A boiler temperature field real-time monitoring method, device and system based on image recognition

Abstract

The application discloses a kind of based on image recognition's boiler temperature field real-time monitoring method, device and system, method includes: obtaining detection data in boiler, and the detection data is digitized, obtains boiler combustion image;According to the boiler combustion image, determine the three primary color value corresponding to any one pixel point;Each three primary color value is sequentially input into temperature measurement grid model, and the temperature value corresponding to each pixel point is output, and the temperature field distribution of the boiler is generated based on the temperature value.The present application solves the technical problems that the amount of temperature information collected in the prior art is too small, it is difficult to accurately reflect the combustion state in the furnace, and the temperature distribution of the entire hearth cannot be described.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of boiler detection, and in particular to a boiler temperature field real-time monitoring method, device and system based on image recognition. BACKGROUND

[0002] The combustion temperature measurement in the furnace is one of the key factors affecting the operation efficiency of power station boilers, so accurate detection of the combustion temperature in the furnace is of great significance. With the increase in energy demand and the improvement in environmental protection standards, efficient and stable operation of boilers has become particularly important.

[0003] Real-time monitoring of the combustion temperature field can help optimize the combustion process, reduce fuel consumption, reduce harmful gas emissions, and prevent boilers from malfunctioning due to overheating or uneven temperature. Currently, power station boilers in operation usually use thermocouples arranged in the furnace to measure temperature, and the combustion state in the furnace is determined by the temperature information collected by the thermocouples. However, due to the heat resistance of the temperature sensing element, the thermocouples can only measure for a short time, and when there are problems such as flame deflection, uneven temperature distribution, and upward and downward shift of the high-temperature area in the furnace, the amount of temperature information collected is too small due to the limited number of thermocouples arranged in the furnace, making it difficult to accurately reflect the combustion state in the furnace, and thus unable to depict the temperature distribution of the entire furnace. SUMMARY

[0004] The present application provides a boiler temperature field real-time monitoring method, device and system based on image recognition to solve the technical problem that the amount of temperature information collected is too small in the prior art, making it difficult to accurately reflect the combustion state in the furnace and unable to depict the temperature distribution of the entire furnace.

[0005] To solve the above technical problems, the present application embodiment provides a boiler temperature field real-time monitoring method based on image recognition, comprising:

[0006] Real-time acquisition of detection data in the boiler, and digital processing of the detection data to obtain a boiler combustion image;

[0007] According to the boiler combustion image, determining the three primary color values corresponding to any one pixel point;

[0008] Inputting each of the three primary color values into a temperature measurement grid model in sequence, outputting to obtain the temperature value corresponding to each pixel point, and generating the temperature field distribution of the boiler based on the temperature value.

[0009] As a preferred scheme, the real-time acquisition of detection data in the boiler, and the digital processing of the detection data to obtain a boiler combustion image, specifically comprises:

[0010] The detection signal of the one-dimensional time sequence is digitally processed to discretize the detection signal of the one-dimensional time sequence into a detection data signal.

[0011] The detection signal of the one-dimensional time sequence is digitally processed to discretize the detection signal of the one-dimensional time sequence into a detection data signal.

[0012] According to the detection digital signal, an image is generated to obtain a boiler combustion image.

[0013] As a preferred solution, the detection signal of the one-dimensional time sequence is digitally processed to discretize the detection signal of the one-dimensional time sequence into a detection data signal, specifically including:

[0014] The detection signal of the one-dimensional time sequence is converted into a frequency domain signal, and the converted frequency domain signal is extracted to obtain a corresponding amplitude feature.

[0015] The corresponding continuous amplitude feature is binned to obtain a detection data signal related to the amplitude size.

[0016] The detection signal of the one-dimensional time sequence is extracted to obtain a cutoff frequency and an original distribution, and the extracted cutoff frequency and original distribution are converted to obtain a detection data signal related to a spatial coordinate.

[0017] The detection data signal includes signals related to the amplitude size and the spatial coordinate.

[0018] As a preferred solution, the three primary color values corresponding to any one pixel point are determined based on the boiler combustion image, specifically including:

[0019] Based on a true color image card, color recognition is performed on the boiler combustion image to obtain the gray scale quantization values of the red, green and blue three colors corresponding to each pixel point coordinate in the boiler combustion image as the three primary color values corresponding to any one pixel point.

[0020] As a preferred solution, each of the three primary color values is sequentially input into a temperature measurement grid model to output a temperature value corresponding to each pixel point, and a temperature field distribution of the boiler is generated based on the temperature value, specifically including:

[0021] Each of the three primary color values is input into a temperature measurement grid model as an input of the temperature measurement grid model, so that the temperature measurement grid model calculates the input three primary color values to output a two-dimensional temperature value corresponding to each pixel point in the boiler combustion image.

[0022] Based on the two-dimensional temperature value and the corresponding spatial coordinate, a temperature field distribution of the boiler is generated.

[0023] As a preferred solution, the method further comprises:

[0024] dividing the pixel points in the boiler combustion image into regions, and determining a middle pixel point in each region as a target pixel point according to the regions of the divided pixel points; wherein the number and size of the pixel points in each region are equal;

[0025] inputting the three primary color values corresponding to the target pixel points into the temperature measurement grid model respectively, so that the temperature measurement grid model calculates the input three primary color values and outputs two-dimensional temperature values corresponding to each target pixel point in the boiler combustion image;

[0026] performing edge smoothing on the target pixel points in each region and the target pixel points in adjacent regions respectively based on the region where each target pixel point is located, so as to calculate two-dimensional temperature values corresponding to the remaining pixel points in each region respectively;

[0027] generating the temperature field distribution of the boiler by using the two-dimensional temperature values of all pixel points and their corresponding spatial coordinates.

[0028] As a preferred solution, the method for constructing the temperature measurement grid model comprises:

[0029] obtaining sample data and performing normalization processing on the sample data; wherein the sample data comprises a plurality of groups of flame radiation temperatures and three primary color values corresponding to each flame radiation temperature respectively;

[0030] setting an initial temperature measurement network model and determining a penalty factor, a Gaussian kernel function and a kernel parameter of the initial temperature measurement network model;

[0031] training the set initial temperature measurement network model based on the sample data, so as to obtain a best penalty factor, a best kernel parameter and a minimum mean standard error corresponding to the initial temperature measurement network model;

[0032] updating the best penalty factor, the best kernel parameter and the minimum mean standard error in the initial temperature measurement network model, so as to obtain a final temperature measurement grid model through training.

[0033] As a preferred solution, the method for obtaining sample data comprises:

[0034] obtaining a black furnace flame image and a plurality of groups of flame radiation temperatures in a preset black furnace;

[0035] Based on the true color image card, color recognition is performed on the black furnace flame image, and the gray quantization values corresponding to the red, green and blue colors of each pixel coordinate in the black furnace flame image are obtained.

[0036] Based on the monochromatic emissivity of the preset black furnace, the flame radiation temperature and the wavelengths corresponding to the three colors, the correction coefficients of any two components in the three colors are determined.

[0037] According to the correction coefficients and the wavelengths corresponding to the three colors, and combined with the monochromatic emissivity of the flame, the monochromatic radiation intensity of the burning flame spectrum of each color is determined, so as to obtain the light intensity signal value under the wavelength corresponding to each color light.

[0038] Based on the light intensity signal value, a relationship between the corresponding temperature and the light wavelength and the monochromatic emissivity of the preset black furnace is obtained, and the gray quantization values corresponding to the three colors of the black furnace flame image are calibrated according to the relationship and the correction coefficients, so as to obtain the three primary color values corresponding to each flame radiation temperature.

[0039] As a preferred scheme, the three primary color values corresponding to any one pixel point are determined according to the boiler combustion image, and the method further comprises the following steps:

[0040] Based on the true color image card, color recognition is performed on the black furnace flame image, and the gray quantization values corresponding to the red, green and blue colors of each pixel coordinate in the black furnace flame image are obtained.

[0041] According to the relationship and the correction coefficients, the gray quantization values corresponding to the three colors of the black furnace flame image are calibrated, so as to determine the three primary color values corresponding to any one pixel point.

[0042] Correspondingly, the application also provides a boiler temperature field real-time monitoring device based on image recognition, which comprises an acquisition module, a three primary color module and a temperature module.

[0043] The acquisition module is used for acquiring detection data in the boiler in real time, and performing digital processing on the detection data to obtain a boiler combustion image.

[0044] The three primary color module is used for determining the three primary color values corresponding to any one pixel point according to the boiler combustion image.

[0045] The temperature module is used for inputting each three primary color value into a temperature measurement grid model in sequence, outputting the temperature value corresponding to each pixel point, and generating the temperature field distribution of the boiler based on the temperature value.

[0046] Correspondingly, the application also provides a boiler temperature field real-time monitoring method system based on image recognition, comprising: an optical system, an air cooling system, a combustion state acquisition system and an image processing system.

[0047] The optical system comprises: a shell, an image transmission fiber penetrating a front-end cavity of the shell and a mirror rod; wherein the shell comprises a front-end cavity and a rear-end cavity, and a front end of the front-end cavity is provided with a multi-hole plate air outlet.

[0048] The air cooling system comprises: a cooler and a cooling air inlet and a cooling air outlet connected with the optical system respectively.

[0049] The combustion state acquisition module is connected with the second end of the mirror rod, comprising: a camera arranged in the rear-end cavity of the shell, a color filter arranged in front of the camera, a microcomputer arranged in the rear-end cavity of the shell, an image acquisition card connected with the microcomputer and arranged on the rear-end cavity wall of the shell and a power supply arranged in the rear-end cavity of the shell.

[0050] The image processing system is used for executing the boiler temperature field real-time monitoring method based on image recognition as claimed in any one of the above, comprising: an image acquisition card and an image processing unit; the image acquisition card is used for signal conditioning and analog-digital signal conversion; the image processing unit is used for image denoising, image enhancement processing and temperature field calculation.

[0051] As a preferred scheme, the first end of the image transmission fiber is connected with a collimating lens, the second end of the image transmission fiber is connected with a color filter, and the image transmission fiber is wrapped with high-temperature-resistant protective material.

[0052] The front end of the mirror rod is provided with a lens, and a plurality of optical lenses and a double-band filter are arranged in the mirror rod at intervals.

[0053] As a preferred scheme, the power supply is connected to the camera and the microcomputer through a power switch.

[0054] As a preferred scheme, a plurality of multi-hole support frames are arranged in the front-end cavity of the shell along the direction of the first end and the second end of the front-end cavity, and the multi-hole support frames are used to support the image transmission fiber and the mirror rod.

[0055] As a preferred scheme, a fireproof plate is further arranged outside the front-end cavity of the shell, and a handheld handle is further arranged at the position of the second end outside the front-end cavity of the shell.

[0056] Correspondingly, the present application also provides a terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the image recognition based real-time boiler temperature field monitoring method according to any one of the above when executing the computer program.

[0057] Correspondingly, the present application also provides a computer readable storage medium comprising a stored computer program, wherein the computer readable storage medium controls the device where the computer readable storage medium is located to execute the image recognition based real-time boiler temperature field monitoring method according to any one of the above when the computer program runs.

[0058] Compared with the prior art, the embodiment of the present application has the following beneficial effects:

[0059] The technical scheme of the present application acquires detection data in the boiler in real time, thereby digitally processing the corresponding detection data, obtaining the corresponding boiler combustion image, and further determining the three primary color values of any point in the image, so that each three primary color value can be input into the temperature measurement grid model in turn, and the temperature value corresponding to each pixel point can be output, and the temperature field distribution of the boiler can be generated based on the temperature value, thereby accurately and quickly obtaining the distribution of the flame temperature field in the current boiler furnace, and the temperature field state in the boiler can also be detected comprehensively and accurately in a timely manner, avoiding the problems that the existing device can only be measured for a short time, and the combustion state in the furnace cannot be accurately reflected, so that the temperature distribution of the entire furnace cannot be depicted. BRIEF DESCRIPTION OF DRAWINGS

[0060] In order to more clearly illustrate the technical scheme in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0061] Figure 1 : a flow chart of the image recognition based real-time boiler temperature field monitoring method provided by the embodiment of the present application;

[0062] Figure 2 : a structural diagram of the image recognition based real-time boiler temperature field monitoring method device provided by the embodiment of the present application;

[0063] Figure 3 : a structural diagram of the image recognition based real-time boiler temperature field monitoring method system provided by the embodiment of the present application;

[0064] Figure 4A collimating lens light ray incidence schematic diagram provided by the embodiment of the present application;

[0065] Figure 5 A double band-pass filter and a CCD camera spectral response characteristic curve diagram provided by the embodiment of the present application;

[0066] In the drawings, the reference signs of the drawings of the specification are as follows:

[0067] 1-first interface; 2-power switch; 3-image processing system; 4-power supply; 5-CCD camera; 6-color filter; 7-housing; 8-cooling air outlet; 9-porous support frame; 10-image transmission optical fiber; 11-fire stop plate; 12-high temperature protective material; 13-collimating lens; 14-flame; 15-lens; 16-optical lens; 17-mirror rod; 18-cooler; 19-cooling air inlet; 20-handheld handle; 21-double band-pass filter; 22-microcomputer; 23-image acquisition card. DETAILED DESCRIPTION

[0068] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0069] Embodiment one

[0070] Please refer to Figure 1 A boiler temperature field real-time monitoring method based on image recognition provided by the embodiment of the present application includes the following steps S101-S103:

[0071] Step S101: Real-time acquisition of detection data in the boiler, and digital processing of the detection data to obtain a boiler combustion image.

[0072] In the embodiment, the corresponding detection data can be acquired by a CCD camera. Preferably, the CCD camera collects and outputs a one-dimensional time sequence signal, and therefore the detection data of the collected one-dimensional time sequence signal is digitally processed, so that the corresponding data is converted into a digital signal to obtain the corresponding boiler combustion image.

[0073] Step S102: Determination of three primary color values corresponding to any one pixel point according to the boiler combustion image.

[0074] In the embodiment, based on the boiler combustion image, preferably, a 24-bit true color image card is adopted, R, G and B of each pixel in the boiler combustion image represent gray quantization values of red, green and blue three colors at each coordinate, so that three primary color values corresponding to each pixel point are obtained. The 24-bit true color image card can distinguish one million colors, and the corresponding quantization level number is 256, and each color channel is 8 bits.

[0075] Step S103: sequentially input each three primary color value into the temperature measurement grid model, output the temperature value corresponding to each pixel point, and generate the temperature field distribution of the boiler based on the temperature value.

[0076] In the embodiment, the R, G and B values obtained by the image processing system are brought into the temperature measurement grid model of the established multi-layer perceptron neural network, and the cross-section temperature field distribution of the power plant boiler furnace is obtained. It should be noted that the network of the temperature measurement grid model is an SVM (Support Vector Machine) network, the number of input nodes is 3, and the number of output nodes is 1. Based on the characteristics of the SVM network, a 3-layer SVM network is constructed, the input layer is three, that is, the R, G and B three color values, and the output layer is one, that is, the two-dimensional temperature T of the flame radiation energy picture.

[0077] The above embodiment has the following effects:

[0078] The technical scheme of the present application can obtain the detection data in the boiler in real time, and then digitally process the corresponding detection data to obtain the corresponding boiler combustion image, and then determine the three primary color values of any point in the image, so that each three primary color value can be sequentially input into the temperature measurement grid model, and the temperature value corresponding to each pixel point can be output, and the temperature field distribution of the boiler can be generated based on the temperature value, so that the distribution of the flame temperature field in the current boiler furnace can be accurately and quickly obtained, and the temperature field state in the boiler can also be detected comprehensively and accurately in real time, avoiding the problems that the existing equipment can only be measured for a short time, and the combustion state in the furnace cannot be accurately reflected, so that the temperature distribution of the entire furnace cannot be depicted.

[0079] Embodiment two

[0080] The present embodiment is a preferred implementation of the boiler temperature field real-time monitoring method based on image recognition.

[0081] As a preferred scheme of the present embodiment, the detection data in the boiler is obtained in real time, and the detection data is digitally processed to obtain the boiler combustion image, which specifically includes:

[0082] The detection signal of one-dimensional time sequence is obtained by setting the lens in the boiler observation hole, adjusting the integral time, and according to the preset sampling frequency; the detection signal of one-dimensional time sequence is digitally processed, so that the detection signal of one-dimensional time sequence is discretized into a detection data signal; the detection digital signal is used for image generation, and a boiler combustion image is obtained.

[0083] In the embodiment, after the portable air compressor, the camera power switch and the computer are turned on, the whole system is started to run. Further, the lens that can be arranged in the boiler observation hole can be the lens of the CCD camera, and further, through the lens, the detection data information in the boiler can be obtained; preferably, the CCD camera can be controlled through the software interface program installed on the computer to start collecting radiation data and radiation image data, and the sampling frequency can be set to 1Hz; wherein, during the sampling process, the appropriate integral time can be adjusted, so as to ensure that the spectral data collected by the spectrometer and the image data collected by the CCD camera are not saturated, and have high signal-to-noise ratio.

[0084] In the embodiment, preferably, the CCD camera outputs a one-dimensional time sequence signal, which needs to be digitally processed, that is, a finite number array or a number table is used to represent a flame image (i.e. a boiler combustion image), and then each frame of video image is discretized from continuous analog signal to digital signal.

[0085] Specifically, the detection signal of one-dimensional time sequence is digitally processed, including amplitude digitalization and spatial coordinate digitalization.

[0086] As a preferred scheme of the embodiment, the detection signal of one-dimensional time sequence is digitally processed, so that the detection signal of one-dimensional time sequence is discretized into a detection data signal, specifically including:

[0087] The detection signal of one-dimensional time sequence is converted into a frequency domain signal, and the converted frequency domain signal is extracted to obtain the corresponding amplitude feature; the corresponding continuous amplitude feature is binned to obtain the detection data signal related to the amplitude; the detection signal of one-dimensional time sequence is extracted to obtain the cutoff frequency and the original distribution, and the corresponding detection data signal related to the spatial coordinate is obtained according to the extracted cutoff frequency and the original distribution; wherein, the detection data signal includes the signal related to the amplitude and the spatial coordinate.

[0088] In the embodiment, the digitalization of the image consists of two parts, which are amplitude digitalization and spatial coordinate digitalization, and the image sampling is the digitalization process of the spatial coordinate (x, y). For example, a flame image with a cutoff frequency f xg , f ygof the flame image, which has an original distribution f(x, y), will have a distribution f g (i, j) after sampling. It can be understood that image sampling is a process of digitizing continuous image space coordinates, that is, converting continuous space coordinates of an image into discrete pixel points. For a flame image, which has an original distribution f(x, y), the image distribution f g (i, j) after sampling is f g (i, j), where i and j represent the discretized x and y coordinates, respectively.

[0089] In this embodiment, the sampling frequency needs to be set, which determines the resolution of the image. In a two-dimensional image, there are usually two sampling frequencies f x and f y , corresponding to the sampling intervals in the x and y directions, respectively, and the embodiment is 1 Hz. In the sampling process, the continuous coordinates (x, y) of the image are mapped to the discrete pixel coordinates (i, j) by dividing the continuous coordinates by the sampling interval and taking the integer part. In the sampling process, the continuous gray scale or color value of the image also needs to be quantized to a limited value, in which the gray scale or color range of the image is divided into a limited number of levels, and a digital value is assigned to each level.

[0090] As a preferred scheme of this embodiment, according to the boiler combustion image, the three primary color values corresponding to any one pixel point are determined, which specifically includes:

[0091] Based on the true color image card, the color recognition of the boiler combustion image is performed, and the gray scale quantization values corresponding to the red, green and blue three colors of each pixel point coordinate in the boiler combustion image are obtained as the three primary color values corresponding to any one pixel point.

[0092] In this embodiment, the color information in the image can be gray scale quantized based on a 24-bit true color image card, so as to obtain a gray scale quantization value as a three primary color value. The 24-bit true color image card can distinguish one million colors, and the corresponding quantization level number is 256, and each color channel is 8 bits. The gray scale quantization values of the red, green and blue three colors at each coordinate are represented by R, G and B.

[0093] In this embodiment, each pixel point in the boiler combustion image corresponds to its corresponding spatial coordinate, and each spatial coordinate is the coordinate corresponding to the pixel point. Each pixel point has its corresponding gray scale quantization values of red, green and blue three colors, that is, the three primary color values corresponding to the pixel point.

[0094] As a preferred scheme of the embodiment, each of the three primary color values is sequentially input into the temperature measurement grid model, and a temperature value corresponding to each pixel point is output, and a temperature field distribution of the boiler is generated based on the temperature value, and specifically includes:

[0095] Each of the three primary color values is input into the temperature measurement grid model as an input, so that the temperature measurement grid model calculates the input three primary color values, and a two-dimensional temperature value corresponding to each pixel point in the boiler combustion image is output; and a temperature field distribution of the boiler is generated based on the two-dimensional temperature value and the corresponding spatial coordinates.

[0096] In the embodiment, the three primary color values respectively corresponding to each pixel point are input into the temperature measurement grid model as inputs, that is, the values corresponding to R, G and B of each pixel point are simultaneously input into the temperature measurement grid model as inputs, wherein the temperature measurement grid model has three input nodes, so that a two-dimensional temperature value corresponding to the pixel point can be directly output, and until all the pixel points in the image output the corresponding two-dimensional temperature values, a distribution map of the two-dimensional temperature values of the entire image can be obtained.

[0097] As a preferred scheme of the embodiment, each of the three primary color values is sequentially input into the temperature measurement grid model, and a temperature value corresponding to each pixel point is output, and a temperature field distribution of the boiler is generated based on the temperature value, and specifically further includes:

[0098] The pixel points in the boiler combustion image are regionally divided, and the middle pixel points of each region are determined as target pixel points according to the regions of the divided pixel points; wherein the number and size of the pixel points in each region are equal; the three primary color values respectively corresponding to the target pixel points are input into the temperature measurement grid model as inputs, so that the temperature measurement grid model calculates the input three primary color values, and a two-dimensional temperature value corresponding to each target pixel point in the boiler combustion image is output; the target pixel points in each region are edge smoothed with the target pixel points in the adjacent regions, based on the region where each target pixel point is located, so that the two-dimensional temperature values respectively corresponding to the remaining pixel points in each region are calculated; and a temperature field distribution of the boiler is generated through the two-dimensional temperature values of all the pixel points and the corresponding spatial coordinates.

[0099] In the embodiment, since the number of pixel points in the boiler combustion image is very large, if the pixel points in the boiler image are calculated in sequence by the temperature measurement grid model to obtain the corresponding temperature value, a huge amount of time will be consumed. By dividing the pixel points in the boiler combustion image into regions, preferably, the pixel points of 3*3 or 5*5 size can be taken as a region, so as to determine the middle pixel point of each region as a target pixel point, that is, the three primary color values corresponding to the target pixel point are taken as the input of the temperature measurement grid model, so that the temperature measurement grid model calculates the three primary color values to obtain the two-dimensional temperature value corresponding to the middle pixel point of each region.

[0100] In the embodiment, by performing temperature edge smoothing calculation on each region and its corresponding adjacent region, the temperature values corresponding to the remaining pixel points in each region except the middle pixel point can be directly calculated, so that a large amount of pixel point information can be avoided from being directly input into the temperature measurement grid model for calculation, the model calculation amount is reduced, the operation resources are optimized, and the accuracy of temperature detection can be ensured by the smoothing algorithm such as Gaussian filtering.

[0101] It can be understood that for a pair of boiler combustion images, the most important information for detecting the combustion temperature field is to obtain the temperature of the flame combustion, and the temperature inside the flame body during the flame combustion will not change too much, so the corresponding smoothing algorithm can be directly used to greatly reduce the corresponding calculation amount, and the temperature of the surrounding of the flame body will not change too much, so the corresponding calculation amount can also be greatly reduced. Since the temperature data of the flame body already exist, there will be a obvious temperature mutation edge between the flame body and the edge space thereof, for example, the position corresponding to the center pixel point in a region is the flame body, the temperature thereof is a, the position corresponding to the center pixel point in the adjacent region is the edge space, the temperature thereof is b, so the temperature between a and b greater than a preset value can be directly determined as the temperature mutation edge between the two regions, and the temperature mutation edge can be finally calculated by setting the corresponding smoothing weight and other coefficients to obtain the temperature value of the corresponding edge pixel point, so that a large amount of similar data calculation can be greatly avoided.

[0102] As a preferred scheme of the embodiment, the method for constructing the temperature measurement grid model comprises:

[0103] The sample data is acquired and normalized, wherein the sample data includes a plurality of groups of flame radiation temperatures and three primary color values corresponding to each flame radiation temperature; an initial temperature measurement network model is set, and a penalty factor, a Gaussian kernel function and a kernel parameter of the initial temperature measurement network model are determined; the initial temperature measurement network model after setting is trained based on the sample data, so as to obtain a best penalty factor, a best kernel parameter and a minimum mean standard error corresponding to the initial temperature measurement network model; the best penalty factor, the best kernel parameter and the minimum mean standard error are updated and set in the initial temperature measurement network model, so as to obtain a final temperature measurement grid model through training.

[0104] In the embodiment, the sample data is acquired and then normalized, and then the training of the temperature measurement grid model can be performed, wherein the temperature measurement grid model can be an SVM model. Then, the initial SVM model is set, that is, the input node number of the SVM network is selected as 3 and the output node number is selected as 1. The sample data used for model training is shown in Table 1.

[0105]

[0106] In the embodiment, the kernel function of the initial SVM model is selected as the RBF Gaussian kernel function:

[0107]

[0108] wherein σ is a standard deviation of the Gaussian function, affects the width of the kernel function, the greater the value of σ, the wider the kernel function, the smoother the decision boundary, the smaller the value of σ, the narrower the kernel function, the more complex the decision boundary; x and x i are kernel function values of two pixel data points, respectively.

[0109] In the embodiment, the value range of the penalty factor (c) is set as -10 to 10 and the value range of the kernel parameter (g) is set as -5 to 5 during the network training process to perform parameter optimization, wherein the values of c and g are logarithmic values. Through training, the best c value (bestc), the best g value (bestg) and the minimum mean standard error (MSE) value of the SVM network parameters are obtained, which are as follows: bestc=90.5097; bestg=0.03125; and the minimum MSE=214.061.

[0110] In the embodiment, the SVM temperature measurement grid model is established, the large sample database is used to train the temperature measurement grid model, the input node number of the SVM network is selected as 3 and the output node number is selected as 1. A three-layer SVM network is constructed by using the characteristics of the SVM network, wherein the input layer is three items, that is, the R, G and B three color values, and the output layer is one item, that is, the two-dimensional temperature value T of the flame radiation energy picture.

[0111] As another preferred embodiment, a multi-dimensional three-layer SVM network can also be constructed, wherein each dimension uses an SVM network as a classifier, at this time the model is a hybrid model combining the characteristics of deep learning and SVM, so that the corresponding detection temperature value can be more accurately output through the temperature grid model.

[0112] As a preferred scheme of the embodiment, sample data is acquired, specifically including:

[0113] The black furnace flame image in the preset black furnace and the flame radiation temperature of each group are acquired; color recognition is performed on the black furnace flame image based on a true color image card to acquire the gray quantization values corresponding to the red, green and blue three colors of each pixel point coordinate in the black furnace flame image; the correction coefficients of any two components in the three colors are determined based on the monochromatic emissivity of the preset black furnace, the flame radiation temperature and the wavelengths corresponding to the three colors of light; the monochromatic radiation intensity of the burning flame spectrum of each color is determined according to the correction coefficients and the wavelengths of light corresponding to the three colors in combination with the monochromatic emissivity of the flame, so that the light intensity signal value under the wavelength corresponding to each color of light is obtained; the relationship between the corresponding temperature and the monochromatic emissivity of the preset black furnace and the wavelength is obtained based on the light intensity signal value, and the gray quantization values corresponding to the three colors of the black furnace flame image are calibrated according to the relationship and the correction coefficients, so that the three primary color values corresponding to each flame radiation temperature are obtained.

[0114] In the embodiment, the black furnace flame image is collected and stored by means of the black body furnace by using the temperature control system, so that a plurality of groups of sample data are obtained, and preferably 30 groups. Further, color recognition is performed on the black furnace flame image based on a true color image card, so that the gray quantization values corresponding to the red, green and blue three colors of each pixel point coordinate in the black furnace flame image can be recognized, and the correction coefficients of any two components in the three colors are determined based on the monochromatic emissivity of the preset black furnace, the flame radiation temperature and the wavelengths corresponding to the three colors of light:

[0115]

[0116]

[0117] wherein, is the monochromatic emissivity of the artificial black body; ; R, G and B represent the three primary color lights with wavelengths of 700 nm, 546.1 nm and 435.8 nm respectively; T is the flame temperature of the black body furnace, and the unit is K.

[0118] According to the Wien formula:

[0119]

[0120] wherein, is the monochromatic radiation intensity [W / m2 / nm] of the flame; 3 ]; is the monochromatic emissivity of the flame; is the wavelength (m) and T is the absolute temperature (K); is the Planck constant, respectively; .

[0121] Take the light intensity signal value at 700 nm, 546.1 nm, 435.8 nm three different wavelengths , take the ratio of two light intensity values at two different wavelengths (preferably, take the ratio of light intensity signal values between R and G and between G and B), then:

[0122]

[0123]

[0124] Thus, the two formulas can be divided to obtain:

[0125] Further, the relationship between the flame radiation temperature of each group and the gray value of the radiation thermal image output by the color CCD camera can be obtained, and then the system is calibrated, and the ideal fitting curve is obtained by a suitable fitting method, and the calibrated expression is:

[0126]

[0127] wherein, , are the correction coefficients for components G and B, respectively; R, G, and B are the three primary color values corresponding to a pixel point on the flame picture obtained by the CCD camera, and R', G', and B' are the calibrated results of R, G, and B, respectively.

[0128] As a preferred scheme of the embodiment, according to the boiler combustion image, the three primary color values corresponding to an arbitrary pixel point are determined, and the method further comprises the following steps:

[0129] Based on the true color image card, color recognition is performed on the boiler combustion image to obtain the gray quantization values corresponding to the red, green, and blue colors of each pixel point coordinate in the boiler combustion image; the gray quantization values corresponding to the three colors of the boiler combustion image are calibrated according to the relationship and the correction coefficient, so as to determine the three primary color values corresponding to an arbitrary pixel point.

[0130] In the embodiment, since there is a corresponding calibration expression when the temperature measurement grid model is established, when the gray quantization values corresponding to the red, green and blue of each pixel point coordinate in the specific boiler combustion image are obtained, the corresponding system calibration can also be performed, that is, the gray quantization values corresponding to R, G and B of the boiler combustion image are calibrated according to the corresponding relationship and the correction coefficient, so that the three primary color values after final calibration can be input into the temperature measurement grid model, thereby greatly improving the accuracy of obtaining image data for temperature measurement detection, and having higher precision than direct single model detection.

[0131] The above embodiment has the following effects:

[0132] The technical scheme of the present application acquires detection data in the boiler in real time, digitizes the corresponding detection data, obtains a corresponding boiler combustion image, and determines the three primary color values of any point in the image, so that each three primary color value can be input into the temperature measurement grid model in turn, and the temperature value corresponding to each pixel point can be output, and the temperature field distribution of the boiler can be generated based on the temperature value, so that the distribution of the flame temperature field in the current boiler furnace can be accurately and quickly obtained, and the temperature field state in the boiler can also be detected comprehensively and accurately in real time, avoiding the problems that the existing equipment can only measure for a short time and cannot accurately reflect the combustion state in the furnace, so that the temperature distribution of the entire furnace cannot be depicted.

[0133] Embodiment three

[0134] Please refer to Figure 2 The present application provides a kind of based on image recognition's boiler temperature field real-time monitoring method device, comprising: acquisition module 201, three primary color module 202 and temperature module 203;

[0135] Acquisition module 201, for real-time acquisition of detection data in the boiler, and detection data is digitized, and the boiler combustion image is obtained;

[0136] Three primary color module 202, for determining the three primary color value corresponding to any one pixel point according to the boiler combustion image;

[0137] Temperature module 203, for inputting each three primary color value into temperature measurement grid model in turn, outputting the temperature value corresponding to each pixel point, and generating the temperature field distribution of the boiler based on the temperature value.

[0138] As a preferred scheme, real-time acquisition of detection data in the boiler, and detection data is digitized, and the boiler combustion image is obtained, specifically comprising:

[0139] The detection signal of the one-dimensional time sequence is digitized to discretize the detection signal of the one-dimensional time sequence into a detection data signal.

[0140] The detection signal of the one-dimensional time sequence is digitized to discretize the detection signal of the one-dimensional time sequence into a detection data signal.

[0141] According to the detection digital signal, an image is generated to obtain a boiler combustion image.

[0142] As a preferred solution, the detection signal of the one-dimensional time sequence is digitized to discretize the detection signal of the one-dimensional time sequence into a detection data signal, specifically including:

[0143] The detection signal of the one-dimensional time sequence is converted into a frequency domain signal, and the converted frequency domain signal is extracted to obtain corresponding amplitude features.

[0144] The corresponding continuous amplitude features are binned to obtain a detection data signal related to the amplitude size.

[0145] The detection signal of the one-dimensional time sequence is extracted to obtain a cutoff frequency and an original distribution, and the extracted cutoff frequency and original distribution are converted to obtain a detection data signal related to spatial coordinates.

[0146] The detection data signal includes signals related to the amplitude size and spatial coordinates.

[0147] As a preferred solution, according to the boiler combustion image, the three primary color values corresponding to any one pixel point are determined, specifically including:

[0148] Based on the true color image card, the color of the boiler combustion image is identified to obtain the gray quantization value corresponding to the red, green and blue three colors of each pixel point coordinate in the boiler combustion image as the three primary color values corresponding to any one pixel point.

[0149] As a preferred solution, each three primary color value is sequentially input into a temperature measurement grid model to output the temperature value corresponding to each pixel point, and a temperature field distribution of the boiler is generated based on the temperature value, specifically including:

[0150] Each three primary color value is input into the temperature measurement grid model as an input to the temperature measurement grid model to calculate the input three primary color value and output a two-dimensional temperature value corresponding to each pixel point in the boiler combustion image.

[0151] Based on the two-dimensional temperature value and the corresponding spatial coordinates, a temperature field distribution of the boiler is generated.

[0152] As a preferred solution, each three primary color value is sequentially input into the temperature measurement grid model, and the temperature value corresponding to each pixel point is output, and the temperature field distribution of the boiler is generated based on the temperature value, and specifically further comprises:

[0153] The pixel points in the boiler combustion image are regionally divided, and the middle pixel points of each region are determined as target pixel points according to the regions of the divided pixel points; wherein the number and size of the pixel points in each region are equal;

[0154] The three primary color values corresponding to the target pixel points are respectively input into the temperature measurement grid model, so that the temperature measurement grid model calculates the input three primary color values, and outputs the two-dimensional temperature value corresponding to each target pixel point in the boiler combustion image;

[0155] Based on the region where each target pixel point is located, edge smoothing is performed on the target pixel points of the adjacent regions, so that the two-dimensional temperature values corresponding to the remaining pixel points in each region are calculated respectively;

[0156] The two-dimensional temperature values of all pixel points and their corresponding spatial coordinates are used to generate the temperature field distribution of the boiler.

[0157] As a preferred solution, the method for constructing the temperature measurement grid model comprises:

[0158] Obtain sample data and normalize the sample data; wherein the sample data includes: a plurality of groups of flame radiation temperatures and three primary color values corresponding to each flame radiation temperature respectively;

[0159] Set an initial temperature measurement network model and determine the penalty factor, Gaussian kernel function and kernel parameter of the initial temperature measurement network model;

[0160] Train the initial temperature measurement network model based on the sample data, so as to obtain the best penalty factor, the best kernel parameter and the minimum mean standard error corresponding to the initial temperature measurement network model;

[0161] Update the best penalty factor, the best kernel parameter and the minimum mean standard error in the initial temperature measurement network model, so as to train the final temperature measurement grid model.

[0162] As a preferred solution, the sample data is obtained, specifically comprising:

[0163] Obtain the black furnace flame image and each group of flame radiation temperature in the preset black furnace;

[0164] Based on the true color image card, the color of the black furnace flame image is identified, and the gray quantization values corresponding to the red, green and blue of each pixel point coordinate in the black furnace flame image are obtained;

[0165] Determine the correction coefficient of any two components in the three colors based on the preset black furnace monochromatic emissivity, the flame radiation temperature and the wavelength corresponding to the three colors respectively;

[0166] Determine the monochromatic radiation intensity of the combustion flame spectrum of each color based on the correction coefficient, the light wavelength corresponding to the three colors respectively and the monochromatic emissivity of the flame, so as to obtain the light intensity signal value under the wavelength corresponding to each color light;

[0167] Based on the light intensity signal value, obtain the relationship between the corresponding temperature and the light wavelength and the monochromatic emissivity of the preset black furnace, and calibrate the gray quantization value corresponding to the three colors of the black furnace flame image respectively according to the relationship and the correction coefficient, so as to obtain the three primary color values corresponding to each flame radiation temperature.

[0168] As a preferred scheme, the three primary color values corresponding to any one pixel point are determined based on the boiler combustion image, and the specific process further comprises:

[0169] Based on the true color image card, color recognition is performed on the boiler combustion image to obtain the gray quantization value corresponding to the red, green and blue three colors respectively of each pixel point coordinate in the boiler combustion image;

[0170] The gray quantization value corresponding to the three colors of the boiler combustion image is calibrated according to the relationship and the correction coefficient, so as to determine the three primary color values corresponding to any one pixel point.

[0171] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described device can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0172] The above embodiments have the following effects:

[0173] The technical scheme of the present application obtains the detection data in the boiler, thereby digitally processing the corresponding detection data to obtain the corresponding boiler combustion image, and further determines the three primary color values of any point in the image, so that each three primary color value can be input into the temperature measurement grid model in turn, and the temperature value corresponding to each pixel point can be output, and the temperature field distribution of the boiler can be generated based on the temperature value, so that the distribution of the flame temperature field in the current boiler furnace can be accurately and quickly obtained, and the temperature field state in the boiler can also be detected comprehensively and accurately in a timely manner, thereby avoiding the problems that the existing equipment can only be measured for a short time, and the combustion state in the furnace cannot be accurately reflected, so that the temperature distribution of the entire furnace cannot be depicted.

[0174] Embodiment four

[0175] Please refer to Figure 3The application also provides a boiler temperature field real-time monitoring method system based on image recognition, which comprises an optical system, an air cooling system, a combustion state acquisition system and an image processing system.

[0176] The optical system comprises a shell 7, an image transmission fiber 10 penetrating a front-end cavity of the shell 7 and a mirror rod 17; wherein the shell 7 comprises the front-end cavity and a rear-end cavity, and a front end of the front-end cavity is provided with a multi-hole plate air outlet.

[0177] As a preferred scheme of the embodiment, please refer to Figure 4 The first end of the image transmission fiber 10 is connected with a collimating lens 13, the second end of the image transmission fiber 10 is connected with a color filter 6, and the image transmission fiber 10 is wrapped with a high-temperature-resistant protective material 12; the front end of the mirror rod 17 is provided with a lens 15, and a plurality of optical lenses 16 and a double-band filter 21 are arranged in the mirror rod 17.

[0178] In the embodiment, the number of the optical lenses 16 can be set according to actual needs, and preferably, the number of the optical lenses 16 is greater than or equal to 3.

[0179] As a preferred scheme of the embodiment, a plurality of multi-hole support frames 9 are arranged in the front-end cavity of the shell 7 along the direction of the first end and the second end of the front-end cavity, and the multi-hole support frames 9 are used to support the image transmission fiber 10 and the mirror rod 17.

[0180] As a preferred scheme of the embodiment, a fireproof plate 11 is further arranged outside the front-end cavity of the shell 7, and a hand-held handle 20 is further arranged at the position of the second end outside the front-end cavity of the shell 7.

[0181] In the embodiment, the fireproof plate can prevent the high-temperature gas or coal ash sprayed in the fire observation hole from burning the operator.

[0182] The air cooling system comprises a cooler 18 and a cooling air inlet 19 and a cooling air outlet 8 connected with the optical system respectively.

[0183] The combustion state acquisition module is connected with the second end of the mirror rod 17, and comprises a camera arranged in the rear-end cavity of the shell 7, the color filter 6 arranged in front of the camera, a microcomputer 22 arranged in the rear-end cavity of the shell 7, an image acquisition card 23 connected with the microcomputer 22 and arranged on the wall of the rear-end cavity of the shell 7, and the power supply 4 arranged in the rear-end cavity of the shell 7.

[0184] As a preferred scheme of the embodiment, the power supply 4 is connected to the camera and the microcomputer 22 through the power switch 2.

[0185] Optionally, the camera can be a CCD camera, which can directly acquire the image of the flame 14, please refer to Figure 5 which is a spectral response characteristic curve diagram of the double-band filter and the CCD camera.

[0186] The image processing system 3 for executing the furnace image combustion temperature field detection method according to any one of the above embodiments comprises an image acquisition card 23 and an image processing unit; the image acquisition card 23 is used for signal conditioning and analog-digital signal conversion; and the image processing unit is used for image denoising, image enhancement processing and temperature field calculation.

[0187] In the embodiment, the image processing system 3 comprises the image acquisition card 23 and the image processing unit; and the image processing unit comprises functions of image denoising, image enhancement processing and temperature field calculation.

[0188] In the embodiment, the image processing system 3 is connected with the image acquisition card 23 through the first interface 1.

[0189] In the embodiment, in the image processing system, the CCD camera 5 has a 24-bit true color image card, and a color CCD three-color temperature measurement method is adopted, in which R, G and B represent the gray quantization values of red, green and blue three colors at each coordinate, and then system calibration is performed to establish the relationship between the flame radiation temperature and the gray value of the radiation heat image output by the color CCD camera 5, and finally the R, G and B three-color values are obtained.

[0190] In the embodiment, in the image processing system, the R, G and B values obtained by the image processing system are brought into the established multi-layer perception neural network to obtain the temperature field distribution of the power station boiler furnace cross section.

[0191] Optionally, the first interface 1 is a gigabit network interface, and the gigabit network interface can be connected to a computer.

[0192] Optionally, the image transmission fiber 10 is an armored fiber, the lens 15 is a pinhole lens, and the power supply 4 is a lithium battery.

[0193] In the embodiment, the compressed air introduced into the boiler can enter the device through the cooling air inlet 19, and then the cooling air is divided into two paths; one path flows out through the cooling air outlet 8, and the other path flows out through the multi-hole plate air outlet 2. The multi-hole plate air outlet 2 can cool the lens 10 and the collimating lens 6, and the cooling air entering the device can also blow and protect the lens 10 and the collimating lens 6 to keep them clean. The collimating lens 6 only allows parallel light in front of it to enter, and the light incidence diagram is as shown in Figure 4The radiation of the coal powder combustion flame in the boiler furnace enters the collimating lens 6 and then enters the lens 15 along the image transmission fiber 10, and then enters the double band-pass filter 21 after passing through the lens group composed of multiple optical lenses 16, and the image on the double band-pass filter 21 is captured by the CCD camera 5. The color CCD camera decomposes the incident light into red (R), green (G) and blue (B) three-color images according to different wavelengths, and the wavelengths are 700 nm, 546.1 nm and 435.8 nm respectively. The porous structure of the multi-hole support frame 17 can also be cooled by the cooling air. The control command and data transmission of the CCD camera 5 are both carried out through the gigabit network cable, and the gigabit network interface is connected with the subsequent image processing system 3, which can simultaneously play the role of power supply and data transmission control. The CCD camera 5 is powered by the built-in lithium battery, and the power supply can be controlled to be turned on or off through the power switch 2.

[0194] It can be understood that the existing detection method can also be reconstructed by using the sound wave method or the laser absorption spectrum method, but both methods have certain technical disadvantages. The sound wave method needs to install a certain number of transmitters and receivers on the furnace, which increases the construction difficulty and hardware cost, and factors such as field vibration will affect the stability of the system operation, and the sound wave transmitter is large in size, and the furnace wall opening is also large, which will also affect the temperature distribution in the furnace to some extent. The cost of the laser absorption spectrum method is relatively high, and the requirements for on-site operation are also relatively high, which is not convenient for the use of power plant on-site staff. Therefore, the embodiment of the present application has the advantages of simple structure, low cost, high stability and convenient operation, and the device and the detection method can realize the wireless image recognition-based real-time monitoring method of the boiler temperature field, which can accurately and comprehensively detect the temperature state in the boiler.

[0195] The above embodiment has the following effects:

[0196] The optical system of the embodiment comprises a shell, an image transmission fiber and a mirror rod penetrating a front end cavity, a collimating lens connected with a first end of the image transmission fiber, a lens arranged at a front end of the mirror rod, a plurality of optical lenses arranged in the mirror rod and spaced apart from each other from a first end to a second end of the mirror rod, and a double band-pass filter; the air cooling system comprises an air cooler, a cooling air inlet connected with the optical device, and a cooling air outlet; the combustion state acquisition system comprises a CCD camera connected with the second end of the mirror rod and arranged in a rear end cavity, a color filter placed in front of the CCD camera, a microcomputer arranged in the rear end cavity, a first interface connected with the microcomputer and used for transmitting spectral data and power supply, an image acquisition card arranged on a wall of the rear end cavity, and a power supply arranged in the rear end cavity and connected with the CCD camera and the microcomputer; and the image processing system comprises the image acquisition card and an image processing unit.

[0197] Embodiment five

[0198] Correspondingly, the application further provides a terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the image recognition based real-time boiler temperature field monitoring method according to any one of the above embodiments when executing the computer program.

[0199] The terminal device of the embodiment comprises a processor, a memory, and a computer program stored in the memory and executable on the processor, and computer instructions. The processor implements the steps in the above embodiment one when executing the computer program, for example, the steps S101 to S103 shown in the figure. Figure 1 Or, the processor implements the functions of the modules / units in the above device embodiment when executing the computer program, for example, the temperature module 203.

[0200] The computer program can be divided into one or more modules / units for example, which are stored in the memory and executed by the processor to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program in the terminal device. For example, the temperature module 203 is configured to input each of the three primary color values into a temperature grid model in sequence, output a temperature value corresponding to each pixel point, and generate a temperature field distribution of the boiler based on the temperature value.

[0201] The terminal device can be a desktop computer, a notebook computer, a palm computer, a cloud server and the like. The terminal device can include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the schematic diagram is only an example of the terminal device, and does not constitute a limitation on the terminal device, and can include more or fewer components than the diagram, or combine certain components, or different components, for example, the terminal device can also include an input / output device, a network access device, a bus and the like.

[0202] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and connects all parts of the terminal device through various interfaces and lines.

[0203] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the terminal device by running or executing the computer program and / or modules stored in the memory, and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required by a function, etc.; and the data storage area can store data created according to the use of the mobile terminal, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0204] The modules / units integrated in the terminal device can be stored in a computer readable storage medium if they are realized in the form of software function units and sold or used as independent products. Based on this understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. The computer program can implement the steps of each method embodiment when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the contents included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0205] Embodiment six

[0206] Correspondingly, the application further provides a computer readable storage medium, including a stored computer program, wherein when the computer program runs, the device where the computer readable storage medium is located executes the image recognition based boiler temperature field real-time monitoring method as any one of the above embodiments.

[0207] The above-described specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above-described specific embodiments are merely examples of the present application and are not intended to limit the protection scope of the present application. It is particularly pointed out that any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for real-time monitoring of boiler temperature field based on image recognition, characterized in that, include: Acquire detection data from the boiler and digitize the data to obtain boiler combustion images; Based on the boiler combustion image, determine the three primary color values ​​corresponding to any pixel. Each of the three primary color values ​​is sequentially input into the temperature measurement grid model, and the temperature value corresponding to each pixel is output. The temperature field distribution of the boiler is then generated based on the temperature value. The method for constructing the temperature measurement grid model includes: Acquire sample data and normalize the sample data; wherein, the sample data includes: several sets of flame radiation temperatures and the three primary color values ​​corresponding to each flame radiation temperature; Set up an initial temperature measurement network model and determine the penalty factor, Gaussian kernel function, and kernel parameters of the initial temperature measurement network model; The initial temperature measurement network model is trained based on the sample data to obtain the optimal penalty factor, optimal kernel parameter, and minimum average standard error of the corresponding initial temperature measurement network model. The optimal penalty factor and the optimal kernel parameter kernel minimum average standard error are updated and set in the initial temperature measurement network model, thereby training the final temperature measurement grid model. The acquisition of sample data specifically includes: Acquire images of the flames in a preset black furnace and the radiation temperatures of each group of flames; Based on a true-color image card, color recognition is performed on the black furnace flame image to obtain the gray quantization values ​​corresponding to the red, green and blue colors of each pixel coordinate in the black furnace flame image. Based on the monochromatic emissivity, flame radiation temperature, and wavelengths corresponding to the three colors of light of the preset black furnace, the correction coefficients for any two components of the three colors are determined. Based on the correction coefficients and the wavelengths of the three colors, combined with the flame monochromatic emissivity, the monochromatic radiation intensity of the combustion flame spectrum for each color is determined, thereby obtaining the light intensity signal value at the wavelength corresponding to each color of light. Based on the light intensity signal value, the relationship between the corresponding temperature and the light wavelength and the monochromatic emissivity of the preset black furnace is obtained. According to the relationship and the correction coefficient, the gray quantization values ​​corresponding to the three colors of the black furnace flame image are calibrated to obtain the three primary color values ​​corresponding to the radiation temperature of each flame. The step of determining the three primary color values ​​corresponding to any pixel based on the boiler combustion image specifically includes: Based on a true-color image card, color recognition is performed on the boiler combustion image to obtain the gray quantization values ​​corresponding to the red, green, and blue colors of each pixel coordinate in the boiler combustion image. Based on the relationship and the correction coefficient, the gray quantization values ​​corresponding to the three colors of the boiler combustion image are calibrated to determine the three primary color values ​​corresponding to any pixel.

2. The method for real-time monitoring of boiler temperature field based on image recognition as described in claim 1, characterized in that, The process of acquiring detection data from the boiler and digitizing the detection data to obtain a boiler combustion image specifically includes: By adjusting the integration time using a lens installed in the boiler observation hole and according to the preset sampling frequency, a one-dimensional time series detection signal is obtained. The detection signal of the one-dimensional time series is digitally processed to discretize the detection signal of the one-dimensional time series into detection data signals; An image is generated based on the detected data signal to obtain a boiler combustion image.

3. The method for real-time monitoring of boiler temperature field based on image recognition as described in claim 2, characterized in that, The digital processing of the detection signal of the one-dimensional time series, so as to discretize the detection signal of the one-dimensional time series into detection data signals, specifically includes: The detection signal of the one-dimensional time series is converted into a frequency domain signal, and the converted frequency domain signal is used to extract features to obtain the corresponding amplitude features; The corresponding continuous amplitude features are binned to obtain the detection data signal involving amplitude magnitude; The cutoff frequency and original distribution of the detection signal of the one-dimensional time series are extracted, and the corresponding detection data signal involving spatial coordinates is obtained based on the extracted cutoff frequency and original distribution. The detection data signal includes signals involving amplitude and spatial coordinates.

4. A method for real-time monitoring of boiler temperature field based on image recognition as described in any one of claims 1-3, characterized in that, The step of determining the three primary color values ​​corresponding to any pixel based on the boiler combustion image specifically includes: Based on a true-color image card, color recognition is performed on the boiler combustion image to obtain the grayscale values ​​corresponding to the red, green, and blue colors of each pixel coordinate in the boiler combustion image, which are used as the three primary color values ​​corresponding to any pixel.

5. The method for real-time monitoring of boiler temperature field based on image recognition as described in claim 1, characterized in that, The process of sequentially inputting each of the three primary color values ​​into the temperature measurement grid model, outputting the temperature value corresponding to each pixel, and generating the temperature field distribution of the boiler based on the temperature values ​​specifically includes: Each of the three primary color values ​​is used as input to the temperature measurement grid model, so that the temperature measurement grid model calculates the input three primary color values ​​and outputs the two-dimensional temperature value corresponding to each pixel in the boiler combustion image. The temperature field distribution of the boiler is generated based on the two-dimensional temperature values ​​and their corresponding spatial coordinates.

6. The method for real-time monitoring of boiler temperature field based on image recognition as described in claim 1, characterized in that, The step of sequentially inputting each of the three primary color values ​​into the temperature measurement grid model, outputting the temperature value corresponding to each pixel, and generating the temperature field distribution of the boiler based on the temperature values, specifically also includes: The pixels in the boiler combustion image are divided into regions, and the middle pixel of each region is determined as the target pixel based on the divided regions; wherein the number and size of pixels in each region are equal. The three primary color values ​​corresponding to the target pixels are used as inputs to the temperature measurement grid model, so that the temperature measurement grid model can calculate the input three primary color values ​​and output the two-dimensional temperature value corresponding to each target pixel in the boiler combustion image. Based on the region where each target pixel is located, edge smoothing is performed on the target pixels in the adjacent regions, thereby calculating the two-dimensional temperature value corresponding to the remaining pixels in each region. The temperature field distribution of the boiler is generated by using the two-dimensional temperature values ​​of all pixels and their corresponding spatial coordinates.

7. A real-time boiler temperature field monitoring device based on image recognition, applied to the real-time boiler temperature field monitoring method based on image recognition as described in claim 1, characterized in that, include: Acquisition module, three primary color module, and temperature module; The acquisition module is used to acquire detection data in the boiler and digitize the detection data to obtain a boiler combustion image; The three-primary-color module is used to determine the three-primary-color value corresponding to any pixel based on the boiler combustion image. The temperature module is used to input each of the three primary color values ​​sequentially into the temperature measurement grid model, output the temperature value corresponding to each pixel, and generate the temperature field distribution of the boiler based on the temperature value.

8. A real-time monitoring system for boiler temperature field based on image recognition, characterized in that, include: Optical system, air-cooling system, combustion status acquisition system, and image processing system; The optical system includes: a housing, an image transmission optical fiber penetrating the front cavity of the housing, and a lens rod; wherein, the housing includes a front cavity and a rear cavity, and the front end of the front cavity is provided with a perforated plate air outlet; The air-cooling system includes: a cooler and a cooling air inlet and a cooling air outlet respectively connected to the optical system; The combustion state acquisition module is connected to the second end of the lens rod and includes: a camera disposed in the rear cavity of the housing, a color filter disposed in front of the camera, a microcomputer disposed in the rear cavity of the housing, an image acquisition card connected to the microcomputer and disposed on the rear cavity wall of the housing, and a power supply disposed in the rear cavity of the housing. The image processing system is used to execute the real-time monitoring method for boiler temperature field based on image recognition as described in any one of claims 1-6, comprising: an image acquisition card and an image processing unit; the image acquisition card is used for signal conditioning and analog-to-digital signal conversion; the image processing unit is used for image denoising, image enhancement processing, and temperature field calculation.

9. The real-time boiler temperature field monitoring system based on image recognition as described in claim 8, characterized in that, The first end of the image transmission fiber is connected to a collimating lens, the second end of the image transmission fiber is connected to a color filter, and the image transmission fiber is wrapped with a high-temperature resistant protective material. A lens is provided at the front end of the lens rod, and multiple optical lenses and a dual bandpass filter are arranged at intervals inside the lens rod.

10. The real-time boiler temperature field monitoring system based on image recognition as described in claim 8, characterized in that, The power supply is connected to the camera and the microcomputer via a power switch.

11. A real-time boiler temperature field monitoring system based on image recognition as described in any one of claims 8-10, characterized in that, Multiple porous support frames are provided in the front cavity of the housing along the direction of the first end and the second end of the front cavity. The porous support frames are used to support the image transmission optical fiber and the mirror rod.

12. A real-time boiler temperature field monitoring system based on image recognition as described in any one of claims 8-10, characterized in that, A fire-resistant plate is provided on the outer side of the front cavity of the housing, and a hand handle is provided at the second end of the outer side of the front cavity of the housing.

13. A terminal device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the image recognition-based real-time monitoring method for boiler temperature field as described in any one of claims 1 to 6.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the image recognition-based real-time monitoring method for boiler temperature field as described in any one of claims 1 to 6.

Images