CN113989218B

CN113989218B - Double electrostatic dust collection method and device based on parameter prediction model

Info

Publication number: CN113989218B
Application number: CN202111244675.7A
Authority: CN
Inventors: 熊松林; 施根荣; 蔡勇峰
Original assignee: Dongguan Gaoqi Printing Co ltd
Current assignee: Dongguan Gaoqi Printing Co ltd
Priority date: 2021-10-26
Filing date: 2021-10-26
Publication date: 2022-05-20
Anticipated expiration: 2041-10-26
Also published as: CN113989218A

Abstract

The application discloses a double electrostatic dust removal method based on a parameter prediction model, which is used for carrying out dust splashing treatment on a first article; shooting to obtain a first image; adjusting parameters of the first electrostatic precipitation roller, and performing primary precipitation treatment; acquiring a specified dust removal position; carrying out secondary dust removal treatment; shooting to obtain a second image; generating a third image; modifying the first parameter and the second parameter for multiple times to obtain multiple third images; calculating a similarity value and constructing a first parameter group; carrying out n times of pattern modification processing to obtain n parameter sets; training by using the parameter group as sample data to obtain a parameter prediction model; obtaining an image to be analyzed; inputting an image to be analyzed into a parameter prediction model to obtain a first prediction parameter and a second prediction parameter; and parameters of the first electrostatic precipitation roller and parameters of the second electrostatic precipitation roller array are adjusted at the same time, and the second article is subjected to dedusting treatment, so that the aim of improving the dedusting effect is fulfilled.

Description

Double electrostatic dust collection method and device based on parameter prediction model

Technical Field

The application relates to the field of computers, in particular to a double electrostatic dust collection method and device based on a parameter prediction model.

Background

The dust removal technology comprises an electrostatic adsorption technology and a non-electrostatic adsorption technology. The non-electrostatic adsorption technology, for example, uses a dust-free cloth to smear alcohol, and performs dust removal treatment on the surface of an object, wherein alcohol generally needs to be smeared on the dust-free cloth, and since alcohol is volatile, alcohol needs to be frequently added, and the dust-free cloth intercepts dirt, and after the dirt is accumulated, secondary ink type dirt can be generated due to alcohol smearing, and meanwhile, the surface can be scratched by dirt particles which are harder than the dirt particles, and the like. Therefore, the effect of this non-electrostatic adsorption technique is low.

Although the electrostatic adsorption technology can overcome the defects of the dust removal technology, a single electrostatic adsorption scheme is adopted in a general electrostatic adsorption scheme, namely, only one electrostatic dust removal roller is adopted, and electrostatic voltage is applied to the electrostatic dust removal roller for adsorbing dust on the surface of an object. Its adsorption effect on dust is still not ideal enough.

Disclosure of Invention

The application provides a double electrostatic dust collection method based on a parameter prediction model, which comprises the following steps:

s1, according to the first pattern drawn in advance, carrying out dust sprinkling treatment on the first article placed on the transmission belt in advance, so that dust on the first article presents the first pattern; shooting the surface of a first article to obtain a first image;

s2, starting the conveyor belt, and adjusting the preset parameter of the first electrostatic precipitation roller to be the preset first parameter to enable the first article to be subjected to primary precipitation treatment through the first electrostatic precipitation roller; the first electrostatic precipitation roller is a fixed precipitation roller;

s3, calling a corresponding relation table of preset dust patterns and dust removal positions to obtain a specified dust removal position corresponding to the first pattern; adjusting the preset parameter of a movable second electrostatic precipitation roller array to be a preset second parameter, placing the second electrostatic precipitation roller array at a specified precipitation position, and performing secondary precipitation treatment when the first article passes through the second electrostatic precipitation roller array; shooting the surface of the first article to obtain a second image;

s4, correspondingly subtracting the color value of each pixel point in the second image from the color value of each pixel point in the first image respectively, and generating a third image;

s5, modifying the first parameters and the second parameters for multiple times, and repeating the steps S1-S4 to obtain a plurality of third images; according to a preset pattern similarity calculation method, calculating a similarity value between the third image and the first pattern, selecting a designated third image corresponding to the maximum similarity value, acquiring a numerical value of a first parameter and a numerical value of a second parameter corresponding to the designated third image, and constructing a first parameter group consisting of the first image, the numerical value of the first parameter corresponding to the designated third image and the numerical value of the second parameter;

s6, performing n times of pattern modification processing to modify the first pattern into a second pattern, a third pattern, … and an nth pattern, and repeating the steps S1-S5 to obtain a second parameter group, a third parameter group, … and an n parameter group; n is an integer greater than 3;

s7, calling a preset convolutional neural network model, and training the convolutional neural network model in a semi-supervised learning mode by adopting the first parameter group, the second parameter group, the third parameter group, the … parameter and the n parameter group as sample data to obtain a parameter prediction model;

s8, shooting the surface of a second object to be dedusted to obtain an image to be analyzed; inputting an image to be analyzed into the parameter prediction model for processing to obtain a first prediction parameter and a second prediction parameter output by the parameter prediction model; wherein the second item is of the same type as the first item;

and S9, adjusting the parameter of the first electrostatic precipitation roller as a first prediction parameter, adjusting the parameter of the second electrostatic precipitation roller array as a second prediction parameter, and performing precipitation treatment on the second article.

Further, the first parameter comprises a minimum distance between the first electrostatic precipitation roller and the first article, an electrostatic voltage of the first electrostatic precipitation roller and a radius of the first electrostatic precipitation roller;

the second parameters include a minimum spacing of each sub-roller of the second array of electrostatic precipitation rollers from the first article, an electrostatic voltage of each sub-roller, and a radius of each sub-roller.

Further, the minimum distance between the first electrostatic precipitation roller and the first article is more than 0 cm; each of the sub-rollers of the second array of electrostatic precipitation rollers has a minimum separation from the first article of greater than 0 cm.

Furthermore, the number of the first electrostatic precipitation rollers is two, and the first electrostatic precipitation rollers are symmetrically arranged relative to the plane where the conveying belt is located;

the number of second electrostatic precipitator roller array is two, and one of them second electrostatic precipitator roller array sets up in the top of transmission band, and another second electrostatic precipitator roller array sets up in the below of transmission band.

Further, the first electrostatic precipitation roller is also in contact with a preset precipitation silica gel rod, and the precipitation silica gel rod is used for transferring the dust adsorbed by the static electricity.

Further, according to a first pre-drawn pattern, carrying out dust sprinkling treatment on a first article which is placed on a conveying belt in advance, so that dust on the first article presents the first pattern; the step S1 of performing a photographing process on the surface of the first article to obtain a first image includes:

s101, dyeing the pre-collected dust into a specified color;

s102, according to a first pre-drawn pattern, carrying out dust splashing treatment on a first article which is placed on a conveying belt in advance, so that dust on the first article presents the first pattern with a specified color;

s103, shooting the surface of the first article to obtain a first image.

Further, modifying the first parameter and the second parameter for multiple times, and repeating the steps S1-S4 to obtain multiple third images; step S5, executed by the computer, of calculating a similarity value between the third image and the first pattern according to a preset pattern similarity calculation method, selecting a designated third image corresponding to a maximum similarity value, obtaining a value of a first parameter and a value of a second parameter corresponding to the designated third image, and constructing a parameter group i consisting of the first image, the value of the first parameter corresponding to the designated third image, and the value of the second parameter, includes:

s501, modifying the first parameter and the second parameter for multiple times, and repeating the steps S1-S4 to obtain a plurality of third images;

s502, converting the color values of the plurality of third images into a plurality of first vectors respectively according to a preset vector conversion method, and simultaneously converting the color values of the first patterns into second vectors;

s503, according to the formula:

calculating a similarity value R between one third image and the first pattern so as to obtain a plurality of similarity values respectively corresponding to the plurality of third images; the method comprises the following steps that Pi is a numerical value of an ith vector in a first vector, Qi is a numerical value of an ith vector in a second vector, the first vector and the second vector are m-dimensional vectors, m is an integer larger than 1, and a, b and c are preset parameters larger than 0;

s504, selecting a designated similarity value with the maximum value from the plurality of similarity values, and acquiring a designated third image corresponding to the designated similarity value;

s505, obtaining the numerical value of the first parameter and the numerical value of the second parameter corresponding to the appointed third image, and constructing a first parameter group consisting of the numerical value of the first parameter and the numerical value of the second parameter corresponding to the first image and the appointed third image.

Further, the step S7 of calling a preset convolutional neural network model, and training the convolutional neural network model in a semi-supervised learning manner by using the first parameter group, the second parameter group, the third parameter group, …, and the n parameter group as sample data to obtain a parameter prediction model includes:

s701, automatically labeling sample data to obtain n sample data, wherein the images in the first parameter group are labeled with a numerical value of a first parameter and a numerical value of a second parameter in the first parameter group, the images in the second parameter group are labeled with a numerical value of a first parameter and a numerical value of a second parameter in the second parameter group, …, and the images in the n parameter group are labeled with a numerical value of a first parameter and a numerical value of a second parameter in the n parameter group;

s702, dividing the n sample data according to a preset proportion to obtain a plurality of training data and a plurality of verification data;

s703, calling a preset convolutional neural network model, and inputting the training data into the convolutional neural network model for training to obtain a temporary parameter prediction model;

s704, verifying the temporary parameter prediction model by adopting a plurality of verification data, and judging whether the result of the verification processing is passed;

s705, if the result of the verification process is that the verification is passed, the temporary parameter prediction model is regarded as the final parameter prediction model.

The application provides a dual electrostatic precipitator device based on parameter prediction model includes:

a first image capturing unit configured to perform step S1 of performing dust-dusting processing on a first article placed on the conveyor belt in advance in accordance with a first pattern drawn in advance so that dust on the first article assumes the first pattern; shooting the surface of a first article to obtain a first image;

a primary dust removal processing unit, configured to execute step S2, start the conveyor belt, adjust a preset parameter of the first electrostatic dust removal roller to a preset first parameter, and perform primary dust removal processing on the first article by using the first electrostatic dust removal roller; the first electrostatic precipitation roller is a fixed precipitation roller;

a secondary dust removal processing unit, configured to perform step S3, call a preset correspondence table between dust patterns and dust removal positions, so as to obtain a specified dust removal position corresponding to the first pattern; adjusting the preset parameter of a movable second electrostatic precipitation roller array to be a preset second parameter, placing the second electrostatic precipitation roller array at a specified precipitation position, and performing secondary precipitation treatment when the first article passes through the second electrostatic precipitation roller array; shooting the surface of the first article to obtain a second image;

the color value subtraction unit of the pixel point is used for executing the step S4, and correspondingly subtracting the color value of each pixel point in the second image from the color value of each pixel point in the first image respectively, so as to generate a third image;

a pattern similarity calculation unit for performing step S5, modifying the first parameter and the second parameter a plurality of times, and repeating steps S1-S4 to obtain a plurality of third images; according to a preset pattern similarity calculation method, calculating a similarity value between the third image and the first pattern, selecting a designated third image corresponding to the maximum similarity value, acquiring a numerical value of a first parameter and a numerical value of a second parameter corresponding to the designated third image, and constructing a first parameter group consisting of the first image, the numerical value of the first parameter corresponding to the designated third image and the numerical value of the second parameter;

a pattern modification processing unit, configured to perform step S6, perform n times of pattern modification processing to modify the first pattern into a second pattern, a third pattern, …, an nth pattern, and repeat steps S1-S5 to obtain a parameter set number two, a parameter set number three, …, a parameter set number n; n is an integer greater than 3;

a parameter prediction model obtaining unit, configured to execute step S7, invoke a preset convolutional neural network model, and train the convolutional neural network model in a semi-supervised learning manner by using the first parameter group, the second parameter group, the third parameter group, …, and the n parameter group as sample data, so as to obtain a parameter prediction model;

the image to be analyzed shooting unit is used for executing the step S8 and shooting the surface of the second object to be dedusted to obtain an image to be analyzed; inputting an image to be analyzed into the parameter prediction model for processing to obtain a first prediction parameter and a second prediction parameter output by the parameter prediction model; wherein the second item is of the same type as the first item;

and a prediction parameter adjusting unit, configured to perform step S9, adjust the parameter of the first electrostatic precipitation roller to be a first prediction parameter, adjust the parameter of the second electrostatic precipitation roller array to be a second prediction parameter, and perform precipitation processing on the second article.

The present application provides a computer device comprising a memory storing a computer program and a processor implementing the steps of any of the above methods when the processor executes the computer program.

According to the parameter prediction model-based double electrostatic dust collection method, the device and the computer equipment, dust splashing treatment is carried out on a first article; shooting to obtain a first image; adjusting parameters of the first electrostatic precipitation roller, and performing primary precipitation treatment; acquiring a specified dust removal position; adjusting parameters of a movable second electrostatic dust removal roller array, and performing secondary dust removal treatment; shooting to obtain a second image; generating a third image; modifying the first parameter and the second parameter for multiple times to obtain multiple third images; calculating the similarity value, selecting a designated third image corresponding to the maximum similarity value, and constructing a first parameter group; carrying out n times of pattern modification processing to obtain an n number parameter set; training the convolutional neural network model by adopting a parameter group as sample data in a semi-supervised learning mode to obtain a parameter prediction model; obtaining an image to be analyzed; inputting the image to be analyzed into the parameter prediction model to obtain a first prediction parameter and a second prediction parameter; and adjusting the parameter of the first electrostatic precipitation roller as a first prediction parameter, adjusting the parameter of the second electrostatic precipitation roller array as a second prediction parameter, and performing precipitation treatment on the second article, so that the aim of improving the precipitation effect is fulfilled.

This application adopts dual electrostatic precipitator, and first heavy whole removes dust, and the second is the pertinence to remove dust. Because the second is the electrostatic precipitation roller array, each small electrostatic precipitation roller can adopt different parameters, thereby carrying out targeted precipitation on different positions.

Moreover, when the dust removal is carried out on a special object, the effect that the common electrostatic dust removal scheme cannot achieve can be achieved, namely, the dust removal effect is improved, and the damage to the object in the dust removal process can be avoided.

Specifically, for a special object when the material on the surface of the object has a weak adsorption capacity with the main body of the object (for example, cloth that is not suitable for washing with water, and gold powder is brushed or sprinkled at the stage of fast production), if a large electrostatic adsorption force is used during electrostatic adsorption, the particles are easily attracted by static electricity and leave the object, thereby causing damage to the surface of the object; if a smaller electrostatic adsorption force is used in electrostatic adsorption, a satisfactory dust removal effect cannot be achieved. And the application can find out appropriate parameters to provide appropriate electrostatic adsorption force, so that in the special scene, the damage to objects in the dust removal process can be avoided.

Drawings

FIG. 1 is a schematic flow chart of a method for dual electrostatic precipitator based on a parametric prediction model according to an embodiment of the present application;

FIG. 2 is a schematic block diagram of a dual electrostatic precipitator based on a parametric prediction model according to an embodiment of the present application;

fig. 3 is a block diagram illustrating a structure of a computer device according to an embodiment of the present application.

The implementation, functional features and advantages of the objectives of the present application will be further explained with reference to the accompanying drawings.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

Referring to fig. 1, an embodiment of the present application provides a dual electrostatic dust removal method based on a parameter prediction model, including the following steps:

s5, modifying the first parameter and the second parameter for multiple times, and repeating the steps S1-S4 to obtain a plurality of third images; according to a preset pattern similarity calculation method, calculating a similarity value between the third image and the first pattern, selecting a designated third image corresponding to the maximum similarity value, acquiring a numerical value of a first parameter and a numerical value of a second parameter corresponding to the designated third image, and constructing a first parameter group consisting of the first image, the numerical value of the first parameter corresponding to the designated third image and the numerical value of the second parameter;

s8, shooting the surface of a second object to be dedusted to obtain an image to be analyzed; inputting an image to be analyzed into the parameter prediction model for processing to obtain a first prediction parameter and a second prediction parameter output by the parameter prediction model; wherein the second article is of the same type as the first article;

The parameter prediction model-based dual electrostatic dust removal method can be suitable for surface dust removal of any feasible material, preferably surface dust removal of a material with particles or coatings with low adsorption force attached to the surface, and can achieve excellent dust removal effect on common materials such as common cloth, plates, paper and other any feasible materials.

The design of double electrostatic precipitation is adopted, wherein the first heavy electrostatic precipitation is carried out by adopting a first electrostatic precipitation roller and is used for carrying out comprehensive basic precipitation, and the electrostatic adsorption capacity generated by the first heavy electrostatic precipitation roller is smaller (relative to a certain sub-roller or a whole sub-roller in a second electrostatic precipitation roller array) by adopting a first parameter; and the second electrostatic dust removal roller array is used for targeted local dust removal, and the second parameter is adopted, so that the second electrostatic dust removal roller array can form larger electrostatic adsorption capacity in a local range, and the adsorption capacity of the second electrostatic dust removal roller array on the dust is improved.

It should be noted that the present application does not employ a large electrostatic adsorption capacity directly on the first electrostatic precipitation roller, because if it is configured, a large overall electrostatic attraction force is generated, but the electrostatic adsorption capacity is not required for all positions of the material surface to adsorb dust, so that there may occur a case where a portion of the material surface having a weak bonding force is adsorbed, thereby causing unnecessary damage on the material surface. Therefore, the scheme of double electrostatic dust removal is adopted in the application, so that the dust removal is realized in stages and pertinently.

The size of the first electrostatic precipitation roller can be any feasible size, but the preferred size is larger, such as: the axial length of the first electrostatic precipitation roller is greater than or equal to the width of the first article. The size of each electrostatic precipitation roller in the second array of electrostatic precipitation rollers may be any feasible size, but is preferably small, for example: and the axial length of each electrostatic precipitation roller in the second electrostatic precipitation roller array is less than or equal to half of the axial length of the first electrostatic precipitation roller. The second electrostatic precipitation roller array can be formed into an array in any feasible mode, for example, all the sub-rollers are positioned on the same straight line, but have a certain distance between each other, so that the second electrostatic precipitation roller array is formed; or the axial directions of all the sub-rollers are parallel to each other but not all in the same straight line.

An electrostatic precipitation roll, as the name implies, is a roll that is electrostatically charged for removing dust. The electrostatic precipitation roller in the present application may be in any feasible configuration, for example, it is arranged as a columnar structure composed of an electrode and a dielectric layer wrapped outside the electrode.

In the present application, the first parameter includes a minimum distance between the first electrostatic precipitation roller and the first article, an electrostatic voltage of the first electrostatic precipitation roller, and/or a radius of the first electrostatic precipitation roller;

the second parameters include a minimum spacing of each sub-roller of the second array of electrostatic precipitation rollers from the first article, an electrostatic voltage of each sub-roller, and/or a radius of each sub-roller.

The present application is directed to modulating the first parameter and the second parameter in order to find a suitable electrostatic adsorption capacity, since an excessive electrostatic adsorption capacity as described above is not required. In the case of the electrostatic precipitation roller, the main factors affecting the electrostatic adsorption capacity are a minimum distance, an electrostatic voltage, and a radius of the electrostatic precipitation roller (more precisely, a radius of the cylindrical electrode and a radius of the entire electrostatic precipitation roller may be also divided), and the like. And the difficulty is high when the proper parameters are required to be manually set, so that the proper parameters are predicted by adopting a parameter prediction model based on a convolutional neural network model.

In addition, in order to further avoid damage to the surface of the material, electrostatic dust removal is performed in a non-contact mode, namely the minimum distance between the first electrostatic dust removal roller and the first article is larger than 0 cm; each of the sub-rollers of the second array of electrostatic precipitation rollers has a minimum separation from the first article of greater than 0 cm.

Furthermore, because the front side and the back side of common materials needing dust removal are provided with dust removal modes at the same time, the two sides of the common materials are provided with dust removal modes, namely the number of the first electrostatic dust removal rollers is two, and the two first electrostatic dust removal rollers are symmetrically arranged relative to the plane where the transmission belt is located; the number of second electrostatic precipitator roller array is two, and one of them second electrostatic precipitator roller array sets up in the top of transmission band, and another second electrostatic precipitator roller array sets up in the below of transmission band. Therefore, the pollution on the front surface of the material can be cleaned, the back surface of the material is cleaned at the same time, the machine fault caused by the dust brought in is avoided, and the integral dust removal efficiency is finally improved.

Further, the first electrostatic precipitation roller is also in contact with a preset precipitation silica gel rod, and the precipitation silica gel rod is used for transferring the dust adsorbed by the static electricity. The dedusting silica gel rod is a rod-shaped object, all or at least the surface of the dedusting silica gel rod is made of silica gel, and the first electrostatic dedusting roller can transfer adsorbed dust to the dedusting silica gel rod after rotating for one circle. Wherein, the surface of the first electrostatic precipitation roller is preferably coated with a silica gel film.

Further, in the dust-sprinkling process of the first pattern of the present application, that is, according to the first pattern drawn in advance, the first article placed on the conveyor belt in advance is subjected to dust-sprinkling process, so that when the first pattern is presented by the dust on the first article, the corresponding thickness of the dust at each position on the first pattern also needs to be considered.

As described in the above steps S1-S4, the first article previously placed on the conveying belt is subjected to dust-dusting processing in accordance with the first pattern drawn in advance, so that the dust on the first article assumes the first pattern; shooting the surface of a first article to obtain a first image; starting a conveying belt, and adjusting the preset parameter of a first electrostatic precipitation roller to be a preset first parameter, so that the first article is subjected to primary precipitation treatment by the first electrostatic precipitation roller; the first electrostatic precipitation roller is a fixed precipitation roller; calling a corresponding relation table of a preset dust pattern and a dust removal position to obtain a specified dust removal position corresponding to the first pattern; adjusting the preset parameter of a movable second electrostatic precipitation roller array to be a preset second parameter, placing the second electrostatic precipitation roller array at a specified precipitation position, and performing secondary precipitation treatment when the first article passes through the second electrostatic precipitation roller array; shooting the surface of the first article to obtain a second image; correspondingly subtracting the color value of each pixel point in the second image from the color value of each pixel point in the first image respectively, thereby generating a third image; modifying the first parameter and the second parameter for multiple times, and repeating the steps S1-S4 to obtain multiple third images; according to a preset pattern similarity calculation method, calculating a similarity value between the third image and the first pattern, selecting a designated third image corresponding to the maximum similarity value, acquiring a numerical value of a first parameter and a numerical value of a second parameter corresponding to the designated third image, and constructing a first parameter group consisting of the first image, the numerical value of the first parameter corresponding to the designated third image and the numerical value of the second parameter.

The steps S1-S6 of the present application are to obtain training data, because the present application uses the prediction model of basic machine learning to predict suitable parameters, and therefore needs sufficient training data, and the manner of obtaining the training data is special, which is a place different from the ordinary machine learning. Steps S1-S5 are for generating a specific sample data, and steps S1-S4 are for generating a basic data of the specific sample data. The sample data of the present application is called as special sample data, on one hand, the present application adopts an automatic labeling mode to label the sample data, and the common scheme of machine learning adopts a manual labeling mode to label the sample data. The automatic labeling means that the first parameter group generated by the application is sample data, and the numerical value of the first parameter and the numerical value of the second parameter are corresponding labels, so that manual labeling processing is not needed, and the labeling is based on the objective effect of dust removal, so that the labeling accuracy is higher.

The first pattern may be any feasible pattern, and one pattern corresponds to one sample data, i.e. in the subsequent steps, the first pattern is modified to become a second pattern, etc., so that no excessive restriction on the pattern is needed. One characteristic of this application lies in, can carry out the dust and stamp out the face, and this kind stamps out the dust on the article surface that the processing obtained, can be with in the actual production life unanimous, also can be inconsistent, because the dust that can follow-up adopt multiple pattern stamps out, consequently can obtain more types of sample data to improve the robustness of the parameter prediction model that the training obtained. And shooting the surface of the first article to obtain a first image. The first image at this time is a first image with dust attached thereto, and the dust has a first pattern.

Further, performing dust-sprinkling treatment on the first article which is placed on the conveying belt in advance according to a first pre-drawn pattern, so that dust on the first article presents the first pattern; before the step S1 of performing the shooting process on the surface of the first article to obtain the first image, the method further includes: and carrying out manual fine dust removal treatment so that dust is not adsorbed on the surface of the first article. The manual fine dust removal processing at this time seems to be contradictory to the subsequent double electrostatic dust removal, but actually, this is not true because the steps S1-S5 are for obtaining suitable sample data, not an actual dust removal process, and therefore the manual fine dust removal processing is for ensuring that the steps S1-S5 can obtain more accurate training data, and the actual dust removal process does not include the manual fine dust removal step.

This application utilizes the transmission band, carries out the article transmission to pass through first electrostatic precipitator roller and second electrostatic precipitator roller array in proper order at the in-process of article transmission, thereby realize dual electrostatic precipitator. The parameters of the electrostatic precipitation roller include a minimum distance, an electrostatic voltage, a radius of the electrostatic precipitation roller (more precisely, the parameters may be divided into a radius of the cylindrical electrode and a radius of the entire electrostatic precipitation roller), and the like. Generally, however, the first parameter will be less than a second parameter of a second, subsequent array of electrostatic precipitation rollers (or the first parameter will provide a lower electrostatic attraction force than the second parameter) in order to protect the surface material of the article.

Wherein, first electrostatic precipitator roller is fixed dust removal roller, and this is a characteristic of comparing in second electrostatic precipitator roller array, and the length of first electrostatic precipitator roller should be bigger than the width of first article, and even the length of first electrostatic precipitator roller should be bigger than the width of transmission band to provide comprehensive first heavy basis and remove dust.

And calling a preset corresponding relation table of the dust patterns and the dust removal positions to acquire a specified dust removal position corresponding to the first pattern, so that the electrostatic dust removal roller can perform targeted dust removal at a proper position. The generation method of the correspondence table may be any feasible method, for example, a method set manually, because the dust pattern is preset, it is predictable which positions are more difficult to remove the dust. Alternatively, an existing image recognition technology is used to perform recognition processing on the first image or the first pattern to obtain dust positions (e.g., which positions have thicker dust, etc.) that are difficult to remove, so as to obtain the specified dust removal position. Further, since the first pattern is drawn in advance, which positions are thick with dust can be predicted, and the correspondence table can be generated more easily.

The second electrostatic precipitation roller array is formed by a plurality of sub-rollers of smaller size, each sub-roller having a similar structure to the first electrostatic precipitation roller, with the main difference that the sub-rollers have a smaller axial length, e.g. each sub-roller has an axial length less than half the axial length of the first electrostatic precipitation roller. The parameters of the second electrostatic precipitator roller array, i.e. the parameters of each sub-roller, are similar to the parameters of the first electrostatic precipitator roller, and are not described herein again. Since the second array of electrostatic precipitation rollers should remove dust that cannot be removed by the first electrostatic precipitation rollers, the second parameter should be such that the second array of electrostatic precipitation rollers has a greater electrostatic attraction, e.g. a certain sub-roller is closer to the first article.

The second array of electrostatic precipitation rollers is movable so that it can be moved to a designated precipitation position. The power for moving the second electrostatic precipitation roller array can be provided by a preset motor, and the motor can enable the sub-rollers in the second electrostatic precipitation roller array to move in the horizontal direction through a mechanical force transmission assembly (gears, belts and the like). This power transmission is a common mechanical power transmission and is not deployed here. After the secondary dust removal processing, the surface of the first article is photographed to obtain a second image, so that if the dust removal effect is ideal, the second image should reflect the surface of the first article without dust.

And correspondingly subtracting the color value of each pixel point in the second image from the color value of each pixel point in the first image respectively, thereby generating a third image. The third image at this time can reflect the quality of the dust removal effect. For example, when there is still much dust not removed, the third image will have a large difference from the first pattern; when the dust is removed clean, then the third image will be almost identical to the first pattern; when the dust is not only removed cleanly but also a portion of the surface of the article is damaged, the third image will have a certain difference from the first pattern. Therefore, the dust removal effect can be automatically obtained by comparing the third image with the first pattern, so that the part for manually verifying the dust removal effect is omitted, and the efficiency and the accuracy of subsequent sample data labeling are improved. Note that the third image is compared with the first pattern at this time, and not the third image with the first pattern.

It is temporarily impossible to determine whether the parameter at this time is the optimal parameter, so the present application modifies the first parameter and the second parameter a plurality of times, and repeats steps S1-S4 to obtain a plurality of third images. So that optimum parameters can be selected from these third images. The specific process is as follows: according to a preset pattern similarity calculation method, calculating a similarity value between the third image and the first pattern, selecting a designated third image corresponding to the maximum similarity value, acquiring a numerical value of a first parameter and a numerical value of a second parameter corresponding to the designated third image, and constructing a first parameter group consisting of the first image, the numerical value of the first parameter corresponding to the designated third image and the numerical value of the second parameter. At this time, the parameters in the first parameter group are the optimal parameters corresponding to the first image, so the first parameter group can be used as sample data for use in the subsequent training process of machine learning.

Further, according to a first pre-drawn pattern, carrying out dust splashing treatment on a first article which is placed on a conveying belt in advance so as to enable dust on the first article to present the first pattern; the step S1 of performing a photographing process on the surface of the first article to obtain a first image includes:

s101, dyeing the pre-collected dust into a specified color;

s103, shooting the surface of the first article to obtain a first image.

So as to improve the efficiency and accuracy of image contrast. Because the dust is in the designated color, the surface of the first article is also in the first pattern in the designated color, and therefore the subsequent judgment of the generation of the third image and the similarity calculation between the third image and the first pattern are easier and quicker.

s503, according to the formula:

calculating a similarity value R between one third image and the first pattern so as to obtain a plurality of similarity values respectively corresponding to the plurality of third images; wherein Pi is the value of the ith vector in the first vector, Qi is the value of the ith vector in the second vector, the first vector and the second vectorThe vectors are m-dimensional vectors, m is an integer greater than 1, and a, b and c are all preset parameters greater than 0;

The method adopts a special similarity calculation formula to calculate the similarity value between two images. In the method, the color values of the plurality of third images are respectively converted into a plurality of first vectors according to a preset vector conversion method, and the color values of the first patterns are simultaneously converted into second vectors by any feasible method, but the conversion methods of the first vectors and the second vectors should be the same. The method is, for example, to use the color value of one pixel or the sum of the color values of a plurality of pixels as a numerical value of one dimension, thereby forming a high-dimension vector. Then according to the formula:

and calculating the similarity value R between one third image and the first pattern, thereby being compatible with the numerical difference and the angle difference between the vectors corresponding to the two images and improving the calculation accuracy. And then the first parameter and the second parameter corresponding to the designated similarity value with the maximum numerical value and the first image form a first parameter group together.

Performing n times of pattern modification processing to modify the first pattern into a second pattern, a third pattern, … and an nth pattern, and repeating the steps S1 to S5 to obtain a parameter set II, a parameter set III, … and a parameter set n, as described in the above steps S6 to S9; n is an integer greater than 3; calling a preset convolutional neural network model, and training the convolutional neural network model in a semi-supervised learning mode by adopting the first parameter group, the second parameter group, the third parameter group, … and the n parameter group as sample data to obtain a parameter prediction model; shooting the surface of a second object to be dedusted to obtain an image to be analyzed; inputting an image to be analyzed into the parameter prediction model for processing to obtain a first prediction parameter and a second prediction parameter output by the parameter prediction model; wherein the second item is of the same type as the first item; and adjusting the parameters of the first electrostatic precipitation roller as first prediction parameters, adjusting the parameters of the second electrostatic precipitation roller array as second prediction parameters, and performing precipitation treatment on the second article.

Although one sample data can be obtained in the foregoing step S5, a plurality of sample data are required in the training process of machine learning, and the more the number of sample data is, the higher the accuracy of the trained model is. Therefore, the present application performs the pattern modification process n times to modify the first pattern into the second pattern, the third pattern, …, the nth pattern, and repeats steps S1 to S5, thereby obtaining n parameter sets. And then, a convolutional neural network model suitable for image analysis is used as a basic model, n parameter groups are respectively used as n sample data, and training processing is carried out, so that a parameter prediction model is obtained. The convolutional neural network model is suitable for classifying image data, and the parameter prediction model of the present application can be actually regarded as a classification model, that is, an input image to be analyzed is classified into a suitable type, and the suitable type refers to a type corresponding to the first prediction parameter and the second prediction parameter, which is why the present application needs to perform parameter labeling in an automatic labeling manner.

When the device is actually used, surface shooting is carried out on a second object needing dust removal so as to obtain an image to be analyzed, and appropriate first prediction parameters and second prediction parameters can be predicted by using a parameter prediction model. The second article is the same as the first article in type, so that the adaptability of the parameter prediction model is ensured. The first prediction parameter and the second prediction parameter are suitable parameters of the electrostatic precipitation roller, so that the parameter of the first electrostatic precipitation roller is adjusted to be the first prediction parameter, the parameter of the second electrostatic precipitation roller array is adjusted to be the second prediction parameter, and the second article is subjected to precipitation treatment, so that proper precipitation treatment can be realized, and the method is not only suitable for surface precipitation of common articles, but also suitable for surface precipitation of special articles with weak binding force between surface materials and the surface.

Further, the step S7 of calling a preset convolutional neural network model, training the convolutional neural network model in a semi-supervised learning manner by using the first parameter group, the second parameter group, the third parameter group, …, and the n parameter group as sample data, and obtaining a parameter prediction model includes:

And then, the sample data is divided to obtain a plurality of training data and a plurality of verification data. And training, and verifying by adopting homologous verification data to ensure that the obtained final parameter prediction model can adapt to the parameter prediction task. The training process can adopt a random gradient descent method and a back propagation algorithm to update the parameters of the network layer. Since the sample data are all images marked with the most suitable parameters, the training process of the application is actually obtained by training in a semi-supervised learning mode.

According to the double electrostatic dust collection method based on the parameter prediction model, dust is splashed on a first article; shooting to obtain a first image; adjusting parameters of the first electrostatic precipitation roller, and performing primary precipitation treatment; acquiring a specified dust removal position; adjusting parameters of a movable second electrostatic dust removal roller array, and performing secondary dust removal treatment; shooting to obtain a second image; generating a third image; modifying the first parameter and the second parameter for multiple times to obtain multiple third images; calculating a similarity value, selecting a designated third image corresponding to the maximum similarity value, and constructing a first parameter group; carrying out n times of pattern modification processing to obtain n parameter sets; training a convolutional neural network model in a semi-supervised learning mode by using a parameter group as sample data to obtain a parameter prediction model; obtaining an image to be analyzed; inputting an image to be analyzed into the parameter prediction model to obtain a first prediction parameter and a second prediction parameter; and adjusting the parameter of the first electrostatic precipitation roller as a first prediction parameter, adjusting the parameter of the second electrostatic precipitation roller array as a second prediction parameter, and performing precipitation treatment on the second article, so that the aim of improving the precipitation effect is fulfilled.

Referring to fig. 2, an embodiment of the present application provides a dual electrostatic precipitator device based on a parameter prediction model, including:

a first image capturing unit 10 for performing step S1 of performing dust-dusting processing on a first article placed in advance on the conveying belt in accordance with a first pattern drawn in advance so that dust on the first article assumes the first pattern; shooting the surface of a first article to obtain a first image;

a primary dust removal processing unit 20, configured to execute step S2, start the conveyor belt, adjust a preset parameter of the first electrostatic dust removal roller to a preset first parameter, and perform primary dust removal processing on the first article by using the first electrostatic dust removal roller; the first electrostatic precipitation roller is a fixed precipitation roller;

a secondary dust removal processing unit 30, configured to execute step S3, retrieve a preset correspondence table between dust patterns and dust removal positions, so as to obtain a specified dust removal position corresponding to the first pattern; adjusting the preset parameter of a movable second electrostatic precipitation roller array to be a preset second parameter, placing the second electrostatic precipitation roller array at a specified precipitation position, and performing secondary precipitation treatment when the first article passes through the second electrostatic precipitation roller array; shooting the surface of the first article to obtain a second image;

the color value subtraction unit 40 of the pixel point is configured to execute step S4, and correspondingly subtract the color value of each pixel point in the second image from the color value of each pixel point in the first image, so as to generate a third image;

a pattern similarity calculation unit 50 for performing step S5, modifying the first parameter and the second parameter a plurality of times, and repeating steps S1-S4 to obtain a plurality of third images; according to a preset pattern similarity calculation method, calculating a similarity value between the third image and the first pattern, selecting a designated third image corresponding to the maximum similarity value, acquiring a numerical value of a first parameter and a numerical value of a second parameter corresponding to the designated third image, and constructing a first parameter group consisting of the first image, the numerical value of the first parameter corresponding to the designated third image and the numerical value of the second parameter;

a pattern modification processing unit 60, configured to perform step S6, perform n times of pattern modification processing to modify the first pattern into a second pattern, a third pattern, …, an nth pattern, and repeat steps S1-S5 to obtain a parameter set number two, a parameter set number three, …, a parameter set number n; n is an integer greater than 3;

a parameter prediction model obtaining unit 70, configured to execute step S7, invoke a preset convolutional neural network model, and train the convolutional neural network model in a semi-supervised learning manner by using the first parameter group, the second parameter group, the third parameter group, …, and the n parameter group as sample data, so as to obtain a parameter prediction model;

the image to be analyzed shooting unit 80 is configured to execute step S8, and perform shooting processing on the surface of the second object to be dedusted to obtain an image to be analyzed; inputting an image to be analyzed into the parameter prediction model for processing to obtain a first prediction parameter and a second prediction parameter output by the parameter prediction model; wherein the second item is of the same type as the first item;

the prediction parameter adjusting unit 90 is configured to execute step S9, adjust the parameter of the first electrostatic precipitation roller to be the first prediction parameter, adjust the parameter of the second electrostatic precipitation roller array to be the second prediction parameter, and perform the precipitation process on the second article.

The operations performed by the above units are in one-to-one correspondence with the steps of the parameter prediction model-based dual electrostatic precipitator method according to the foregoing embodiment, and are not described herein again.

According to the double electrostatic dust removal device based on the parameter prediction model, dust splashing treatment is carried out on a first article; shooting to obtain a first image; adjusting parameters of the first electrostatic precipitation roller, and performing primary precipitation treatment; acquiring a specified dust removal position; adjusting parameters of a movable second electrostatic dust removal roller array, and performing secondary dust removal treatment; shooting to obtain a second image; generating a third image; modifying the first parameter and the second parameter for multiple times to obtain multiple third images; calculating a similarity value, selecting a designated third image corresponding to the maximum similarity value, and constructing a first parameter group; carrying out n times of pattern modification processing to obtain n parameter sets; training a convolutional neural network model in a semi-supervised learning mode by using a parameter group as sample data to obtain a parameter prediction model; obtaining an image to be analyzed; inputting an image to be analyzed into the parameter prediction model to obtain a first prediction parameter and a second prediction parameter; and adjusting the parameter of the first electrostatic precipitation roller as a first prediction parameter, adjusting the parameter of the second electrostatic precipitation roller array as a second prediction parameter, and performing precipitation treatment on the second article, so that the aim of improving the precipitation effect is fulfilled.

Referring to fig. 3, an embodiment of the present invention further provides a computer device, where the computer device may be a server, and an internal structure of the computer device may be as shown in the figure. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the computer designed processor is used to provide computational and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used for storing data used by the parameter prediction model-based dual electrostatic precipitation method. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of dual electrostatic precipitation based on a parametric predictive model. The computer device further comprises a display screen and an input device for displaying the human interactive interface and for receiving input data, respectively.

The processor executes the parameter prediction model-based dual electrostatic precipitation method, wherein the steps included in the method correspond to the steps of the parameter prediction model-based dual electrostatic precipitation method of the foregoing embodiment one to one, and are not described herein again.

It will be understood by those skilled in the art that the structures shown in the drawings are only block diagrams of some of the structures associated with the embodiments of the present application and do not constitute a limitation on the computer apparatus to which the embodiments of the present application may be applied.

The computer equipment performs dust sprinkling treatment on a first article; shooting to obtain a first image; adjusting parameters of the first electrostatic precipitation roller, and performing primary precipitation treatment; acquiring a specified dust removal position; adjusting parameters of a movable second electrostatic dust removal roller array, and performing secondary dust removal treatment; shooting to obtain a second image; generating a third image; modifying the first parameter and the second parameter for multiple times to obtain multiple third images; calculating a similarity value, selecting a designated third image corresponding to the maximum similarity value, and constructing a first parameter group; carrying out n times of pattern modification processing to obtain n parameter sets; training a convolutional neural network model in a semi-supervised learning mode by using a parameter group as sample data to obtain a parameter prediction model; obtaining an image to be analyzed; inputting an image to be analyzed into the parameter prediction model to obtain a first prediction parameter and a second prediction parameter; and adjusting the parameter of the first electrostatic precipitation roller as a first prediction parameter, adjusting the parameter of the second electrostatic precipitation roller array as a second prediction parameter, and performing precipitation treatment on the second article, so that the aim of improving the precipitation effect is fulfilled.

An embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored thereon, and when the computer program is executed by a processor, the method for dual electrostatic dust collection based on a parameter prediction model is implemented, where steps included in the method correspond to steps of the method for dual electrostatic dust collection based on a parameter prediction model in the foregoing embodiment one to one, and are not described herein again.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, apparatus, article, or method that includes the element.

The above description is only a preferred embodiment of the present application, and not intended to limit the scope of the present application, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the specification and the drawings of the present application, or which are directly or indirectly applied to other related technical fields, are also included in the scope of the present application.

Claims

1. A double electrostatic dust collection method based on a parameter prediction model is characterized by comprising the following steps:

2. The parametric prediction model-based dual electrostatic precipitator method according to claim 1, wherein the first parameters comprise a minimum distance between the first electrostatic precipitator roller and the first article, an electrostatic voltage of the first electrostatic precipitator roller, and a radius of the first electrostatic precipitator roller;

3. The parametric prediction model-based dual electrostatic precipitator method of claim 2, wherein the minimum spacing of the first electrostatic precipitation roller from the first article is greater than 0 cm; each of the sub-rollers of the second array of electrostatic precipitation rollers has a minimum separation from the first article of greater than 0 cm.

4. The parameter prediction model-based dual electrostatic precipitation method of claim 1, wherein the number of the first electrostatic precipitation rollers is two, and the first electrostatic precipitation rollers are symmetrically arranged relative to the plane of the conveyor belt;

5. The dual electrostatic precipitation method based on the parameter prediction model according to claim 1, wherein the first electrostatic precipitation roller is further in contact with a preset precipitation silica gel stick for transferring the electrostatically adsorbed dust.

6. The parameter prediction model-based dual electrostatic precipitator method according to claim 1, wherein the dust-dusting process is performed on a first article previously placed on the conveyor according to a first pre-drawn pattern, so that dust on the first article takes on the first pattern; the step S1 of performing a photographing process on the surface of the first article to obtain a first image includes:

s101, dyeing the pre-collected dust into a specified color;

s103, shooting the surface of the first article to obtain a first image.

7. The parametric prediction model-based dual electrostatic precipitator method according to claim 1, wherein the first parameter and the second parameter are modified a plurality of times, and steps S1-S4 are repeated to obtain a plurality of third images; step S5, executed by the computer, of calculating a similarity value between the third image and the first pattern according to a preset pattern similarity calculation method, selecting a designated third image corresponding to a maximum similarity value, obtaining a value of a first parameter and a value of a second parameter corresponding to the designated third image, and constructing a parameter group i consisting of the first image, the value of the first parameter corresponding to the designated third image, and the value of the second parameter, includes:

s503, according to the formula:

8. The parameter prediction model-based dual electrostatic precipitator method according to claim 1, wherein the step S7 of retrieving a preset convolutional neural network model, and using the parameter group i, the parameter group ii, the parameter group iii, …, and the parameter group n as sample data to train the convolutional neural network model in a semi-supervised learning manner, so as to obtain a parameter prediction model, includes:

s701, automatically labeling sample data to obtain n sample data, wherein the images in the first parameter group are labeled with the numerical value of the first parameter and the numerical value of the second parameter in the first parameter group, the images in the second parameter group are labeled with the numerical value of the first parameter and the numerical value of the second parameter in the second parameter group, …, and the images in the n parameter group are labeled with the numerical value of the first parameter and the numerical value of the second parameter in the n parameter group;

9. The utility model provides a dual electrostatic precipitator device based on parameter prediction model which characterized in that includes:

10. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 8 when executing the computer program.