains the calculation result: the calculation module is specifically used to calculate the lightness L, chroma a, chroma b, and heart color in the L*a*b* color space. degree C _ab ; calculate L', a'b', hue h' _ab and adjustment factor G of a* axis in CIEL*a*b* color space; calculate lightness difference ΔL, chroma difference ΔC _ab , hue difference ΔH'_ab; Calculate the weight function S _L , S _C , S _H and the rotation function R _T , R _C ; determine the correction coefficient K _L , K _C , K _H ; through the formula compute to obtain the computed result.

It can be seen that, through the SLIC superpixel segmentation-based color difference adaptive online detection method and device provided by the embodiment of the present invention, it solves the situation in the current industry that the color difference detection efficiency of plain cloth is low and the degree of automation is not high. , which saves a lot of computation and greatly improves the detection efficiency.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings that need to be used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For Those of ordinary skill in the art can also obtain other drawings based on these drawings on the premise of not paying creative work.

Fig. 1 is the flowchart of the color difference adaptive online detection method based on SLIC superpixel segmentation provided by the embodiment of the present invention;

Fig. 2 is a specific flow chart of the chromatic aberration adaptive online detection method based on SLIC superpixel segmentation provided by the embodiment of the present invention;

FIG. 3 is a schematic structural diagram of a color difference adaptive online detection device based on SLIC superpixel segmentation provided by an embodiment of the present invention.

Detailed ways

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided for more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

Fig. 1 shows the flow chart of the chromatic aberration adaptive online detection method based on SLIC superpixel segmentation provided by the embodiment of the present invention, referring to Fig. 1, the chromatic aberration adaptive online detection method based on SLIC superpixel segmentation provided by the embodiment of the present invention, include:

S1, image acquisition of the sample to be detected is performed by an online linear array CCD (Charge Coupled Device, charge coupled device image sensor) camera.

Specifically, the sample to be tested is a plain cloth, and the detection method provided by the present invention is a detection of the color difference of the plain cloth.

S2. Perform Gaussian filter processing on the acquired image and convert it into a device-independent uniform color space.

As an optional implementation of the embodiment of the present invention, this step may specifically include: clipping the collected image to obtain the image to be processed; setting conversion parameters according to different illumination bodies and elevation angles; performing Gaussian filtering on the image to be processed, And perform gamma correction; perform color space conversion on the corrected image according to the conversion parameters.

During specific implementation, the steps of performing Gaussian filtering on the acquired image and converting it to a device-independent uniform color space can be performed in the following manner:

S21, first perform Gaussian linear smoothing filter processing on the collected images, and use the formula Gaussian linear filtering is performed, where σ represents the parameters of the Gaussian distribution, the value of σ will ultimately determine the width of the filter, and the value of σ is positively correlated with smoothing.

S22: convert the processed image from RGB to CIELAB space:

Specifically use the following formula:

Among them, the tristimulus values X _n , Y _n , and Z _n in this step have six modes to choose from according to the illumination body and elevation angle, and the parameter table is as follows:

Among them, if the value of X/X _n , Y/Y _n , Z/Z _n is greater than

Or, more specifically, the step of performing Gaussian filtering on the acquired image and converting it to a device-independent uniform color space can be performed as follows:

S21: Crop the acquired original RGB image into an image to be processed with a size of 400×400 pixels.

S22: System preset for the lighting halo environment of the detection system: Different light sources and lighting angles have certain influence on the color space conversion parameters in the later stage, so it needs to be pre-selected and set according to the six regular meeting modes currently calibrated. Among them, the six regular meeting modes can be found in Table 1.

S23: Use a Gaussian filter to smooth the image and suppress noise on the above-mentioned 400×400 pixel image, and then perform gamma correction on it, in order to improve the contrast of the color image, thereby avoiding a large number of small areas generated by SLIC superpixel segmentation in the later stage; Gaussian filtering is a process of weighted average of the entire image, and the value of each pixel is obtained by weighted average of itself and other pixel values in the neighborhood. The specific operation of Gaussian filtering is: use a template (or convolution, mask) to scan each pixel in the image, and use the weighted average gray value of the pixels in the neighborhood determined by the template to replace the value of the pixel in the center of the template. Gaussian smoothing filters are very effective at suppressing noise that follows a normal distribution.

Among them, gamma correction is mainly to nonlinearize the original RGB image, and convert the red, green and blue three channels through the following formula;

Among them, the value ranges of the r, g, and b channels are all [0, 255].

S24: Perform color space conversion on the 400×400 pixel image processed through the above steps, RGB-XYZ-CIELAB.

If the value of X/X _n , Y/Y _n , Z/Z _n is greater than

The values of X _n , Y _n , and Z _n are according to the values set in step S22.

S3, the image is preprocessed using a superpixel segmentation method.

As an optional implementation of the embodiment of the present invention, this step may specifically include: dividing the image into a preset number of superpixels, setting seed points in each superpixel; calculating Gradient values of all pixels, move the seed point to the position with the smallest gradient value and record it as a new seed point; search for similar pixel points in the 2S×2S area centered on the new seed point, perform clustering, and create similar pixels Point assignment class label; measure the distance of each searched similar pixel point, determine the color distance and spatial distance from each similar pixel point to the seed point; determine the color feature value of each super pixel in the image, each ordinary pixel and all superpixel color feature values of the image.

During specific implementation, the step of using the superpixel segmentation method to preprocess the image can be performed in the following manner:

S31: Initialize the seed point. According to the set number of superpixels, the seed points are evenly distributed in the image. Assuming that the picture has M pixels, and the preset is divided into K superpixels of the same size, then the size of each superpixel is M/K, and the distance between adjacent seed points is approximately

S32: Calculate the gradient values of all pixels in the 3×3 neighborhood of the seed point, move the seed point to the position with the smallest gradient value, and record it as a new seed point. This can prevent the seed point from falling near the contour boundary and affect the clustering effect.

S33: Search for similar pixel points in a 2S×2S region centered on the seed point, perform clustering, and assign class labels to them.

S34: Measure the distance of each searched pixel point, including the color distance and spatial distance from each pixel point to the seed point, the calculation method is as follows:

In the formula, d _lab is the color distance, d _s is the space distance, S′ is the distance measure between two pixels, D is the distance between seed points, and m is the balance parameter.

S35: Set the color feature value of each superpixel in the segmented image V _i = (L, B, A), i=1, 2, 3, ..., K, the color feature value of each common pixel For v _i = (l, a, b), j = 1, 2, 3, ..., M/K, then The superpixel color feature values of the entire image are represented by a one-dimensional vector [V _i =(L, A, B) ^T ], i=1, 2, 3, . . . , K.

S4, using the CIELDE2000 color difference formula to calculate the color difference, and obtain the calculation result.

As an optional implementation of the embodiment of the present invention, this step may specifically include: calculating lightness L, chroma a, chroma b, and heart chroma C _ab in the L*a*b* color space; calculating L', a'b', hue h' _ab and adjustment factor G of a* axis in CIEL*a*b* color space; calculate lightness difference ΔL, chroma difference ΔC _ab , hue difference ΔH'_ab; calculate weight function S _L , S _C , S _H and rotation functions R _T , R _C ; determine the correction coefficients K _L , K _C , K _H ; through the formula compute to obtain the computed result.

During specific implementation, the steps of using the CIELDE2000 color difference formula to calculate the color difference and obtain the calculation result can be performed in the following manner:

Specifically, the following formula can be used for color difference calculation:

The calculation steps in the formula are as follows:

S41: Calculate lightness L, chroma a, chroma b, and heart chroma C _ab in the L*a*b* color space.

S42: Calculate L', _a'b ', hue h'ab and the adjustment factor G of the a* axis in the CIEL*a*b* color space.

In the formula is the arithmetic mean of the test image C _ab1 and the standard image C _ab2 .

S43: Calculate lightness difference ΔL, chroma difference ΔC _ab , and hue difference ΔH′ _ab .

S44: Calculate weighting functions S _L , S _C , S _H and rotation functions R _T , R _C .

S45: where K _L , K _C , and K _H are correction coefficients determined according to actual usage conditions. Under CIE standard observation conditions, the default value can be set as K _L =K _C =K _H =1.

S5, comparing the settlement result with the preset color difference threshold, and outputting the comparison result.

Specifically, the calculation result is compared with the preset threshold value in the system. If it meets the requirements, it is judged that the product meets the standard, and if it does not meet the requirements, it is judged that the product does not meet the standard. That is, if the calculation result meets the requirements, it indicates that the sample to be tested meets the requirements, that is, the color difference meets the requirements, and if the calculation result does not meet the requirements, it indicates that the sample to be tested does not meet the requirements, that is, the color difference does not meet the requirements.

It can be seen that, aiming at the problem of low accuracy and slow speed of on-line detection of cloth color difference in the cloth color printing process, the detection method of the present invention is proposed. After collecting the items to be detected and preprocessing the image, based on the idea of superpixels, a simple linear iterative clustering (SLIC, simple linear iterative clustering) algorithm is used to cluster adjacent pixels with similar characteristics to form a compact, Approximately uniform pixel blocks, each pixel block is a superpixel. Superpixels are used to replace multiple similar pixels in the pixel block, and the color features of the standard image and the image to be detected are extracted respectively. Use the appropriate color difference formula for color difference calculation. On the basis of ensuring the accuracy of the detection results, the method can effectively reduce the amount of data calculation and improve the detection efficiency.

Hereinafter, taking Fig. 2 as an example, the color difference adaptive online detection method based on SLIC superpixel segmentation provided by the embodiment of the present invention is specifically described, but the present invention is not limited thereto:

The CCD camera collects images of the sample to be tested (solid color cloth);

Set the light source conditions;

The Gaussian filter is used to process the collected images;

Perform superpixel segmentation on the processed image;

Extract superpixel color features;

CIEDE2000 color difference formula to calculate color difference;

Compare the calculated color difference value with the preset threshold, and judge whether the sample to be tested is qualified.

It can be seen that the color difference adaptive online detection method based on SLIC superpixel segmentation provided by the embodiment of the present invention integrates the detection environment of light source conditions and angles in various scenarios, and has a wide range of applications and a relatively high degree of automation. Meet the testing requirements of most of the current plain fabrics. Moreover, the detection method provided by the embodiment of the present invention has the advantages of faster speed, better accuracy, and improved detection efficiency compared with the current pixel-by-pixel color difference detection method that is widely used.

Fig. 3 shows a schematic structural diagram of a chromatic aberration adaptive online detection device based on SLIC superpixel segmentation provided by an embodiment of the present invention, which is applied to the above-mentioned chromatic aberration based on SLIC superpixel segmentation Adaptive online detection method. The following is only a brief description of the structure of the chromatic aberration adaptive online detection device based on SLIC superpixel segmentation. For other unfinished matters, please refer to the above-mentioned related descriptions of the chromatic aberration adaptive online detection method based on SLIC superpixel segmentation. , which will not be repeated here. Referring to Fig. 3, the chromatic aberration adaptive online detection device based on SLIC superpixel segmentation provided by the embodiment of the present invention includes:

The collection module 301 is used to collect images of samples to be detected by an online linear array CCD camera;

A conversion module 302, configured to perform Gaussian filter processing on the acquired image and convert it to a device-independent uniform color space;

A preprocessing module 303, configured to preprocess the image using a superpixel segmentation method;

Calculation module 304, for adopting CIELDE2000 color difference formula to carry out color difference calculation, obtain calculation result;

The comparison module 305 is configured to compare the settlement result with the preset color difference threshold, and output the comparison result.

As an optional implementation of the embodiment of the present invention, the conversion module 302 performs Gaussian filter processing on the acquired image and converts it to a device-independent uniform color space in the following manner: the conversion module 302 is specifically used to convert the acquired image The image is cropped to obtain the image to be processed; the conversion parameters are set according to the different illumination bodies and elevation angles; the image to be processed is processed by Gaussian filtering and gamma correction; the corrected image is converted into color space according to the conversion parameters.

As an optional implementation of the embodiment of the present invention, the preprocessing module 303 specifically uses the superpixel segmentation method to preprocess the image in the following manner: the preprocessing module 303 is specifically used to divide the image into a preset number of superpixels , set a seed point in each superpixel; calculate the gradient value of all pixels in the 3×3 neighborhood of each seed point, move the seed point to the position with the smallest gradient value and record it as a new seed point; use the new seed point as Search for similar pixels in the central 2S×2S area, perform clustering, and assign class labels to similar pixels; measure the distance of each searched similar pixel, and determine the color of each similar pixel to the seed point Distance and spatial distance; determine the color feature value of each superpixel in the image, the color feature value of each ordinary pixel, and the color feature value of all superpixels in the image.

As an optional implementation of the embodiment of the present invention, the calculation module 304 specifically uses the CIELDE2000 color difference formula to perform color difference calculation in the following manner to obtain the calculation result: the calculation module 304 is specifically used to calculate the lightness in the L*a*b* color space L, chroma a, chroma b, heart chroma C _ab ; calculate L', a'b', hue h' _ab and adjustment factor G of a* axis in CIEL*a ^*b * color space; calculate lightness difference ΔL, chroma difference ΔC _ab , hue difference ΔH′ _ab ; calculate weight functions S _L , S _C , S _H and rotation functions R _T , R _C ; determine correction coefficients K _L , K _C , K _H ; through the formula compute to obtain the computed result.

It can be seen that the color difference adaptive online detection device based on SLIC superpixel segmentation provided by the embodiment of the present invention integrates the detection environment of light source conditions and angles in various scenarios, and has a wide range of applications and a relatively high degree of automation. Meet the testing requirements of most of the current plain fabrics. Moreover, the detection method provided by the embodiment of the present invention has the advantages of faster speed, better accuracy, and improved detection efficiency compared with the current pixel-by-pixel color difference detection method that is widely used.

The above are only examples of the present application, and are not intended to limit the present application. For those skilled in the art, various modifications and changes may occur in this application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims (8)

1. A color difference adaptive online detection method based on SLIC superpixel segmentation, characterized in that, comprising: Image acquisition of the sample to be tested through the online linear array CCD camera; Gaussian filtering is performed on the collected image and converted to a uniform color space that has nothing to do with the device; The image is preprocessed by using a superpixel segmentation method; Use the CIELDE2000 color difference formula to calculate the color difference and get the calculation result; The settlement result is compared with the preset color difference threshold, and the comparison result is output. 2. The method according to claim 1, wherein said performing Gaussian filter processing on the collected image and converting it to a device-independent uniform color space comprises: Crop the collected image to obtain the image to be processed; Set the conversion parameters according to the different lighting bodies and elevation angles; Performing Gaussian filter processing on the image to be processed, and performing gamma correction; Perform color space conversion on the corrected image according to the conversion parameters. 3. method according to claim 1, is characterized in that, described adopting superpixel segmentation method to carry out preprocessing to described image comprises: The image is divided into a preset number of superpixels, and a seed point is set in each superpixel; Calculate the gradient values of all pixels in the 3×3 neighborhood of each seed point, move the seed point to the position with the smallest gradient value and record it as a new seed point; Search for similar pixel points in the 2S×2S area centered on the new seed point, perform clustering, and assign class labels to the similar pixel points; Perform distance measurement on each searched similar pixel point, and determine the color distance and spatial distance from each similar pixel point to the seed point; The color feature value of each superpixel in the image, the color feature value of each common pixel, and the color feature values of all superpixels in the image are determined. 4. method according to claim 1, is characterized in that, described adopting CIELDE2000 color difference formula to carry out color difference calculation, obtains calculation result and comprises: Calculate lightness L, chroma a, chroma b, heart chroma C _ab in the L*a*b* color space; Compute L', _a'b ', hue h'ab and the adjustment factor G for the a* axis in the CIEL*a*b* color space; Calculate lightness difference ΔL, chroma difference ΔC _ab , hue difference ΔH′ _ab ; Calculation of weight functions S _L , S _C , S _H and rotation functions R _T , R _C ; Determine the correction coefficients K _L , K _C , K _H ; by formula compute to obtain the computed result. 5. A color difference adaptive online detection device based on SLIC superpixel segmentation, characterized in that, comprising: The acquisition module is used for image acquisition of the sample to be detected by an online linear array CCD camera; A conversion module, configured to perform Gaussian filtering on the acquired image and convert it to a device-independent uniform color space; A preprocessing module, configured to preprocess the image using a superpixel segmentation method; The calculation module is used to calculate the color difference by using the CIELDE2000 color difference formula to obtain the calculation result; The comparison module is used to compare the settlement result with the preset color difference threshold and output the comparison result. 6. The device according to claim 5, wherein the conversion module performs Gaussian filtering on the collected image and converts it to a device-independent uniform color space in the following manner: The conversion module is specifically used to cut the collected image to obtain the image to be processed; set the conversion parameters according to the different illumination body and elevation angle; perform Gaussian filter processing on the image to be processed, and perform gamma correction; The resulting image undergoes color space conversion according to the conversion parameters. 7. The device according to claim 5, wherein the preprocessing module specifically adopts a superpixel segmentation method to preprocess the image in the following manner: The preprocessing module is specifically used to divide the image into a preset number of superpixels, set a seed point in each superpixel; calculate the gradient of all pixels in the 3×3 neighborhood of each of the seed points value, move the seed point to the position with the smallest gradient value and record it as a new seed point; search for similar pixel points in the 2S×2S area centered on the new seed point, perform clustering, and Assigning class labels to pixel points; performing distance measurement on each searched similar pixel point, determining the color distance and spatial distance from each similar pixel point to the seed point; determining the color feature value of each superpixel in the image , the color feature value of each common pixel and all superpixel color feature values of the image. 8. device according to claim 5, is characterized in that, described calculation module specifically adopts CIELDE2000 color difference formula to carry out color difference calculation in the following manner, obtains calculation result: The calculation module is specifically used to calculate lightness L, chroma a, chroma b, heart chroma C _ab in the L*a*b* color space; calculate L', a'b', hue h' _ab and The adjustment factor G of the a* axis in the CIEL*a*b* color space; calculate the lightness difference ΔL, the chroma difference ΔC _ab , the hue difference ΔH′ _ab ; calculate the weight function S _L , S _C , S _H and the rotation function R _T , R _C ; determine the correction coefficients K _L , K _C , K _H ; through the formula compute to obtain the computed result.