Summary of the invention
The present invention proposes in order to overcome the shortcoming that exists in the prior art, and purpose provides a kind of particle sorting classifying method based on visual identity of using the intelligence of embedded vision image technique and mode identification technology.
Technical scheme of the present invention is: a kind of granular material sorting classifying method based on visual identity may further comprise the steps:
The first, obtain sample image (S1)
A, according to the difference of sorting material, selectively use the color of different radiation sources and reference background plate;
B, sampling, the actual sorting classifying requirement according to material obtains the middle honest material of this material and the sample of defective material respectively, the sorting standard of determining when the selection of sample can be according to this material sorting;
C, honest material and defective material sample are passed through equably the optical imaging system of embedded vision processing controls platform respectively, term harmonization for actual sorting the time, the material sample falls into the imaging region of optical imaging system along the device of making thinner with certain speed, is got access to the image of honest material and defective material sample at last by high speed gray scale or colored CCD;
The second, image feature information preliminary treatment (S2)
A, the honest material that will adopt and defective material image use corresponding image processing algorithm to handle, and can obtain gray value or the RGB color and the shape facility value of image after the processing;
B, the gradation of image value that obtains honest material and defective material or characteristic informations such as RGB color and shape facility value read in the two stack features values of information in the same plane of delineation;
Three, design category device (S3)
A, according to the method for statistical model identification, add up the contour shape of honest material and defective material and gray scale, RGB color threshold characteristic information value respectively;
B, according to the difference of sensitivity curve position, design different graders;
Four, set up characteristic information experts database (S4)
Grader is stored in the memory of embedded vision control processing controls platform, the different sorting classifying of material is with the different graders of correspondence, can in same table apparatus, set up grader for various materials, thereby formed a grader storehouse---characteristic information experts database, experts database is in case set up, its data will be stored in the memory device on the embedded vision processing controls platform, call during for sorting;
Five, material sorting to be selected and characteristic information experts database mate, and carry out sorting classifying (S5);
A, on touch-screen 2, choose the sorting classifying standard for the treatment of sorting material, promptly selected respective classified device in the characteristic information experts database;
B, embedded vision processing controls platform begin to control the feeding device of making thinner material evenly are sent to the optical imagery district, obtain the material image by optical imaging system;
C, by the IMAQ pretreatment system to image information preliminary treatment, characteristic information extraction;
The grader matching ratio is right in d, characteristic information and the characteristic information experts database;
E, according to matching result, if defective material, then the vision control system is sent to reject to instruct to and is rejected executing agency, rejects defective material.
Embedded vision processing controls platform is made up of IMAQ pretreatment system, vision control system and characteristic information experts database.
IMAQ pretreatment system and vision control system mainly are made up of devices such as field programmable gate array and digital signal processors.
The present invention is owing to adopted the embedded vision image recognition technology that granule materials is carried out classification and sorting, traditional accurate rate variance of selection mode, the low disadvantages such as easily causing secondary pollution that reaches of efficient have been overcome, make the classification and sorting quality be guaranteed, and realized fast intelligent, improved the classification and sorting efficient of granule materials.
The specific embodiment
Below, with reference to drawings and Examples the granular material sorting classifying method based on visual identity of the present invention is elaborated:
As shown in Figure 1, 2, a kind of granular material sorting classifying method based on visual identity, wherein separation and classifying device comprises machinery installation carrying platform 1, color touch screen 2 and embedded vision processing controls platform 3.
In machinery installation carrying platform 1, material storing box 4, feeding are installed and the system 5 that makes thinner, optical imaging system 6 and rejecting executing agency 7.
Wherein, feeding and the system of making thinner 5 are different according to the sorting classifying material, can be suitable for the different materials conveyer of making thinner.The material less to particles such as rice, the structure of employing oblique slot type; And the material of larger particles such as broad bean, pearl is adopted caterpillar structure, it is example that the present invention adopts caterpillar structure.
In track structure, the feeding and system's 5 involving vibrations feeders 14, skewed slot 15 and the material conveying device 20 of making thinner of making thinner.
Optical imaging system 6 comprises CCD camera lens 16, light source 17, polyhedron reference background plate 18 and lens (not shown), and CCD camera lens 16 comprises CCD and camera lens.
Wherein, CCD (Charge Coupled Device) is used for optical signal is converted to the signal of telecommunication.
Charge-coupled image sensor is that CCD camera lens 16 is made by a kind of semi-conducting material of ISO, can convert photon to electric charge, has resolution ratio height, accuracy of identification height, portable characteristics such as strong.
Polyhedron background board 18 is under different reference background, the gradation of image value that CCD camera lens 16 gets access to is different with RGB (red, green, orchid) characteristic information value, requirement for the material that satisfies the sorting different colours, adopting a polyhedron is the reference background plate, the present invention is a hexahedron, and every face scribbles different reference background looks.Polyhedron background board 18 is by control circuit control, and can requiring automatically according to the different material sorting classifying during work, the control background board turns to the different background faceted pebble.
Light source 17 adopts linear led light source.
Rejecting executing agency 7 forms by rejecting execution controlling organization 22 and nozzle 19, rejecting execution controlling organization 22 is made up of Drive and Control Circuit, compressed air system and pressure-regulating device etc., nozzle 19 can be designed to the different structure form as required, as the flat structure.
Color touch screen 2 is connected with embedded vision processing controls platform 3 by holding wire 13 as human-computer interaction interface.During sorting,, on color touch screen 2, show the honest material image that sub-elects according to different sorting classifying standards and the image of the defective material of correspondence with it according to the existing corresponding matching image characteristic of sorting material for the treatment of in the characteristic information expert knowledge library 11.Click different images, can determine to have selected different sorting standards.Simultaneously when sorting work, collect the image of current sorting material through optical imaging system 6 after, can shield on 2 in color touch and show in real time, be convenient to observe.Color touch of the present invention screen 2 can be selected color touch screen like MT8121 type or the performance classes for use.
Embedded vision processing controls platform 3 is made up of IMAQ pretreatment system 9, vision control system 10 and characteristic information experts database 11.
IMAQ pretreatment system 9 and vision control system 10 mainly are made up of field programmable gate array (FPGA) and digital signal processor devices such as (DSP), cooperate special-purpose data communication interface, data storage device composition and other digital logic device to form, realize the analyzing and processing of material image.
Wherein, on-site programmable gate array FPGA (Field Programmable Gate Array) is the highest a kind of of integrated level in the special IC (ASIC), and inside comprises configurable logic blocks, output input module and three parts of interconnector.In the present invention, field programmable gate array 6 can the serial EP3C120 chip of Cai Yong Hurricane wind III (Cyclone III) or performance classes like device.
Wherein, digital signal processor (DSP:Digital Signal Processor) is mainly used in digital processing field, is fit to very much high density, and the signal of repetitive operation and large data capacity is handled.Its adopts multibus technology and pipeline organization, program and data storage is separated use multibus, instruction fetch and fetch data and carry out simultaneously.DSP relies on hardware multiplier to finish multiplying in the monocycle, but also has special signal processing instruction, and computational speed is enhanced.Digital signal processor can select that device constitutes like chip TMS320DM64X or the performance classes for use, is used for the analyzing and processing of digital picture and the output of signal.
After embedded image process control system 3 is finished the image feature information comparison operation, send to reject to instruct to reject and carry out controlling organization 22, this moment, Drive and Control Circuit was sent corresponding magnetic valve nozzle 19 OPENs, by nozzle 19 ejection pressure-airs heterochromatic material (defective material) was blown out.Meanwhile, reject the pressure that compressed air system detects in actuating mechanism controls mechanism 22, when rejecting pressure was not enough or too big, the control automatic pressure regulator carried out the air pressure adjustment.
Characteristic information expert knowledge library 11 is according to collecting the image of relevant material and concrete sorting classifying standard, determines its normal material (honest material) and characteristic informations such as the image of the material (defective material) that will reject, shape profile.Characteristic information expert knowledge library 11 is stored in the mass storage on the embedded vision processing controls platform 3 in advance.
Can be divided into several colors such as mould fruit, black fruit, oil fruit and Chinese olive as the fruit of Chinese wolfberry according to its color through the airing oven dry, be divided into according to the exocuticle area of the fruit of Chinese wolfberry, the parameters such as size of maximum gauge: fruit of Chinese wolfberry king, 180 (represent the fruit of Chinese wolfberry number of per 50 grams in the fructus lycii dry fruits, down with), 280,350 and several classes such as little pearls and jewels.In the classification and sorting processing of reality, often determine multiple different standard according to client's different needs.And its criteria for classification of the fruit of Chinese wolfberry in the different places of production, different times is also different.Therefore, need to collect all classification and sorting standards as much as possible, carry out subjective appreciation according to its color, size etc., set up the characteristic information experts database 11 of matrimony vine classifying sorting by corresponding professional.
As shown in Figure 3, utilize said apparatus, the granular material sorting classifying method based on visual identity of the present invention may further comprise the steps:
The first, obtain sample image (S1)
A, according to the difference of sorting material, selectively use the color of different radiation sources 17 (color and spectral region) and reference background plate 18;
B, sampling according to the actual sorting classifying requirement of material, obtain the middle honest material (normal material) of this material and the sample of defective material (desire rejecting material) respectively, definite sorting standard when the selection of sample can be according to this material sorting;
C, honest material and defective material sample are passed through equably the optical imaging system 6 of embedded vision processing controls platform 3 respectively, term harmonization for actual sorting the time, the material sample falls into the imaging region of optical imaging system 6 along the device of making thinner with certain speed, is got access to the image of honest material and defective material sample at last by high speed gray scale or colored CCD;
In order to obtain the image feature information of many as far as possible honest materials and defective material, generally need repeatedly gather the image of honest material and defective material sample, finally obtain multiple image respectively by ccd image sensor.
The second, image feature information preliminary treatment (S2)
A, the honest material that will adopt and defective material image are handled, can obtain its gray value or RGB color and shape facility value, consider the working site circumstance complication of separation system, the image that collects can contain various unwanted interference noises and some unexpected picture distortion, and effect is rejected in influence.In order to restrain the noise in the image, simultaneously the useful information in the image is strengthened, therefore the image that must collect the CCD vision system uses corresponding image processing algorithm to handle.These algorithms comprise medium filtering, even smothing filtering algorithm etc., and this step is mainly finished by the IMAQ pretreatment system 9 of embedded vision processing controls platform 3;
B, obtain the gray value or the characteristic informations such as RGB color and shape facility value of the image of honest material and defective material, the two stack features values of information are read in the same plane of delineation, because the difference of image feature information between the honest material defective material, the image feature information of honest material and defective material will be in the diverse location in the plane of delineation.As shown in Figure 4, can see the top that the image of honest material will drop on, defective material drops on the below;
Three, design category device (S3)
A, at first need be according to the method for statistical model identification, add up the contour shape of honest material and defective material and gray scale, RGB color threshold characteristic information value respectively;
B, in two dimensional surface, honest material and the different zoness of different that respectively can fall within plane of defective material owing to characteristic value, but because different material requires the standard variation when sorting, same material may be divided into honest material in various criterion also may be divided into honest material, be the bound fraction in honest material and defective material zone among Fig. 4, can be classified as honest material and also can be classified as defective material;
As shown in Figure 5, define curve by setting a dynamic sensitivity, different with traditional setting threshold method, sensitivity curve is different according to the image feature information distributed areas of honest material and defective material, and a dynamic critical curve of delimiting, dropping on curve upper left is honest material (needing the product of reservation), dropping on curve bottom-right is defective material (the non-product of rejecting), critical curve is divided the difference of position, in the difference that to a certain degree embodies the sorting classifying standard;
C, according to the difference of sensitivity curve position, promptly represented different sorting classifying standards, so the sensitivity curve of four diverse locations among Fig. 5 promptly represents 4 kinds of different sorting classifying standards, also promptly according to 4 different graders of this design.
Comprise following content in the grader as the Xinjiang grape dry material:
Sorting classifying standard 1--grader 1: honest material is pure green raisin and light yellow raisins, and the defective material that need pick out is dark brown and black raisins.
Sorting classifying standard 2--grader 2: honest material is a green raisin, and the defective material that need pick out is yellow, dark brown and black raisins.
Sorting classifying standard 3--grader 3: honest material is that green raisin is pure, and the defective material that need pick out is the raisins (i.e. " blackhead ") of carpopodium position blackout and the raisins of other colors.
Four, set up characteristic information experts database (S4)
A, promptly be stored in embedded vision after errorless in the grader empirical tests of stating design in the 3rd step (S3) and control in the memory of processing controls platform 3.The different sorting classifying of material is with the different graders of correspondence.Can in same table apparatus, set up grader, thereby form a grader storehouse---characteristic information experts database for various materials.Experts database is in case set up, and its data will be stored in the memory device on the embedded vision processing controls platform 3, call during for sorting;
When b, actual sorting classifying, if when new sorting material and sorting classifying standard are arranged again, can carry out classifier design again according to the step of S1~S3 successively, the result writes characteristic information experts database 11 and gets final product the most at last;
C, when sorting, according to different material and sorting classifying requirement, can from memory, call the respective classified device and get final product, realize intelligent sorting.
Five, material sorting to be selected and characteristic information experts database mate, and carry out sorting classifying (S5);
A, on touch-screen 2, choose the sorting classifying standard for the treatment of sorting material, promptly selected respective classified device in the characteristic information experts database 11;
B, embedded vision processing controls platform 3 begin to control the feeding device 5 of making thinner material are sent to the optical imagery district equably, obtain the material image by optical imaging system 6;
C, by 9 pairs of image information preliminary treatment of IMAQ pretreatment system, characteristic information extraction;
The grader matching ratio is right in d, characteristic information and the characteristic information experts database;
E, according to matching result, if defective material, then vision control system 10 is sent to reject to instruct to and is rejected executing agency 7, rejects defective material.
The present invention is owing to adopted the embedded vision image recognition technology that granule materials is carried out classification and sorting, traditional accurate rate variance of selection mode, the low disadvantages such as easily causing secondary pollution that reaches of efficient have been overcome, make the classification and sorting quality be guaranteed, and realized fast intelligent, improved granule materials classification and sorting efficient.