CN102682293B - Image-based recognition method and system for gyroscopic glass bottle bump mold number - Google Patents

Abstract

The invention relates to a method and system for identifying a salient-point mould number of a revolution-solid glass bottle based on images. The method comprises the following steps of: 1) rotating the to-be-identified glass bottle on a turn table, and triggering a camera to collect a panoramic image containing a pitting mould number area by virtue of a photoelectric sensor and a rotary encoder;2) preprocessing the collected image to improve the quality of the image; 3) on the basis of smoothening the image rapidly by virtue of an integral image, segmenting the image by virtue of a dynamic threshold, and carrying out morphologic processing on the image to improve the segmentation effect, screening salient points of a mould according to the features (such as shapes, area and coordinates and the like) of the salient points of the mould, and extracting the coordinates of each salient point; 4) according to the distance relationships among the salient points, calculating corresponding symbolic expressions; and 5) inquiring a glass bottle mould code table, and determining the mould number according to the symbolic expressions obtained in the step 4). According to the invention, lightpaths are transformed by a mirror, thereby satisfying the needs on the space layout and adjustment of equipment; the collection of the high-definition salient-point images of the mould is ensured; and the identification system is simple in structure, low in cost, high in precision, high in speed, and easy to popularize and use.

Description

The salient point mould recognition methods of image-based revolution vial and system thereof

Technical field

The present invention relates to the mould recognition methods of a kind of revolution vial salient point and system, relate in particular to a kind of image-based revolution vial salient point mould recognition methods and system.

Background technology

The vial of different size is made by different moulds and production procedure, near the bottle end root salient point mould number is arranged, and corresponding different moulds of different salient point permutation and combination number by mold code table inquiry decoding, can be identified mould number.Such as artificial interpretation, then efficient is low, even erroneous judgement.

For the square glass bottle, can be taken pictures by the front, through image recognition mould number.But for the revolution vial, as: beer bottle.Perspective during because of imaging changes, and accurately gathers salient point very difficult.In addition, vial is being produced conveyer line, because of the close bottle of a salient point end, the implantation of devices such as camera, light source, Existential Space restricted problem.

Consult the national patent storehouse, existing " automatic recognition system of point mold code of glass bottle/jar " (application number: 200920226221.5) utility model patent, but this patent is with laser sensor and single-chip microcomputer, during laser process point-like mould, light intensity changes, and the sensor make-and-break signal is identified mould number, when practical operation, a code-point easily occurs and make mistakes, cause the whole mould situation that number identification makes mistakes.

Summary of the invention

The objective of the invention is in order to overcome the deficiencies in the prior art, a kind of image-based revolution mould for glass bottle recognition methods and system thereof are provided, it has advantages of the problem that solves simultaneously device layout space constraint and high-precision mold Real time identification.

To achieve these goals, the present invention adopts following technic relization scheme:

A kind of image-based revolution vial salient point mould recognition system, it comprises the LED area source, rotation platform, rotary encoder, photoelectric sensor, ccd video camera, level crossing, product travelling belt, wherein, rotation platform is connected with rotary encoder, the rotation platform rear is provided with the LED area source, product travelling belt and rotation platform join, the oblique the place ahead of rotation platform is provided with ccd video camera, be provided with level crossing between ccd video camera and the product travelling belt, be provided with photoelectric sensor in the opposite side of product travelling belt, oblique the place ahead of rotation platform.

The recognition methods of a kind of image-based revolution vial salient point mould the steps include:

1) be guided out the vial image by level crossing, by photoelectric sensor and rotary encoder, trigger line frequency and frame frequency image acquisition, control gathers slightly many survey faces of one week of vial, that is: vial rotation 365 degree are to 370 degree; Shooting comprises convex dot shape mould number district's close-up view;

2) carry out quick filter and noise suppression preprocessing to gathering the mould image, improve picture quality;

3) utilize dynamic thresholding method, to Image Segmentation Using, learn with the burn into expansion form again and process, improve segmentation effect, utilize shape corresponding to salient point, area, coordinate Feature Selection to go out salient point, extract the salient point coordinate;

4) according to the die point coordinate, calculate adjacent bump pitch from, draw mould character expression corresponding to die point; Because each vial orientation is random on the travelling belt, the situation that continuous point-like mould number is truncated can appear, and will block die point and reconnect;

5) the inquiry mould code table that prestores identifies mould corresponding to each salient point number, provides the mould number.

In the described step 3): utilize integrogram mean filter method to former figure filtering, try to achieve the average image, then former figure and the average image are subtracted each other and obtain error image, error image is carried out dynamic threshold segmentation, obtain the binary segmentation image.

In the described step 4), with blocking the method that die point reconnects be: for row's mould salient point, distance between its 2 left and the rightest end points all is 13 times of minimum adjacent bump pitch, and the spacing maximum between the adjacent salient point also only has the situation of 3 times of adjacent spacings of minimum to occur; Distance must be die point to have occurred to block in gathering image greater than the situation of 3 times of adjacent spacings of minimum between the salient point if find to occur in gathering image, can only be from breaking when blocking with originally continuous salient point; To occur this moment die point block before and character string corresponding to salient point that from image, read afterwards be recorded in respectively in first and second character strings, be combined into a new character strings by second character string and first character string, from new character strings, take out the character expression corresponding to salient point of 13 correspondences, thereby just can inquire about the mould code table with it, try to achieve mould number.

The present invention adopts the polishing of LED area source, uses the optics shift theory, and by photoelectric sensor and rotary encoder location, ccd video camera gathers vial panorama picture, can keep the cross ratio invariability of mould salient point.By image processing algorithm, first positioning convex point zone, the pre-service such as rapid image filtering improves quality again, and then split image extracts the salient point position relationship, inquires about at last the mould code table, identifies mould number.

Beneficial effect of the present invention:

1) adopts ccd video camera and image processing algorithm, realize convolution vial salient point mould Real time identification;

2) vial rotates a circle manyly, by the rotary encoder pulse signal CCD camera frame frequency trigger pip is set, and guarantees to collect whole salient point information, has cross ratio invariability between image bumps and the actual salient point simultaneously;

3) adopt integrogram quick filter method to realize quick filter, then carry out dynamic threshold segmentation;

4) the inquiry mould code table that prestores apart from corresponding symbol code, identifies mould number by the mould salient point;

5) by minute surface conversion light path, satisfy device space layout and regulatory demand;

6) cooperating of LED area source and CCD camera, guarantee to gather high-resolution mould salient point information;

7) recognition system is simple in structure, and cost is low, and precision is high, and speed is fast, is easy to promote the use of.

Description of drawings

Fig. 1 is image-based revolution vial salient point mould recognition system schematic diagram;

Fig. 2 is vial salient point mould photo;

Fig. 3 is that CCD gathers the salient point areal map;

Fig. 4 a is the continuous situation of mould salient point;

Fig. 4 b is the situation that the mould salient point is truncated.

Among the figure, the 1:LED area source; 2: rotation platform; 3: rotary encoder; 4: photoelectric sensor; 5: vial; The 6:CCD video camera; 7: level crossing; 8: the product travelling belt.

Embodiment

The invention will be further described below in conjunction with accompanying drawing and embodiment.

As shown in Figure 1, a kind of image-based revolution vial salient point mould recognition system, it comprises LED area source 1, rotation platform 2, rotary encoder 3, photoelectric sensor 4, ccd video camera 5, level crossing 7, product travelling belt 8, wherein, rotation platform 2 is connected with rotary encoder 3, rotation platform 2 rears are provided with LED area source 1, product travelling belt 8 joins with rotation platform 2, rotation platform 2 oblique the place aheads are provided with ccd video camera 6, be provided with level crossing 7 between ccd video camera 6 and the product travelling belt 8, at the opposite side of product travelling belt 8, oblique the place ahead of rotation platform 1 is provided with photoelectric sensor 4.

LED area source 1 irradiation vial 5 to be measured, vial 5 is in rotation platform 2 rotations, by rotary encoder 3 position rotating information, by optical flat mirror 7 first to vial 5 imagings, vial 5 reimagings in 6 pairs of minute surfaces of ccd video camera, pass through image processing algorithm, point-like mould zone, location, because the difference of salient point transmittance, gather image bumps and non-salient point corresponding grey scale and have difference, extract thus each salient point coordinate, utilize the salient point position relationship, inquiry mould code table can identify salient point mould number.

Actual step of the present invention:

1) vial 5 is sent on the rotation platform 2, photoelectric sensor 4 and rotary encoder 3 orientation triggering collected by cameras comprise the panorama sketch of mould salient point;

2) according to actual observation and statistical study, positioning convex point mould regional extent;

3) to gathering the pre-service such as image strengthens, improve picture quality;

4) with integrogram fast filtering method smoothed image, then carry out image segmentation with dynamic thresholding method;

5) process with morphology, improve segmentation effect, utilize the features such as shape, area, coordinate of mould salient point, screening extracts each die point coordinate;

6) by each die point coordinate, calculate distance between the consecutive point, draw mould character expression corresponding to salient point.Because each vial orientation is random on the travelling belt, the situation that continuous point-like mould number is truncated can appear.Analyzing on the mould salient point feature base, the exploitation recognizer will be blocked die point and be reconnected.

7) inquiry mould code table identifies mould number, provides the mould number.

The key of image-based revolution vial salient point mould recognition methods is the range information that accurately extracts between each salient point of vial, and inquiry mould code table (as shown in table 1) is tried to achieve corresponding mould number.

1. based on the mean filter principle of integrogram

The integrogram I of image P refers to, is the point of (x, y) for coordinate in the image, its I (x, y) equal in (x, y) among P point rectangular area, upper left side all the pixel values of point with, that is:

$I(x,y)=\underset{i\≤x,j\≤y}{\mathrm{\Σ}}P(i,j)---\left(1\right)$

Filtering with default filter kernel coefficient matrix, is carried out convolution algorithm with image usually.Generally getting the filter window length of side is the filtering core of odd number (width W=2*h+1, height H=2*h+1, h, w are filter window half height and half-breadth, get positive integer).Filtering operation can be expressed as follows:

$P^{\′}(x_0,y_0)=\underset{(x,y)\∈D}{\mathrm{\Σ}}C(x+w-x_0,y+h-y_0)P(x,y)---\left(2\right)$

P ' (x in the formula ₀, y ₀) be (x in the image after the filtering ₀, y ₀) pixel value of position.

The filtering core element is constant during mean filter, gets C (i, j)=1/ (W*H), and (2) can be reduced to:

$P^{\′}(x_0,y_0)=\underset{(x,y)\∈D}{\mathrm{\Σ}}\frac{1}{W\·H}P(x,y)$

$=\frac{1}{W\·H}\underset{(x,y)\∈D}{\mathrm{\Σ}}P(x,y)---\left(3\right)$

Integration item in the following formula

In the expression original image in the D of rectangular area the gray-scale value of each pixel and, it can be calculated by integrogram, so,

$P^{\′}(x_0,y_0)=\frac{1}{W\·H}[I(x_0+w,y_0+h)-I(x_0+w,y_0)$

$-I(x_0,y_0+h)+I(x_0-w,y_0-h)]---\left(4\right)$

By (4) as seen, only need carry out integrogram I (x, y), the average computing is converted into the simple operation of the integrated value in the calculation of filtered window.Owing to need not every pixel double counting weighted sum, operand no longer changes with the filter window size, has improved operation efficiency.

2. dynamic threshold segmentation method

Dynamic threshold segmentation carries out threshold operation with the local mean value of object pixel as Rule of judgment, can be described as:

f(x,y)＝(f(x,y)-mean≥c) foregroud:background (5)

F (x in the formula, y) be (x in the image, y) pixel value of position, mean refers to the pixel local mean value of image slices vegetarian refreshments, the threshold value of c for setting, foregroud and background are respectively display foreground and the background gray levels of setting, f ' (x, y) be the pixel value of (x, y) position in the segmentation result image.During practical operation, with integrogram mean filter method the average image is tried to achieve in former figure filtering, then former figure and the average image are subtracted each other and obtain error image, error image is carried out (5) formula carry out Threshold segmentation, obtain dynamic binary segmentation image.

3. block the join algorithm of die point

After studying various mould for glass bottle salient points, find: for row's mould salient point, distance between its 2 left and the rightest end points all is 13 times of minimum adjacent bump pitch, and the spacing maximum between the adjacent salient point also only has the situation of 3 times of adjacent spacings of minimum to occur.Distance must be die point to have occurred to block in gathering image greater than the situation of 3 times of adjacent spacings of minimum between the salient point if find to occur in gathering image, can only be from breaking when blocking with originally continuous salient point.To occur this moment die point block before and character string corresponding to salient point that from image, read afterwards be recorded in respectively in first and second character strings, be combined into a new character strings by second character string and first character string, from new character strings, take out the salient point character expression of 13 correspondences.Thereby just can inquire about the mould code table with it, try to achieve mould number.

Although above-mentionedly by reference to the accompanying drawings the specific embodiment of the present invention is described; but be not limiting the scope of the invention; one of ordinary skill in the art should be understood that; on the basis of technical scheme of the present invention, those skilled in the art do not need to pay various modifications that creative work can make or distortion still in protection scope of the present invention.

Claims (3)

1. A recognition method based on an image-based revolving body glass bottle bump mold number recognition system, said system comprising LED surface light source, rotary platform, rotary encoder, photoelectric sensor, CCD camera, plane mirror, product conveyor belt, wherein, rotating The platform is connected to the rotary encoder, LED surface light source is installed behind the rotary platform, the product conveyor belt is connected to the rotary platform, a CCD camera is installed obliquely in front of the rotary platform, and a flat mirror is installed between the CCD camera and the product conveyor belt. There are photoelectric sensors on the side and oblique front of the rotating platform; It is characterized in that the identification method comprises the following steps: 1) The image of the glass bottle is guided by the plane mirror, and the photoelectric sensor and the rotary encoder trigger the line frequency and frame frequency image acquisition to control the acquisition of the side of the glass bottle for a little more than a week, that is: the glass bottle rotates from 365 degrees to 370 degrees; the shooting includes The overall view of the number area of the bump-shaped mold; 2) Perform fast filtering and denoising preprocessing on the collected mold number images to improve image quality; 3) Use the dynamic threshold method to segment the image, and then use erosion and expansion morphology to improve the segmentation effect, use the shape, area, and coordinate characteristics corresponding to the bumps to filter out the bumps, and extract the bump coordinates; 4) According to the coordinates of the mold points, calculate the distance between adjacent bumps, and obtain the symbolic expression corresponding to the bumps of the mold; because the orientation of each glass bottle on the conveyor belt is random, there will be a situation where continuous point mold numbers are truncated, and the mold will be truncated point to reconnect; 5) Query the pre-stored mold code table, identify the mold number corresponding to each bump, and give the mold number data. the 2. The identification method according to claim 1, characterized in that, in said step 3), the original image is filtered by the mean value filter method of the integral image to obtain the average image, and then the original image is subtracted from the average image to obtain The difference image, performing dynamic threshold segmentation on the difference image to obtain a binary segmented image. the 3. The identification method according to claim 1, characterized in that, in the above step 4), the method of reconnecting the truncated mold points is: for a row of mold bumps, the leftmost and rightmost 2 The distance between two endpoints is 13 times the minimum distance between adjacent bumps, and the maximum distance between adjacent bumps is only 3 times the minimum distance between adjacent bumps; If the distance is greater than 3 times the minimum adjacent spacing, it must be that the mold point is truncated in the collected image. When truncating, the original continuous bumps can only be disconnected from it; The character strings corresponding to the convex points read in the image are respectively recorded in the first and second character strings, and a new character string is composed of the second character string and the first character string, and 13 characters are taken out from the new character string The corresponding embossed character expression, so it can be used to query the mold code table to obtain the mold number. the

Images