CN115601412A - Method for detecting crystalline polymer powder impurities based on multispectral image processing - Google Patents
Method for detecting crystalline polymer powder impurities based on multispectral image processing Download PDFInfo
- Publication number
- CN115601412A CN115601412A CN202211272598.0A CN202211272598A CN115601412A CN 115601412 A CN115601412 A CN 115601412A CN 202211272598 A CN202211272598 A CN 202211272598A CN 115601412 A CN115601412 A CN 115601412A
- Authority
- CN
- China
- Prior art keywords
- image
- detected
- impurities
- crystalline polymer
- polymer powder
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10036—Multispectral image; Hyperspectral image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30242—Counting objects in image
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a method for detecting impurities in crystalline polymer powder based on multispectral image processing, which comprises the following steps: shooting a crystalline polymer powder sample to be detected to obtain a multispectral image as an original image, and selecting a partial image channel or an enhanced original image as an image to be detected; carrying out thresholding segmentation on an image to be detected to obtain a binary image; processing the binary image by adopting a connected region analysis algorithm to obtain a plurality of adjacent image regions with the same pixel intensity as connected regions; and screening impurity regions according to the geometrical characteristics of the impurity particles and the distribution condition of impurities in the powder, and calculating the quantity and the area of the impurities. The method can accurately and efficiently detect the impurities in the powder, reduce the influence of human factors in the impurity detection process, improve the stability of the impurity detection result, and is widely suitable for powder impurity detection of various crystalline polymers.
Description
Technical Field
The invention relates to the technical field of digital image processing, in particular to a method for detecting powder impurities of a crystalline polymer based on multispectral image processing.
Background
The crystalline polymer is generally in the form of powder or pellets, and the impurities of the crystalline polymer are usually detected manually and recognized by a skilled inspector with the aid of the naked eye. However, the manual detection has the problems of difficult personnel training, slow detection speed, different personnel standards, great influence of environmental conditions, physical conditions and psychological factors on the detection precision of the same personnel and the like. The existing detection device has a good detection result on the granules with larger particle sizes, but is limited by equipment and a detection method, and the precision cannot reach the manual level when the powder with smaller particle sizes is detected. Therefore, it is necessary to develop a high-precision impurity detection method.
The crystalline regions of the crystalline polymer are randomly oriented and the direction of reflected light is random, and a nearly white color is exhibited in the visible light band. The crystalline polymer is decomposed under certain conditions, and the degree of crystallinity and the degree of aromaticity of the polymer are reduced by the decomposition reaction. The decrease in crystallinity weakens the light reflection capability and the increase in the degree of aromatization increases the absorption capability of long-wave light. As the degree of decomposition increases, the light absorption capacity of the crystalline polymer gradually extends from ultraviolet to infrared bands, and a color change process of white-light yellow-deep yellow-red-black is exhibited in visible bands. Therefore, the degree of decomposition of the crystalline polymer can be estimated from the change in the spectral characteristics of the crystalline polymer in a specific wavelength band and used as a criterion for discriminating impurities.
CN114266746A discloses a resin impurity detection method and system based on computer vision, relating to the technical field of computer vision. The method comprises the following steps: acquiring a primary overall image; processing the primary integral image to obtain a target image; performing impurity contour analysis on the target image, and calculating to obtain the number of contour pixels of the target area, the contour area of the target area and the contour perimeter of the target area; and generating an impurity detection result according to the number of the contour pixels of the target region, the contour area of the target region, the contour perimeter of the target region and a preset contrast parameter threshold, so that the working intensity of operators can be greatly reduced, and the detection efficiency and the detection accuracy can be improved. However, in the invention, the original image must be converted into the gray image, and the image to be detected is selected according to the average brightness of the gray image, so that the detection accuracy is not high enough.
CN103512883A discloses a method and system for detecting geometrical characteristics of impurities in polyolefin material based on digital image processing, which aims to detect impurities in material with smaller size, mainly based on some morphological characteristics of impurities in the image, so that a watershed method is adopted to process fiber image. The method has the main limitations that microscopic equipment is required to be used for shooting, and the processing equipment is limited, so that the method has the defects of limited sample amount in single detection, high popularization cost, capability of detecting some microscopic impurities only according to morphological characteristics, and incapability of judging whether the impurities consist of different materials.
Therefore, the prior art still lacks a method for quickly and effectively detecting impurities in the polymer, in particular an accurate and stable detection means aiming at the impurities in the crystalline polymer.
Disclosure of Invention
Based on the current situation that the crystalline polymer powder in the prior art lacks an effective impurity detection technology, the invention provides a crystalline polymer powder impurity detection method based on multispectral image processing, so as to realize efficient, accurate and stable detection of impurities in the crystalline polymer powder.
In order to achieve the purpose, the invention adopts the technical scheme that:
a method for detecting impurities in crystalline polymer powder based on multispectral image processing comprises the following steps:
s1, acquiring an image to be detected: shooting a crystalline polymer powder sample to be detected by using image acquisition equipment to obtain a multispectral image as an original image, and selecting a partial image channel of the original image or the enhanced original image as the image to be detected;
s2, segmenting the image to be detected: carrying out thresholding segmentation on an image to be detected to obtain a binary image;
s3, calculating the quantity and area of impurities: processing the binary image by adopting a connected region analysis algorithm to obtain a plurality of adjacent image regions with the same pixel intensity as connected regions;
and S4, screening out standard-meeting connected regions from the plurality of connected regions obtained in the S3 as impurity regions according to the geometrical characteristics of the impurity particles and the distribution condition of the impurities in the powder, and calculating the quantity and the area of the impurities.
The standard in step S4 is established according to the geometric characteristics of the impurity particles and the impurity distribution condition of the powder, and the parameter types include the particle diameters of normal powder and typical impurity particles, the aspect ratio of the connected region, the impurity distribution density in a certain size region, the color value range of the typical impurity in the image, and the position of the connected region in the image.
In some embodiments of the present invention, the process of acquiring an original image comprises the steps of:
s11, selecting a spectrum sensitive waveband: selecting a wave band with large spectral characteristic difference as a spectral sensitive wave band and a wave band with small spectral characteristic difference as an inert wave band according to the spectral characteristics of the normal sample and the impurities;
s12, selecting a camera or a camera combination with a sensor working waveband covering the spectral sensitivity waveband as image acquisition equipment by the image acquisition equipment;
s13, shooting an original image: spreading a to-be-detected crystalline polymer powder sample on a plane, and shooting the to-be-detected sample by using image acquisition equipment under the condition of sufficient illumination to obtain a multispectral image as an original image;
the degree of crystallinity of the polymer is reduced and the degree of aromatic hydrocarbon is increased after the crystalline polymer is decomposed. The decrease in crystallinity weakens the reflectivity of light and the increase in the degree of aromatization increases the absorption of long-wave light. As the degree of decomposition deepens, the light absorption capacity of the crystalline polymer gradually extends from ultraviolet to infrared bands, and a color change process of white-light yellow-dark yellow-red-black appears in visible light bands. Therefore, the degree of decomposition of the crystalline polymer can be estimated from the change in the spectral characteristics of the crystalline polymer in a specific wavelength band and used as a criterion for discriminating impurities.
According to the spectral characteristic difference degrees of the normal sample and impurities in various wave bands from infrared to ultraviolet of the spectrum, part of the wave bands with more obvious differences in the spectral sensitive wave bands are selected as the spectral sensitive wave bands, the signal-to-noise ratio of the image to be detected is enhanced, and the detection precision is improved.
Typically the degree of difference in spectral characteristics between the normal sample and the contaminant is obtained by scanning the uv-vis or nir absorption spectrum.
In the invention, a camera is adopted to photograph the surface of the sample, the main property influencing the image signal is the reflectivity of the material, and the reflectivity is greatly influenced by the absorption spectrum of the material under the condition that other conditions such as a powder preparation process, particle morphology and the like are not changed, so the spectral sensitivity waveband is selected according to the difference degree of the reflection spectrum characteristics of a normal sample and impurities.
In some embodiments of the present invention, the process of obtaining the image to be measured specifically includes the steps of:
s14, selecting image channels or combinations of channels corresponding to a plurality of sensitive bands from the original image as an image to be detected according to the spectrum sensitive band and the inert band; or selecting a weighted combination of image channels corresponding to the sensitive wave band and the inert wave band to enhance the original image and taking the enhanced original image as the image to be detected.
In some embodiments of the invention, the image acquisition device comprises a black and white camera, a visible light camera, an infrared camera, an ultraviolet camera, a hyperspectral camera, or any combination of the foregoing cameras; selecting image acquisition equipment according to the spectral characteristic difference of a normal sample and impurities, for example, selecting a visible light camera, a near ultraviolet camera, a hyperspectral camera or any combination thereof with a sensor working waveband covering the waveband if the spectrum sensitive waveband is a blue-violet light waveband;
in some embodiments of the present invention, the segmenting the image to be measured includes: and respectively carrying out thresholding segmentation on different areas of the image to be detected by adopting an absolute threshold or a relative threshold according to the uniformity of the brightness spatial distribution of the image to be detected.
Specifically, for the light source which is far in a point shape or the light source which is in a planar shape and the illumination intensity of each point is close, the brightness spatial distribution of the sample image pixels is uniform, and the light source and the photographing conditions of the detection scene, such as distance, angle and the like, are basically unchanged, the threshold segmentation is carried out by adopting an absolute threshold; otherwise, the relative threshold value is adopted for thresholding segmentation.
The image to be measured is preferably segmented using a relative threshold.
In some embodiments of the present invention, thresholding of the absolute threshold converts the image to be measured into a binary image by comparing the relative magnitude of the absolute threshold to the pixel intensity of each pixel.
In some embodiments of the present invention, the thresholding segmentation of the relative threshold calculates the difference between the pixel intensity of a single pixel and the weighted average of the pixel intensities of the neighborhood of a certain size, and converts the image to be measured into a binary image by comparing the relative threshold with the relative size of the difference.
The neighborhood diameter is at least 3 times the average particle size of the sample.
The size of the neighborhood needs to meet the sample size of normal distribution sampling, so that at least 3 times of particle size is taken as the diameter of the neighborhood; since the powder has a certain particle size distribution, the number of pixels occupied by the sample particles in the image is not completely fixed, and it is preferable to take 5 times the average particle size as the neighborhood diameter.
In some embodiments of the present invention, when the polymer powder sample is loose in bulk or the powder particle size is large, the low-pass filtering method is used to smooth the image to be detected before segmenting the image to be detected, so as to reduce the high-frequency noise generated by the image acquisition device or the powder gap in the image.
The low-pass filtering method includes, but is not limited to, mean filtering, gaussian filtering, median filtering, bilateral filtering, and the like, and combinations thereof.
In some embodiments of the present invention, when calculating the amount and area of the impurities, according to the geometric characteristics of the impurity particles and the distribution of the impurities in the powder, including but not limited to the geometric length, the cross-sectional area, the aspect ratio, the geometric position, the impurity distribution density of the powder and combinations thereof in the image of the crystalline polymer, a region meeting the standard is screened out from a plurality of connected regions obtained by the connected region analysis method as a connected region, and the amount and area of the impurities are calculated.
Compared with the prior art, the invention has the following beneficial effects:
the method for detecting the impurities in the crystalline polymer powder based on the multispectral image processing is designed according to the spectral characteristics of the crystalline polymer for the first time, utilizes the spectral difference between a normal sample and the impurities to selectively process image channels in the acquired multispectral image, combines a connected region analysis algorithm to process a binary image, has complete theoretical and experimental data support, can accurately and efficiently detect the impurities in the powder, reduces the influence of human factors in the impurity detection process, improves the stability of the impurity detection result, is the basis of the automatic technology for detecting the impurities in the crystalline polymer powder, and is widely applicable to the powder impurity detection of various crystalline polymers.
Drawings
FIG. 1 shows UV-visible absorption spectra of polyvinylidene chloride at different decomposition times.
FIG. 2 shows an original image and an image channel of a yellow impurity in example 1.
Fig. 3 is a pixel histogram of the yellow impurity image to be measured (fig. 2- (3)) in example 1.
Fig. 4 is a yellow impurity binary image of example 1.
Fig. 5 shows the original image of yellow impurities and the enhanced original image of example 2.
Fig. 6 is a pixel histogram of the yellow impurity image to be measured (fig. 5- (3)) in example 2.
Fig. 7 is a yellow impurity binary image of example 2.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. Those skilled in the art should understand that they can make modifications and equivalents without departing from the spirit and scope of the present invention, and all such modifications and equivalents are intended to be included within the scope of the present invention.
In the following examples, which are illustrated with reference to polyvinylidene chloride (PVDC) powder as a sample, a normal sample of PVDC powder should be recognized by the human eye as a color close to white, and impurities should be recognized by the human eye as yellow, red and black or a transition color of the aforementioned colors, according to the color development principle after decomposition of the crystalline polymer.
Reference is made to the UV-visible spectrum of the impurities of PVDC in the documents https:// doi.org/10.1016/0141-3910 (87) 90026-7, photo-chemistry of polymers of vinylidine chloride and vinyl chloride, part1 wavelet efficiencies on disportionation, as shown in FIG. 1.
Example 1
A method for detecting powder impurities of PVDC based on multispectral image processing comprises the following steps:
s1, acquiring an image to be detected
S11, selecting a sensitive wave band: according to the uv-vis spectrum of PVDC of fig. 1, it can be observed that as the decomposition time increases, these impurities exhibit a stronger absorption capacity in the near uv to blue wavelength band than the normal sample, and the degree of macroscopically yellow color increases, so that the wavelength range of 300nm to 500nm is selected as the spectrum sensitive wavelength band, and the wavelength range of 600nm to 800nm is selected as the spectrum inactive wavelength band.
S12, selecting an image acquisition device: as the visible light camera has a bluish purple waveband, the near ultraviolet camera has a near ultraviolet waveband, the hyperspectral camera has a bluish purple waveband and a near ultraviolet waveband, and the three cameras and the combination thereof can meet the conditions. The center of the working waveband of the blue light sensor of the common visible light camera is near 460nm and comprises a spectrum sensitive waveband; the center of the operating band of the red light sensor of a common visible light camera is near 610nm and comprises a spectrally inert band, and the embodiment uses the visible light camera as an image acquisition device.
S13, shooting an original image of the sample: the polyhaloolefin powder is spread out on a flat surface to smooth its outer surface. Under the condition of sufficient ambient light, a clear sample is shot by a visible light camera lens along the vertical direction close to the powder spreading plane at a proper distance matched with the focal length of the lens, and an obtained multispectral image is used as an original image, as shown in fig. 2- (1).
S14, acquiring an image to be detected: in the original image taken by the visible light camera, the red channel of the image corresponds to the spectral inert band, as shown in fig. 2- (2); the blue channel of the image corresponds to the spectrally sensitive band, as in fig. 2- (3). As can be seen from fig. 2, the normal sample and the yellow impurity show the largest difference in the blue channel of the original image, and therefore the blue channel of the image is selected as the image to be detected.
S2, segmenting the image to be detected
S21, low-pass filtering: the average particle size of the polyhaloolefin powder in the embodiment is about 0.3mm, and the particle size is larger, so that the image noise is reduced by adopting a Gaussian filtering method, and the detection accuracy is improved. The gaussian filter equation is as follows:
where x and y are two-dimensional coordinates of the pixel, I and j are offset coordinates from a point, I D (x, y) represents a pixel value of the image after noise reduction, I (x, y) represents a pixel value of the image to be measured at coordinates (x, y), G (I, j) represents a probability value of the two-dimensional normal distribution at a point (I, j) deviated from the origin, I (x + I, y + j) represents a neighborhood pixel value deviated from (I, j) with respect to the point (x, y), k represents a gaussian kernel radius, the size of the gaussian kernel should be close to the particle size, and the particle diameter in the image is about 6 pixels, so that k is 3, the gaussian kernel size is 7 × 7, and σ represents a standard deviation of the two-dimensional normal distribution, which is generally specified as
e denotes the base of the natural logarithm.
S22, obtaining a binary image by thresholding segmentation: in this embodiment, a relative threshold segmentation method is taken as an example, that is, the size of the threshold is adaptively adjusted according to the spatial distribution of image brightness, and this method is used to take the standard deviation of pixel values 2-5 adjacent pixels lower than that of adjacent pixels as possible impurity pixels, because impurities in the powder generally exhibit stronger light absorption capability than that of a normal sample, and therefore exhibit darker color. The relative threshold method formula is as follows:
T(x,y)＝f(x,y)+α×σ (4)
in the formula I c (x, y) represents the classification result of each pixel, T (x, y) represents the threshold of each pixel, f (x, y) is the weighted sum of the neighborhood pixels, α is a constant that is generally negative, σ represents the standard deviation of the image pixel values, μ is the mean of the image pixels to be measured, I is the average of the image pixels to be measured D (x + i, y + j) represents a neighborhood pixel value of the noise-reduced image at a position deviated from (i, j) with respect to the point (x, y), and n represents the number of neighborhood pixels, i.e., n = (2k + 1) 2 。
The pixel standard deviation σ calculated according to fig. 2- (3) and equation (6) is 7.59; obtaining the pixel histogram of fig. 2- (3) results in fig. 3, and assuming that the pixel values obey a gaussian distribution according to fig. 3, then taking 3 times the standard deviation as the threshold criterion, i.e., α takes-3, then α × σ = -22.76.
The particle diameter in the image is about 6 pixels, and according to the sample condition, the proportion of impurity regions in the neighborhood is not higher than 95% under the assumption that the area of a single impurity particle does not exceed 95% of the size of the neighborhood, 5 times of the particle diameter is taken as the size of the neighborhood, the Gaussian kernel radius k is 15, and the Gaussian kernel size is 31 multiplied by 31.
Thus, the pixel value is less than the neighbor weighted value by 22.76 or moreThe pixels can be classified as trash pixels. Adopting a thresholding segmentation method of relative threshold values, processing the image to be measured by formulas (2) - (4) and processing the image to be measured 2- (3) to classify the image to obtain a binary image, and obtaining a classification result I c The graph of (x, y) is shown in fig. 4, the white dots in the graph represent the pixel points classified as 1, and it can be seen that the positions and sizes of the white dots in the classification result are the same as those of the black dots in fig. 2- (3).
S3, calculating the quantity and the area of impurities
S31, analyzing a connected region: and processing the binary image 4 by adopting a connected region analysis algorithm to obtain a plurality of quintuple groups representing the connected regions in the binary image, wherein each quintuple group consists of a horizontal coordinate value, a vertical coordinate value, a horizontal maximum length, a vertical maximum length and the number of pixels of the connected regions. The results of the connected component analysis performed on FIG. 4 are [0, 600,359974] and [272,115,6,5,26], identifying the black and white connected components in FIG. 4, respectively.
S32, screening an impurity region: setting screening conditions, and acquiring some statistical distribution conditions of the powder in advance according to the delivery standard or production monitoring standard of a sample powder product, wherein the statistical distribution conditions comprise the content distribution of impurities, the particle size distribution of the powder, the geometric length, the sectional area size, the length-width ratio, the geometric position, the powder impurity distribution density, the combination of the powder and the impurity distribution density in an image and the like, and the statistical distribution conditions are used for judging the image characteristics of a normal sample and the impurities.
And determining partial screening conditions according to the image characteristics of the polymer powder, and screening the white connected region obtained in the connected region analysis to obtain a connected region which finally represents the impurities.
The screening conditions are that the side length of the impurity region is in the powder particle size range, the impurity region is close to the powder particle shape, the area of the impurity region is close to the powder sectional area, the distribution density of impurities in the powder is low, and the like. Taking 0.5 to 2 times of the typical particle size of 6 pixels in an image as a particle size range, taking the side length range of a connected region as 3 to 12 pixels, taking the area range as 10 to 200 pixels, taking the region length-width ratio not more than 3 (the particles are nearly spherical), and taking the impurity region number not more than 5 in the neighborhood of every 600 multiplied by 600 (obtained by observing the sample condition and practical experience rules), so that the connected region representing the impurities is a white connected region marked by a quintuple [272,115,6,5,26 ].
Example 2
The process of embodiment 2 is different except for obtaining the image to be measured, and all the method steps are completely the same as those of embodiment 1, that is, the same original image, visible light camera, definition of the spectrum sensitive wave band and the inert wave band and the same processing method (low-pass filtering processing, thresholding segmentation, connected region analysis, impurity region screening and the like) are adopted. Again for the same PVDC powder as in example 1.
S1, acquiring an image to be detected
S11, S12 and S13, selecting a sensitive wave band, selecting an image acquisition device and shooting a sample original image, wherein the process is as in embodiment 1, and the shot original image is as shown in figure 5- (1).
S14, acquiring an image to be detected: enhancing the original image: in the original image taken by the visible-light camera, a blue channel (see fig. 5- (2)) corresponding to a spectrally sensitive band in the image is selected as an image to be enhanced. This embodiment selects a weighted combination λ (Δ I) of the blue channel and the red channel respectively corresponding to the spectrally sensitive band and the inert band in the original image b (x,y)-αI r (x, y)) as an enhancement value, the weighting values are λ and- λ, respectively.
The image enhancement formula is as follows,
I aug (x,y)＝I b (x,y)+λ(αI b (x,y)-αI r (x,y)) (7)
αI b (x,y)＝I b (x,y)-μ b (8)
ΔI r (x,y)＝I r (x,y)-μ r (9)
in the formula I aug (x, y) denotes the enhanced original image pixel value, λ (Δ I) b (x,y)-ΔI r (x, y)) represents a weighted combination of the blue and red channels, Δ I b (x, y) and Δ I r (x, y) represent the difference of the pixel-to-channel mean of the blue and red channels, respectively, of the original image, I b (x, y) and I r (x, y) denote the pixels, μ, of the blue and red channels, respectively, of the original image b And mu r Respectively representing the pixel mean values of a blue channel and a red channel of an original image, wherein lambda is a constant of a weighted value, and an index of 2 is recommended.
Taking λ =2, processing the blue channel raw image of the spectral sensitivity band like 5- (2) using equations (7) - (9) yields an enhanced raw image like 5- (3). It can be seen that compared with the left image, the impurity area with black marks can be seen more clearly in fig. 5- (3), and the enhanced original image fig. 5- (3) is taken as the image to be measured.
S2, segmenting the image to be detected
S21, low-pass filtering: following the procedure of example 1, the particle diameter in the image was about 6 pixels, so k was taken to be 3, the gaussian kernel size was 7 x 7, and the standard deviation of the gaussian function was taken to be 5. The enhanced original image pixel value I aug Substituting (x, y) as I (x, y) into formula (1) to obtain pixel value I after image noise reduction D (x,y)。
S22, thresholding segmentation: the pixel standard deviation σ was calculated to be 9.32 according to fig. 5- (3) and equation (6). Obtaining the pixel histogram of fig. 5- (3) results in fig. 6, and assuming that the pixel values obey gaussian distribution according to fig. 6, then taking 3 times the standard deviation as the threshold criterion, i.e., α takes-3, then α × σ = -27.96. The particle diameter in the image was about 6 pixels, the gaussian kernel radius k was taken to be 15, and the gaussian kernel size was 31 x 31.
Therefore, pixels with a pixel value more than 27.96 less than the weighted value of the neighboring pixels can be classified as foreign pixels. Adopting a thresholding segmentation method of relative threshold values, processing a graph to be mapped 5- (3) by using formulas (2) - (4), and classifying a result I c The graph of (x, y) is shown in fig. 7, the white dots in the graph represent the pixel points classified as 1, and it can be seen that the positions and sizes of the white dots in the classification result are the same as those of the black dots in fig. 5- (3).
S3, calculating the quantity and the area of impurities
S31, analyzing a connected region: the connected region analysis of FIG. 7 was performed according to the procedure of example 1, and the results were [0, 600,359973] and [413,252,6,5,27], which respectively identify the black connected region and the white connected region of FIG. 7.
S32, screening impurity regions: according to the procedure of example 1, a white connected region finally representing the impurity and identified by the quintuple [413,252,6,5,27] was obtained according to the same screening conditions as example 1.
Claims (10)
1. The method for detecting the impurities in the crystalline polymer powder based on multispectral image processing is characterized by comprising the following steps of:
s1, acquiring an image to be detected: shooting a crystalline polymer powder sample to be detected by using image acquisition equipment to obtain a multispectral image as an original image, and selecting a partial image channel of the original image or the enhanced original image as the image to be detected;
s2, segmenting the image to be detected: carrying out thresholding segmentation on an image to be detected to obtain a binary image;
s3, calculating the quantity and area of impurities: processing the binary image by adopting a connected region analysis algorithm to obtain a plurality of adjacent image regions with the same pixel intensity as connected regions; and according to the geometrical characteristics of the impurity particles and the distribution condition of impurities in the powder, screening out a region meeting the standard from the plurality of communicated regions as an impurity region, and calculating the quantity and the area of the impurities.
2. The method for detecting powdery impurities in crystalline polymer based on multispectral image processing as claimed in claim 1, wherein the process of obtaining the original image comprises the steps of:
s11, selecting a spectrum sensitive wave band: selecting a wave band with large spectral characteristic difference as a spectral sensitive wave band and a wave band with small spectral characteristic difference as an inert wave band according to the spectral characteristics of the normal sample and the impurities;
s12, selecting an image acquisition device: selecting a camera or a camera combination with a sensor working waveband covering the spectrum sensitive waveband as image acquisition equipment;
s13, shooting an original image: and (3) spreading the to-be-detected crystalline polymer powder sample on a plane, and shooting the to-be-detected sample by using image acquisition equipment under the condition of sufficient illumination to obtain a multispectral image as an original image.
3. The method for detecting impurities in crystalline polymer powder based on multispectral image processing according to claim 1 or 2, wherein the step of obtaining the image to be detected specifically comprises the steps of:
s14, selecting image channels or combinations of channels corresponding to a plurality of sensitive wave bands from the original image as an image to be detected according to the spectrum sensitive wave bands and the inert wave bands; or selecting a weighted combination of image channels corresponding to the sensitive wave band and the inert wave band to enhance the original image and taking the enhanced original image as the image to be detected.
4. The method for detecting impurities in crystalline polymer powder based on multispectral image processing as claimed in claim 1 or 2, wherein the image capturing device comprises a black and white camera, a visible light camera, an infrared camera, an ultraviolet camera, a hyperspectral camera or any combination of the foregoing cameras.
5. The method for detecting impurities in crystalline polymer powder based on multispectral image processing as claimed in claim 1, wherein the step of segmenting the image to be detected comprises: and respectively carrying out thresholding segmentation on different areas of the image to be detected by adopting an absolute threshold or a relative threshold according to the uniformity of the brightness spatial distribution of the image to be detected.
6. The method for detecting impurities in crystalline polymer powder based on multispectral image processing as claimed in claim 5, wherein the thresholding segmentation of the absolute threshold converts the image to be detected into a binary image by comparing the absolute threshold with the relative intensity of the pixel intensity of each pixel.
7. The method for detecting impurities in crystalline polymer powder based on multispectral image processing as claimed in claim 5, wherein the difference between the pixel intensity of a single pixel and the weighted average of the pixel intensities of a certain neighborhood is calculated by thresholding segmentation with respect to a threshold value, and the image to be detected is converted into a binary image by comparing the relative threshold value with the relative magnitude of the difference.
8. The method according to claim 7, wherein the neighborhood diameter is at least 3 times the average particle size of the sample.
9. The method for detecting impurities in crystalline polymer powder based on multispectral image processing as claimed in claim 1, wherein when the image to be detected is segmented, a low-pass filtering method is used to smooth the image to be detected when the polymer powder sample is loose or the particle size of the powder is large.
10. The method for detecting impurities in a crystalline polymer powder based on multispectral image processing as recited in claim 1, wherein the geometric features of the impurity particles and the distribution of impurities in the powder comprise geometric length, cross-sectional area size, aspect ratio, geometric position, powder impurity distribution density, and combinations thereof of the crystalline polymer in the image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211272598.0A CN115601412A (en) | 2022-10-18 | 2022-10-18 | Method for detecting crystalline polymer powder impurities based on multispectral image processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211272598.0A CN115601412A (en) | 2022-10-18 | 2022-10-18 | Method for detecting crystalline polymer powder impurities based on multispectral image processing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115601412A true CN115601412A (en) | 2023-01-13 |
Family
ID=84847858
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211272598.0A Pending CN115601412A (en) | 2022-10-18 | 2022-10-18 | Method for detecting crystalline polymer powder impurities based on multispectral image processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115601412A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116359420A (en) * | 2023-04-11 | 2023-06-30 | 烟台国工智能科技有限公司 | Chromatographic data impurity qualitative analysis method based on clustering algorithm and application |
-
2022
- 2022-10-18 CN CN202211272598.0A patent/CN115601412A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116359420A (en) * | 2023-04-11 | 2023-06-30 | 烟台国工智能科技有限公司 | Chromatographic data impurity qualitative analysis method based on clustering algorithm and application |
| CN116359420B (en) * | 2023-04-11 | 2023-08-18 | 烟台国工智能科技有限公司 | Chromatographic data impurity qualitative analysis method based on clustering algorithm and application |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kleynen et al. | Development of a multi-spectral vision system for the detection of defects on apples | |
| Markovic et al. | Color measurement of food products using CIE L* a* b* and RGB color space. | |
| Kurniawati et al. | Investigation on image processing techniques for diagnosing paddy diseases | |
| Leemans et al. | Defects segmentation on ‘Golden Delicious’ apples by using colour machine vision | |
| JP5496509B2 (en) | System, method, and apparatus for image processing for color classification and skin color detection | |
| CN104732235B (en) | A kind of vehicle checking method for eliminating the reflective interference of road at night time | |
| CN108387516A (en) | Threading defects detection device and detection method based on CCD camera image processing technologies | |
| Wspanialy et al. | Early powdery mildew detection system for application in greenhouse automation | |
| CN116758083B (en) | Quick detection method for metal wash basin defects based on computer vision | |
| Vu et al. | Spatial and spectral features utilization on a hyperspectral imaging system for rice seed varietal purity inspection | |
| CN115601412A (en) | Method for detecting crystalline polymer powder impurities based on multispectral image processing | |
| CN111122590A (en) | Ceramic surface defect detection device and detection method | |
| Ye et al. | Removing shadows from high-resolution urban aerial images based on color constancy | |
| CN114216867A (en) | Hyperspectral image acquisition and identification device and method | |
| JP3505496B2 (en) | Image processing method and image processing apparatus | |
| Daskalov et al. | Comparative study on the application of visible and near infrared hyperspectral imaging for Fusarium disease assessment of corn seeds | |
| WO2005075961A1 (en) | Brightness evaluation method and brightness evaluation device | |
| Lee et al. | Sensors i: Color imaging and basics of image processing | |
| US20050017186A1 (en) | Method and means for detecting internal larval infestation in granular material | |
| CN107944426B (en) | Wheat leaf powdery mildew spot marking method based on combination of texture filtering and two-dimensional spectral feature space discrimination | |
| CN117495984B (en) | Intelligent calibration method for coaxial hybrid optical zoom pinhole lens | |
| Xu et al. | Citrus fruit recognition using color image analysis | |
| Gumiran et al. | Enhancing Threshold-based Phenotyping by Normalizing Image Luminosity | |
| Zhao et al. | Spectral properties inspired image enhancement for crystalline polymer powder impurity detection | |
| KAYA et al. | Detection of impurities in wheat and classification of durum wheat vitreousness in a dynamic system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |