CN108171705A - The foreign bodies detection algorithm of liquid in a kind of Clear glass bottles and jars - Google Patents
The foreign bodies detection algorithm of liquid in a kind of Clear glass bottles and jars Download PDFInfo
- Publication number
- CN108171705A CN108171705A CN201810056180.3A CN201810056180A CN108171705A CN 108171705 A CN108171705 A CN 108171705A CN 201810056180 A CN201810056180 A CN 201810056180A CN 108171705 A CN108171705 A CN 108171705A
- Authority
- CN
- China
- Prior art keywords
- pixel
- image
- background model
- background
- sample
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 51
- 239000011521 glass Substances 0.000 title claims abstract description 38
- 239000007788 liquid Substances 0.000 title claims abstract description 29
- 238000012545 processing Methods 0.000 claims abstract description 19
- 238000000034 method Methods 0.000 claims description 41
- 230000008569 process Effects 0.000 claims description 22
- 238000000605 extraction Methods 0.000 claims description 4
- 230000002708 enhancing effect Effects 0.000 abstract description 37
- 230000008859 change Effects 0.000 abstract description 3
- 230000000694 effects Effects 0.000 description 27
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 11
- 239000008267 milk Substances 0.000 description 9
- 210000004080 milk Anatomy 0.000 description 9
- 235000013336 milk Nutrition 0.000 description 9
- 238000001914 filtration Methods 0.000 description 8
- 238000013461 design Methods 0.000 description 5
- 230000007423 decrease Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 230000002146 bilateral effect Effects 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 210000000746 body region Anatomy 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 239000006071 cream Substances 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000002715 modification method Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000003389 potentiating effect Effects 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 238000009738 saturating Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20172—Image enhancement details
- G06T2207/20182—Noise reduction or smoothing in the temporal domain; Spatio-temporal filtering
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of foreign bodies detection algorithms of liquid in Clear glass bottles and jars, Steerable filter enhancing processing first is carried out to video image, Clear glass bottles and jars profile is gone out using " Hough " change detection again, algorithm detection zone is contracted to liquid detecting area, each pixel for liquid detecting region establishes a background model, calculate the similarity of current pixel point and background model, it is if similar, then it is classified as background, otherwise it is prospect, calculates the connected region area of prospect, if connected region area is less than given threshold, background is then classified as, is otherwise prospect.The present invention can be detected the foreign matter in nontransparent liquid.
Description
Technical field
The present invention relates to a kind of image processing techniques, the foreign bodies detection algorithm of liquid in particularly a kind of Clear glass bottles and jars.
Background technology
In Clear glass bottles and jars in the intelligent detection equipment of nontransparent fluid product, most crucial detection technique is based on machine
The product defects identification of vision and detection technique, wherein preferable image enhancement processing is the base of foreign bodies detection in subsequent liquid
Plinth.
At present, the preferable image enhancement technique of enhancing effect has following four：Histogram equalization, help secretly Dow process,
Retinex_MSR and Steerable filter method.
(1) histogram equalization be by Cumulative Distribution Function convert based on Histogram Modification Methods, it can generate one
The uniform image of width grey level distribution probability, the image after histogram equalization have larger information content.With regard to contrast dynamic
For range image less than normal, which, which is based on very shirtsleeve operation, can effectively enrich gray level, therefore as image
Automatically the effective means enhanced.
(2) dark channel prior algorithm has preferable effect, but processing time is long, needs for the processing of single image defogging
Very big storage resources and computing resource.Therefore, the shortcoming of the algorithm is that real-time is poor, and through dark channel prior algorithm
The image handled can be darker than original image, also needs that image is exposed processing to treated, again increases the complexity of algorithm
Degree.
(3) Retinex_MSR algorithms are superimposed in simple terms by multiple SSR algorithms, using several different big
Linear weighted function normalization has just obtained MSR algorithms again after small scale parameter filtering, and the essence of the two is actually identical.
MSR algorithms can adaptively adjust the size of scale parameter according to different images, have good adaptive ability, realize preferable
Color effect while keep the preferable detailed information of image.MSR algorithms have color constancy more better than SSR algorithm,
The features such as dynamic range compression.And when handling coloured image, often there is the phenomenon that insufficient color saturation, halation in MSR algorithms.
(4) Steerable filter is a kind of image filtering technology, schemes G (be oriented to and scheme) by a guiding, (defeated to target image I
Enter image) it is filtered so that last output image is generally similar to target image I, but texture part and guiding
It is similar to scheme G.When it is same image that guiding figure G is with input figure I, it is oriented to the effect of figure filtering and the effect of bilateral filtering
Similar, but unlike that bilateral filtering, the filtering of guiding figure can be readily devised an optimization unrelated with filter radius
Algorithm.When input figure I is an initial mask image, the effect of guiding figure filtering, which is similar to, scratches nomography.Steerable filter
Typical case is to protect edge image smoothly and scratch figure, and preferable enhancing effect is also achieved in the enhancing of image defogging.
Invention content
The purpose of the present invention is to provide a kind of foreign bodies detection algorithm of liquid in Clear glass bottles and jars, to saturating in vial
Foreign matter in prescribed liquid or nontransparent liquid is detected.
Realize the object of the invention technical solution be：Include the following steps：
Step 1, image is acquired to Clear glass bottles and jars, obtains the RGB image containing moving target, and to image successively
Carry out gray scale and Steerable filter processing；
Step 2, Hough transform is carried out to the image that step 1 obtains；
Step 3, the pixel gray value in the image that extraction step 2 obtains in preceding N frames image in Clear glass bottles and jars region,
Establish the initial background model M (x) for pixel in Clear glass bottles and jars region；
Step 4, each pixel of each frame image after N+1 frames start and the similarity of background model are calculated, if
It is similar, then background is classified as, goes to step 5；Otherwise it is prospect, goes to step 6；
Step 5, for the similar pixel of each frame image after N+1 frames,The similar pixel of probability updating corresponds to
The sample of background model,The similar pixel F × F neighborhoods of probability updating in certain pixel correspond to the sample of background model,
Go to step 6；
Step 6, it by calculating the connected region area of each frame image after N+1 frames, excludes small in liquid detecting region
In the tiny noise jamming of threshold value；
Step 7, step 4 is repeated to step 6, up to all N+1 to nth frame image detection are completed.
Compared with prior art, the present invention haing the following advantages (1) for foreign bodies detection in more muddy liquid, introducing is led
Image enhancement is carried out to filtering, enhances gradation of image contrast, highlights the useful information in figure, improve the follow-up place of image
Reason ability is more advantageous to detection and extraction to foreign matter；(2) for the random noise point outside liquid detecting region, by " Hough "
Transformation is introduced into VIBE algorithms, and by carrying out straight-line detection to vial profile, detection zone is contracted to liquid in vial
Body region not only reduces the operand of algorithm, improves the speed of service of algorithm, also effectively eliminates outside liquid detecting region
Background area existing for noise jamming；(3) for the tiny noise spot in liquid detecting region, by setting connected region area
Threshold value will be set to background dot less than the connected region of the area threshold.
The invention will be further described with reference to the accompanying drawings of the specification.
Description of the drawings
Fig. 1 is the method for the present invention flow chart.
Fig. 2 is histogram equalization, Steerable filter, Retinex_MSR algorithms and helps Dow process secretly to foreign matters different in clear water
Enhancing effect comparison diagram, wherein the gray-scale map of (a1) primitive frame image；(b1) the enhancing effect figure (c1) of histogram equalization
The enhancing effect figure of Steerable filter；(d1) the enhancing effect figure of Retinex_MSR algorithms；(e1) enhancing effect of Dow process is helped secretly
Figure.
Fig. 3 for histogram equalization, Steerable filter, Retinex_MSR algorithms and help secretly Dow process in slightly troubled liquor not
With the enhancing effect comparison diagram of foreign matter, wherein the gray-scale map of (a1) primitive frame image；(b1) enhancing effect of histogram equalization
Scheme the enhancing effect figure of (c1) Steerable filter；(d1) the enhancing effect figure of Retinex_MSR algorithms；(e1) enhancing of Dow process is helped secretly
Design sketch.
Fig. 4 for histogram equalization, Steerable filter, Retinex_MSR algorithms and help secretly Dow process to compared in troubled liquor not
With foreign matter enhancing effect comparison diagram wherein, the gray-scale map of (a1) primitive frame image；(b1) enhancing effect of histogram equalization
Scheme the enhancing effect figure of (c1) Steerable filter；(d1) the enhancing effect figure of Retinex_MSR algorithms；(e1) enhancing of Dow process is helped secretly
Design sketch.
Fig. 5 is the foreign bodies detection effect contrast figure of connected region before and after the processing, wherein (a) be not using connected region processing
Foreign bodies detection design sketch；(b) the foreign bodies detection design sketch handled using connected region.
Specific embodiment
With reference to Fig. 1, the foreign bodies detection algorithm of liquid, includes the following steps in a kind of Clear glass bottles and jars：
Step 1, image is acquired to Clear glass bottles and jars, obtains the RGB image containing moving target, and to image successively
Carry out gray scale and Steerable filter processing；
Step 2, Hough transform is carried out to the image that step 1 obtains；
Step 3, the pixel gray value in the image that extraction step 2 obtains in preceding N frames image in Clear glass bottles and jars region,
Establish the initial background model M (x) for pixel in Clear glass bottles and jars region；
Step 4, each pixel of each frame image after N+1 frames start and the similarity of background model are calculated, if
It is similar, then background is classified as, goes to step 5；Otherwise it is prospect, goes to step 6；
Step 5, for the similar pixel of each frame image after N+1 frames,The similar pixel of probability updating corresponds to
The sample of background model,The similar pixel F × F neighborhoods of probability updating in certain pixel correspond to the sample of background model,
Go to step 6；
Step 6, it by calculating the connected region area of each frame image after N+1 frames, excludes small in liquid detecting region
In the tiny noise jamming of threshold value；
Step 7, step 4 is repeated to step 6, up to all N+1 to nth frame image detection are completed.
In step 1, gray level image carries out image enhancement step using Steerable filter and is exemplified below：Using images themselves to lead
Xiang Tu, windows radius are sized to 16, and regularization coefficient is set as 0.4^2, the Steerable filter figure of output and the line being oriented between figure
Property coefficient is respectively 5 and 1.
" Hough " transformation is carried out in step 2, the detection zone of algorithm is reduced, improves the speed of service of algorithm and exclude liquid
Noise jamming step outside body detection zone is exemplified below：The peak number to be found of setting is 5 when " Hough " is converted, and sets peak
Be worth threshold size for " ceil (0.3*max (H (:))) ", " H " is obtained " Hough " matrix, only more than the point of the threshold value
Being considered as may peak point.The line segment that combined distance is taken the straight line detected to be less than 5 abandons all length and is less than 7
The method of line segment filters out the profile line segment of Clear glass bottles and jars, records the position of contour line section.To step 2 in enhanced image
Pixel judge it whether in Clear glass bottles and jars contour area, if certain pixel outside Clear glass bottles and jars contour area,
For background, it is otherwise prospect, goes to step 3, the detection zone of algorithm is narrowed down in Clear glass bottles and jars contour area, reduces and calculate
The operand of method.
Background model M (x)={ P in step 31,P2,...,PN, wherein, P1,P2,...,PNSample for background model
This.This example N takes 20, and 20 samples are shared in background model M (x).
Similar deterministic process in step 4 is：
Step 4.1, since N+1 frames, for a certain pixel x of present frame, gray value is P (x), in European color
Defined in space one centered on P (x), R be radius circle SRThe gray value and the corresponding back of the body of (P (x)) to current frame pixel point x
Sample in scape model compares, wherein, R be Model Matching threshold value, SR(P (x)) represents all samples for being less than R with P (x) distances
This set；
Step 4.2, if some sample value of background model is in circle SRIn (P (x)), then it represents that pixel x and the sample
Matching, i.e., with # { SR(P(x))∩{P1,P2,...PNThe similarity of P (x) and background model M (x) described；This example R takes 20；
Step 4.3, given threshold #min, if # { SR(P(x))∩{P1,P2,...PN＜ #min, then pixel x with
Background model M (x) is mismatched, and judges the point for foreground point；Otherwise, it is judged as background dot；This example #min takes 3.
In step 5, the update of background model is the key that moving object detection algorithm, mainly background model is fitted
Answer the continuous variation of background, such as the variation of illumination, the change of background object etc..The similar pixel of probability updating corresponds to the back of the body
The detailed process of the sample of scape model is：With the gray value of the similar pixel of present frame replace withProbability is from background model
The gray value of corresponding pixel in the sample selected in M (x).The similar pixel F × F neighborhoods of probability updating in certain picture
The detailed process that vegetarian refreshments corresponds to the sample of background model is：Probability replaced with the gray value of the similar pixel of present frame
In its F × F neighborhoods in the corresponding background model of some pixel respective point pixel gray value.
In step 6, connected region area threshold is 50, and connected region area in testing result is less than to the foreground zone of threshold value
Domain background is shown, is otherwise judged as foreground target.
Embodiment
Visible light video is acquired first with Visible-light CCD, it will be in video input to computer；It is carried to detect the present invention
The foreign matter detecting method effect of liquid in a kind of Clear glass bottles and jars gone out, now passes through MATLABR2014a developing algorithm simulation models
The Steerable filter algorithm for image enhancement introduced in the present invention and histogram equalization, Retinex_MSR algorithms and dark is first
The enhancing effect of checking method is compared.It is 352*288 to choose size respectively, frame rate be 25 frames/second clear water in it is black
Color scrap gum sport video 1；Ecru eraser chips sport video 2 in clear water；White bits sport video 3 in clear water；In clear water thoroughly
Bright chips of glass sport video 4；Black rubber considers sport video 5 to be worth doing slightly in troubled liquor；Ecru eraser chips are transported slightly in troubled liquor
Dynamic video 6；White bits sport video 7 slightly in troubled liquor；Transparent glass considers sport video 8 to be worth doing slightly in troubled liquor；Compared with turbid solution
Black rubber considers sport video 9 to be worth doing in body；Compared with ecru eraser chips sport video 10 in troubled liquor；Compared with white bits in troubled liquor
Sport video 11；Sport video 12 is considered to be worth doing compared with transparent glass in troubled liquor；
For every frame image gray processing first of original video source, histogram equalization is respectively adopted to the image of gray processing
Change, Steerable filter, Retinex_MSR algorithms and dark channel prior algorithm carry out image enhancement processing.
Foreign matter in Fig. 2, Fig. 3, Fig. 4 in first row image is considered to be worth doing for black rubber, and the foreign matter in secondary series image is cream colour
Color eraser chips, the foreign matter in third row image are white paper scrap, and the foreign matter in the 4th row image is transparent glass bits.
As can be seen from Figure 2, the gray-scale map for scheming the former video frame images shown in (a1) is dark, and the identification of foreign matter is very low, figure
(b1) overall brightness of the image after histogram equalization improves, and gray scale is stretched, and the tonal gradation included in image is more
Abundant, the information that image includes increases, and the foreign matter of detection and the grey-scale contrast of neighborhood territory pixel increase, but effect and unknown
It is aobvious；Scheme (c1) enhancing effect after Steerable filter, the grey-scale contrast in image significantly increases, foreign matter and the neighborhood picture of detection
The gray difference of element is apparent, and enhancing effect is preferable；The effect that figure (d1) is enhanced using Retinex_MSR algorithms, image entirety
Brightness improves, but image entirety is more fuzzy, and enhancing effect is poor.Figure (e1) is to carry out image enhancement using Steerable filter
Effect, enhanced image overall gray value declines, dark compared with original image, and the effect that highlights of foreign matter is not obvious, and increases
It is strong ineffective.
From figure 3, it can be seen that figure (a1) is the gray-scale map of the slightly primitive frame image of troubled liquor, since liquid turbidity increases
Add, human eye is difficult identification foreign matter.(b1) is schemed for slightly troubled liquor frame image effect after histogram equalization, is significantly deposited in image
In " exposure " phenomenon of large area, ecru eraser chips, white paper scrap and transparent glass foreign matter, inspection have been blocked in white exposure area
It is poor to survey effect.In addition, " exposure " of large area may cause noise spot in final detection figure to increase；(c1) is schemed through Steerable filter
Enhancing effect afterwards, the grey-scale contrast in image significantly increase, and the foreign matter of detection and the gray difference of neighborhood territory pixel are apparent, increase
Potent fruit is preferable；The effect that figure (d1) is enhanced using Retinex_MSR algorithms, the brightness of image entirety improves, but image is whole
It is more fuzzy, enhancing effect is poor.Figure (e1) is using the effect for helping Dow process progress image enhancement secretly, foreign matter and field pixel
Grey-scale contrast have increased slightly, but enhanced image overall intensity is dark compared with original image, foreign matter profile is not obvious, enhancing effect
Fruit is bad.
Scheme as can be seen from Figure 4, (a1) is the gray-scale map of the primitive frame image compared with troubled liquor, due to liquid turbidity again
Increase, human eye can not almost recognize foreign matter.(b1) is schemed for figure (a1) effect after histogram equalization, it can be seen that is gone out in image
Now apparent large area " exposure " and " bulk " phenomenon, it is virtually impossible to detect and tell foreign matter, detection result is very poor.In addition,
" exposure " and " bulk " of large area may cause noise spot in final detection figure to increase；Scheme the enhancing of (c1) after Steerable filter
Effect, image overall intensity is relatively low, compares primitive frame gradation of image figure, can distinguish foreign matter position, but detection result is bad；Figure
(d1) effect enhanced using Retinex_MSR algorithms, the brightness of image entirety are improved, but image entirety is very fuzzy, can not
Foreign matter is distinguished, enhancing effect is very poor.It is the enhancing effect figure for helping Dow process secretly to scheme (e1), and overall intensity is dark compared with original image, almost without
Method differentiates foreign matter, and enhancing effect is poor.
To compare the speed of service of different enhancing algorithms, increase clocking capability in former algorithm, detect that different enhancings are calculated
The operation of method takes.
1 algorithm of histogram equalization of table handles a frame image and takes (s)
| Black rubber is considered to be worth doing | Ecru eraser chips | White paper scrap | Transparent glass | |
| Clear water | 0.1067 | 0.1079 | 0.1101 | 0.1084 |
| Slightly muddy milk power solution | 0.1069 | 0.1062 | 0.1059 | 0.1067 |
| More muddy milk power solution | 0.1058 | 0.1061 | 0.1061 | 0.1060 |
2 Steerable filter algorithm process of table, one frame image takes (s)
| Black rubber is considered to be worth doing | Ecru eraser chips | White paper scrap | Transparent glass | |
| Clear water | 0.1272 | 0.1314 | 0.1301 | 0.1278 |
| Slightly muddy milk power solution | 0.1260 | 0.1263 | 0.1267 | 0.1272 |
| More muddy milk power solution | 0.1278 | 0.1292 | 0.1269 | 0.1323 |
3 Retinex_MSR algorithm process of table, one frame image takes (s)
| Black rubber is considered to be worth doing | Ecru eraser chips | White paper scrap | Transparent glass | |
| Clear water | 0.1645 | 0.1689 | 0.1651 | 0.1643 |
| Slightly muddy milk power solution | 0.1740 | 0.1665 | 0.1678 | 0.1791 |
| More muddy milk power solution | 0.1671 | 0.1656 | 0.1691 | 0.1656 |
4 dark algorithm process of table, one frame image takes (s)
| Black rubber is considered to be worth doing | Ecru eraser chips | White paper scrap | Transparent glass | |
| Clear water | 0.3181 | 0.3161 | 0.3184 | 0.3284 |
| Slightly muddy milk power solution | 0.3253 | 0.3184 | 0.3157 | 0.3185 |
| More muddy milk power solution | 0.3145 | 0.3176 | 0.3165 | 0.3185 |
Data in contrast table 1,2,3 and 4 can be seen that in four kinds enhance algorithm, helps Dow process secretly and handles a frame image
Longest is taken, and treatment effect is poor；Steerable filter is long compared with the time of one frame image of histogram equalization processing, compares Retinex_
The time of one frame image of MSR algorithm process is short, and histogram equalization and Retinex_MSR algorithms are for the primitive frame of troubled liquor
The enhancing effect of image is good not as good as the enhancing effect of Steerable filter.Consider, this paper selective guides are filtered to primitive frame image
Carry out enhancing processing, but for the enhancing compared with troubled liquor, treated that image is integrally dark for enhancing.
In order to verify the effect for calculating connected region area to removing tiny noise, it is respectively adopted and is added without calculating connected region
The algorithm that domain area and addition calculate connected region area carries out simulation process to above-mentioned video.
Fig. 5 is the foreign bodies detection effect contrast figure of connected region before and after the processing：(a) not using the foreign matter of connected region processing
Detection result figure；(b) the foreign bodies detection design sketch handled using connected region；
From fig. 5, it can be seen that figure (a) is as a result, the interference due to noise not using the foreign bodies detection of connected region processing
There is the smaller noise spot of connected region area, detection result are caused to decline in foreign matter edge in testing result, liquid detecting region,
In the frame image even occurred in no foreign matter, a small number of noise spots are mistaken for foreign matter target, accuracy of detection is caused to decline.Figure
(b) it is using the foreign matter effect detected after the judgement of connected region area, it can be seen that the noise spot at foreign matter edge is gone completely
It removes, the detection profile of foreign matter is more clear；The smaller noise spot of other connected region areas in liquid detecting region is removed
Effect is also preferable；For the frame image that no foreign matter occurs, using, without foreground point, reducing and miss in connected region treated image
Sentence rate, improve accuracy of detection.
In order to by the foreign matter detecting method of liquid in traditional VIBE algorithms and a kind of Clear glass bottles and jars proposed by the present invention
Detection result carries out detailed comparisons, we have carried out simulation process to above-mentioned video respectively with two kinds of algorithms.
The foreign bodies detection effect of 5 tradition VIBE algorithms of table
As can be seen from Table 5, VIBE algorithms can detect four kinds of foreign matters in clear water, but for transparent glass in clear water
Detection result is poor, the area very little of the chips of glass detected；For in slightly troubled liquor, the detection results of four kinds of foreign matters compared with
Difference considers shape smaller black rubber bits and transparent glass to be worth doing, and VIBE algorithms are nearly no detectable；For more muddy milk powder solution
Four kinds of foreign matters in body, the overwhelming majority can't detect.It can be seen that the muddy degree due to liquid is different, detection result is caused
Decline to some extent, therefore, before foreign bodies detection is carried out, image need to be pre-processed, reduce troubled liquor to foreign matter
Fuzzy Influence, enhance picture contrast, and then improve detection result.
Table 6 is the foreign bodies detection effect of this paper VIBE algorithms
As can be seen from Table 6, compared with the effect of former VIBE algorithms detection foreign matter, the integrity degree for detecting foreign matter significantly carries
Height, former algorithm consider the black rubber in slightly troubled liquor to be worth doing, and transparent glass foreign matter can not detected, and this paper algorithms can be preferable
Detect such foreign matter；Further for compared with the foreign matter in troubled liquor, former VIBE algorithms can not almost detect all kinds of foreign matters,
This paper algorithms can detect that black rubber is considered to be worth doing, ecru eraser chips and white paper scrap, but for transparent glass foreign matter, the inspection of this paper
It is poor to survey effect.
Claims (8)
1. the foreign bodies detection algorithm of liquid in a kind of Clear glass bottles and jars, which is characterized in that include the following steps：
Step 1, image is acquired to Clear glass bottles and jars, obtains the RGB image containing moving target, and image is carried out successively
Gray scale and Steerable filter processing；
Step 2, Hough transform is carried out to the image that step 1 obtains；
Step 3, the pixel gray value in the image that extraction step 2 obtains in preceding N frames image in Clear glass bottles and jars region is established
For the initial background model M (x) of pixel in Clear glass bottles and jars region；
Step 4, each pixel of each frame image after N+1 frames start and the similarity of background model are calculated, if similar,
Background is then classified as, goes to step 5；Otherwise it is prospect, goes to step 6；
Step 5, for the similar pixel of each frame image after N+1 frames,The similar pixel of probability updating corresponds to background mould
The sample of type,The similar pixel F × F neighborhoods of probability updating in certain pixel correspond to the sample of background model, go to step
6；
Step 6, it by calculating the connected region area of each frame image after N+1 frames, excludes in liquid detecting region less than threshold
The tiny noise jamming of value；
Step 7, step 4 is repeated to step 6, up to all N+1 to nth frame image detection are completed.
2. according to the method described in claim 1, it is characterized in that, background model M (x)={ P in step 31,P2,...,
PN,
Wherein, P1,P2,...,PNSample for background model.
3. according to the method described in claim 1, it is characterized in that, the similar deterministic process in step 4 is：
Step 4.1, since N+1 frames, for a certain pixel x of present frame, gray value is P (x), in European color space
Defined in one centered on P (x), R be radius circle SRThe gray value and corresponding background mould of (P (x)) to current frame pixel point x
Sample in type compares, wherein, R be Model Matching threshold value, SR(P (x)) represents all samples for being less than R with P (x) distances
Set；
Step 4.2, if some sample value of background model is in circle SRIn (P (x)), then it represents that pixel x and the sample matches,
Use # { SR(P(x))∩{P1,P2,...PNThe similarity of P (x) and background model M (x) described；
Step 4.3, given threshold #min, if # { SR(P(x))∩{P1,P2,...PN＜ #min, then pixel x and background mould
Type M (x) is mismatched, and judges the point for foreground point；Otherwise, it is judged as background dot.
4. according to the method described in claim 1, it is characterized in that,The similar pixel of probability updating corresponds to background model
The detailed process of sample is：
With the gray value of the similar pixel of present frame replace withIt is corresponding in the sample that probability is selected from background model M (x)
Pixel gray value.
5. according to the method described in claim 1, it is characterized in that,The similar pixel F × F neighborhoods of probability updating in certain
The detailed process that pixel corresponds to the sample of background model is：
Probability replace the corresponding back of the body of some pixel in its F × F neighborhoods with the gray value of the similar pixel of present frame
The gray value of respective point pixel in scape model.
6. according to the method described in claim 1, it is characterized in that, the method that Steerable filter is handled in step 1 includes：
16 are sized to using images themselves as figure, windows radius is oriented to, regularization coefficient is set as 0.4^2, the Steerable filter of output
Linear coefficient between figure and guiding figure is respectively 5 and 1.
7. according to the method described in claim 1, it is characterized in that, the peak value to be found is set in step 2 during Hough transform
Mesh is 5, setting peak threshold size for ceil (0.3*max (H (:))), it is possible peak point that the point more than the threshold value, which is,.
8. according to the method described in claim 1, it is characterized in that, in step 6, connected region area threshold is 50.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810056180.3A CN108171705A (en) | 2018-01-20 | 2018-01-20 | The foreign bodies detection algorithm of liquid in a kind of Clear glass bottles and jars |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810056180.3A CN108171705A (en) | 2018-01-20 | 2018-01-20 | The foreign bodies detection algorithm of liquid in a kind of Clear glass bottles and jars |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108171705A true CN108171705A (en) | 2018-06-15 |
Family
ID=62514954
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810056180.3A Pending CN108171705A (en) | 2018-01-20 | 2018-01-20 | The foreign bodies detection algorithm of liquid in a kind of Clear glass bottles and jars |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108171705A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109615610A (en) * | 2018-11-13 | 2019-04-12 | 浙江师范大学 | A kind of medical band-aid flaw detection method based on YOLO v2-tiny |
| CN112529877A (en) * | 2020-12-15 | 2021-03-19 | 中国联合网络通信集团有限公司 | Method and device for detecting impurities in bottled liquid |
| CN116823835A (en) * | 2023-08-30 | 2023-09-29 | 山东省永星食品饮料有限公司 | Bottled water impurity detection method based on machine vision |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105784722A (en) * | 2016-04-29 | 2016-07-20 | 中南大学 | Method and system of detecting visible foreign matters in medicine liquid bottles on flow line |
| US20160352995A1 (en) * | 2015-05-26 | 2016-12-01 | SK Hynix Inc. | Apparatus for generating image and method thereof |
| CN107220950A (en) * | 2017-05-31 | 2017-09-29 | 常州工学院 | A kind of Underwater Target Detection image enchancing method of adaptive dark channel prior |
-
2018
- 2018-01-20 CN CN201810056180.3A patent/CN108171705A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160352995A1 (en) * | 2015-05-26 | 2016-12-01 | SK Hynix Inc. | Apparatus for generating image and method thereof |
| CN105784722A (en) * | 2016-04-29 | 2016-07-20 | 中南大学 | Method and system of detecting visible foreign matters in medicine liquid bottles on flow line |
| CN107220950A (en) * | 2017-05-31 | 2017-09-29 | 常州工学院 | A kind of Underwater Target Detection image enchancing method of adaptive dark channel prior |
Non-Patent Citations (2)
| Title |
|---|
| OLIVIER BARNICH 等: "ViBe: A Universal Background Subtraction Algorithm for Video Sequences", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
| 陶雄: "液体制剂中异物检测的机器智能识别", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109615610A (en) * | 2018-11-13 | 2019-04-12 | 浙江师范大学 | A kind of medical band-aid flaw detection method based on YOLO v2-tiny |
| CN109615610B (en) * | 2018-11-13 | 2023-06-06 | 浙江师范大学 | Medical band-aid flaw detection method based on YOLO v2-tiny |
| CN112529877A (en) * | 2020-12-15 | 2021-03-19 | 中国联合网络通信集团有限公司 | Method and device for detecting impurities in bottled liquid |
| CN116823835A (en) * | 2023-08-30 | 2023-09-29 | 山东省永星食品饮料有限公司 | Bottled water impurity detection method based on machine vision |
| CN116823835B (en) * | 2023-08-30 | 2023-11-10 | 山东省永星食品饮料有限公司 | Bottled water impurity detection method based on machine vision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115082683B (en) | Injection molding defect detection method based on image processing | |
| CN115375676A (en) | Stainless steel product quality detection method based on image recognition | |
| US9251614B1 (en) | Background removal for document images | |
| CN111915704A (en) | Apple hierarchical identification method based on deep learning | |
| CN111161222B (en) | Printing roller defect detection method based on visual saliency | |
| CN108171705A (en) | The foreign bodies detection algorithm of liquid in a kind of Clear glass bottles and jars | |
| CN111145105B (en) | Image rapid defogging method and device, terminal and storage medium | |
| CN116883408B (en) | Integrating instrument shell defect detection method based on artificial intelligence | |
| CN111754538B (en) | Threshold segmentation method for USB surface defect detection | |
| CN116309191B (en) | Intelligent gas inspection display method based on image enhancement | |
| Shaikh et al. | A novel approach for automatic number plate recognition | |
| CN113221881B (en) | Multi-level smart phone screen defect detection method | |
| CN114708325A (en) | Method for quickly positioning rubber production problem based on rubber blooming defect | |
| CN113256624A (en) | Continuous casting round billet defect detection method and device, electronic equipment and readable storage medium | |
| CN114926407A (en) | Steel surface defect detection system based on deep learning | |
| CN110866915A (en) | Circular inkstone quality detection method based on metric learning | |
| CN114549441A (en) | Sucker defect detection method based on image processing | |
| CN111815563B (en) | Retina optic disc segmentation method combining U-Net and region growing PCNN | |
| CN117094975A (en) | Method and device for detecting surface defects of steel and electronic equipment | |
| CN114018946B (en) | OpenCV-based high-reflectivity bottle cap defect detection method | |
| CN114092682A (en) | Small hardware fitting defect detection algorithm based on machine learning | |
| CN108805854B (en) | Method for rapidly counting tablets and detecting completeness of tablets in complex environment | |
| CN113537037A (en) | Pavement disease identification method, system, electronic device and storage medium | |
| CN111210405A (en) | Volatile organic compound leakage detection method based on infrared imaging | |
| CN117197064A (en) | Automatic non-contact eye red degree analysis method |
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 |