A high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging.

Abstract

This invention discloses a high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging. The method includes: in plume velocity inversion measurement using the improved Farneback dense optical flow method, a clustering algorithm is used to extract plume features from the image and screen the main plume components; image texture intensity is calculated based on the main plume components, and the corresponding optical flow parameters are determined; forward optical flow is calculated based on the optical flow parameters; a consistency error field based on bidirectional optical flow is introduced, and time consistency correction is performed on the forward optical flow based on the consistency error field; the fusion weight factor is determined based on the gradient of the calculated forward optical flow; and the velocity is calculated after weighted total variation denoising of the forward optical flow to obtain the plume velocity inversion result; the plume velocity inversion result is verified; and the plume velocity inversion result is combined with hyperspectral remote sensing imaging to achieve high-precision emission flux quantification, thus improving the accuracy of emission flux calculation for organized industrial emissions.

Description

Technical Field

[0001] This invention belongs to the field of plume velocity inversion method and corresponding accuracy verification technology, specifically involving a high-precision emission flux quantification method for plume velocity inversion, verification and fusion imaging. Background Technology

[0002] Currently, industrial emissions remain one of the main sources of air pollutants. Accurately calculating the emission fluxes of organized emissions from factories and establishing dynamic emission inventories, especially the emission fluxes at chimney outlets, is of great significance for air pollution control. In this process, obtaining information on plume concentration and plume velocity is crucial for calculating chimney outlet emission fluxes.

[0003] Currently, methods for calculating plume velocity information can be mainly divided into: wind speed data approximation, traditional optical flow, and deep learning-based modeling methods. Wind speed data approximation approximates the plume velocity using wind speed data, but ignores the plume's own upward velocity. Furthermore, methods based on in-situ measurements and model simulations suffer from systematic errors due to matching issues in the observation space. Traditional optical flow suffers from plume misidentification, poor temporal consistency, and difficulty in parameter determination. While deep learning-based modeling methods have certain advantages, they are prone to prediction bias when facing scenarios not covered by the training set (such as new emission sources or extreme weather conditions). In addition, deep learning methods have poor interpretability, and training sets in real-world scenarios are difficult to obtain.

[0004] Currently, validation of machine vision-based gas velocity calculation methods (such as optical flow methods and deep learning-based modeling methods) typically employs CFD-based model validation or gas distribution device validation. However, a common drawback of these two methods is their over-reliance on the accuracy of the model or gas distribution device itself, lacking observational data independent of the visual method for validation. Machine vision-based gas velocity calculation methods can be used to quantify chimney emission fluxes, but they still require information on gas concentration distribution.

[0005] Currently, the common method for obtaining concentration data in conjunction with gas flow rate calculations is through gas cameras. However, using gas cameras to obtain concentration distributions presents several challenges. First, due to limitations in filter performance, gas cameras can only acquire very limited information on gas components, typically measuring only components such as sulfur dioxide and nitrogen dioxide. Second, gas cameras cannot effectively eliminate cross-absorption interference between multiple gas components. Furthermore, due to the nonlinearity of the calibration factor, gas cameras may introduce conversion errors, further affecting the accuracy of the concentration data.

[0006] Therefore, how to obtain a plume velocity calculation method with high spatial matching degree and good scene adaptability based on actual observation data, and how to design a corresponding verification method to achieve high-precision emission flux calculation, have become urgent technical problems to be solved. Summary of the Invention

[0007] In view of the above, the purpose of this invention is to provide a high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging. This method provides a high-precision, robust plume velocity calculation method and an accuracy verification method that does not rely on training data for the field of environmental monitoring. Furthermore, it combines a ground-based hyperspectral remote sensing system with rapid high-resolution imaging of multi-component atmospheric trace components to achieve high-precision emission flux quantification, thereby further improving the accuracy of emission flux calculation for organized industrial emissions.

[0008] To achieve the above-mentioned objectives, the embodiments provide a high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging, comprising the following steps:

[0009] In the Farneback dense optical flow method for plume velocity inversion measurement, a clustering algorithm is used to extract plume features from the image and screen the plume subject. Based on the plume subject, the image texture intensity is calculated and the optical flow parameters corresponding to the image texture intensity are determined. Based on the optical flow parameters, the forward optical flow is calculated. A consistency error field based on bidirectional optical flow is introduced and the forward optical flow is corrected for temporal consistency based on the consistency error field. The fusion weight factor is determined based on the gradient of the calculated forward optical flow. After weighted total variation denoising of the forward optical flow based on the fusion weight factor, the flow velocity is calculated to obtain the plume velocity inversion result.

[0010] The results of plume velocity inversion were verified, and the results of plume velocity inversion were combined with hyperspectral remote sensing imaging to achieve high-precision emission flux quantification.

[0011] Preferably, a clustering algorithm is used to extract plume features from the image and filter the plume subjects, including:

[0012] Clustering algorithms are used to cluster pixels to extract plume features. The plume is considered to be the area with the highest brightness, and the cluster with the largest channel value is selected as the main body of the plume.

[0013] Preferably, calculating the image texture intensity based on the plume body and determining the optical flow parameters corresponding to the image texture intensity includes:

[0014] The gradient fields in the horizontal and vertical directions of the feather body are calculated, and the gradient magnitude statistics of each pixel are statistically analyzed based on the gradient fields in the two directions. The standard deviation of the gradient magnitude statistics is used to represent the image texture intensity. The optical flow parameters corresponding to the image texture intensity are determined according to the correspondence between the image texture intensity and the optical flow parameters.

[0015] Preferably, the correspondence is as follows:

[0016] When the image texture intensity σ is in [0,5), the texture level is very low texture, corresponding to a pyramid scaling factor of 0.2, a pyramid layer of 2, a window size of 7, and an iteration count of 2.

[0017] When the image texture intensity σ is in [5,10), the texture level is low texture, the corresponding pyramid scaling factor is 0.25, the number of pyramid layers is 3, the window size is 11, and the number of iterations is 3;

[0018] When the image texture intensity σ is in [10,20), the texture level is medium texture, the corresponding pyramid scaling factor is 0.4, the number of pyramid layers is 4, the window size is 15, and the number of iterations is 4.

[0019] When the image texture intensity σ is in [20,30), the texture level is high texture, the corresponding pyramid scaling factor is 0.5, the number of pyramid layers is 5, the window size is 21, and the number of iterations is 5;

[0020] When the image texture intensity σ is in [30,+∞), the texture level is very high texture, corresponding to a pyramid scaling factor of 0.6, a pyramid layer of 6, a window size of 25, and an iteration count of 6.

[0021] Preferably, the forward optical flow is calculated based on optical flow parameters, a consistency error field based on bidirectional optical flow is introduced, and time consistency correction of the forward optical flow is performed based on the consistency error field, including:

[0022] The Farneback method, based on optical flow parameters, is used to calculate the forward optical flow F. f The consistency error field E based on bidirectional optical flow is introduced as follows:

[0023] E = |F f +F b |

[0024] Among them, F b This is the reverse optical flow obtained through reverse calculation.

[0025] Time consistency correction of the forward optical flow based on the consistency error field is expressed as:

[0026] F = F f -λ·E

[0027] Where λ is the correction weighting factor, and F represents the corrected forward optical flow.

[0028] Preferably, the fusion weighting factor is determined based on the gradient of the forward optical flow, and the flow velocity is calculated after weighted total variation denoising based on the fusion weighting factor to obtain the plume flow velocity inversion result, including:

[0029] Calculate the magnitude gradient G of the forward optical flow flow The linear combination gradient method is used based on the magnitude gradient G of the forward optical flow. flow and gradient magnitude statistic G image Obtain the fusion weight factor w:

[0030] w=α·G flow +(1-α)·G image

[0031] Where α represents the weight of the optical flow magnitude gradient;

[0032] Weighted total variation denoising of the forward optical flow is performed based on a fusion weighting factor. Specifically, the optimization objective of weighted total variation denoising is to make the optical flow smoother while maintaining sharp edges. This is achieved by constructing an energy function based on the weighting factor and minimizing the energy functional E(F):

[0033] E(F)=∫w·G flow (F)dxdy

[0034] G flow (F) represents the gradient of the forward optical flow. After performing variational operations on the energy functional E(F), we obtain the divergence operator. The divergence is a reflection of the energy gradient and serves as a tool for optimizing the energy functional. The discretized gradient divergence is calculated as follows:

[0035]

[0036] ∈ is a minimal quantity. It is the weight update term, (i,j) represents the coordinate position, ΔF x (i,j) and ΔF y (i,j) represents the change in gradient divergence in the horizontal x-direction and the vertical y-direction. Gradient descent optimizes the energy functional through divergence. The formula for iteratively updating the optical flow field by gradient descent is as follows:

[0037]

[0038] η is the learning rate of gradient descent, the update direction is the negative energy gradient, and k represents the update round;

[0039] The forward optical flow after iterative optimization is denoted as F′, which is the gradient divergence F including the horizontal x-direction and the vertical y-direction. x ′ (i,j) and F y ′ (i,j), then we have:

[0040]

[0041] m(i,j) refers to the magnitude of the displacement vector of the (i,j)th pixel in a frame, i.e., the pixel frame velocity. The magnitude of the actual velocity v(i,j) of each pixel is:

[0042] v(i,j)=m(i,j)×pm×FPS

[0043] The direction θ(x,y) of the true velocity is:

[0044] θ(x,y)=tan -1 (F x ′ (i,j),F y ′ (i,j))

[0045] Where pm is the conversion coefficient between pixels and their actual physical relationship, and FPS refers to the video frame rate.

[0046] Preferably, the results of the plume velocity inversion are verified, including:

[0047] Based on the plume simulation case based on the particle tracking model provided in COMSOL software, plume velocity data is exported as model data. Low velocity is filtered out from the model data and the average value of velocity data at different heights in the same frame is used as the ground truth data.

[0048] After filtering out low velocities and averaging the velocities at different heights in the same frame, the plume velocity inversion data were compared with the ground truth data for correlation verification.

[0049] Preferably, the results of the plume velocity inversion are verified, including:

[0050] The field verification of plume velocity inversion results was achieved using a pulsed smoke generator in conjunction with an anemometer, specifically including:

[0051] An anemometer was used to collect anemometer data using a pulsed smoke generator. At the same time, the plume velocity inversion results were matched with the pulse interval and time resolution of the pulsed smoke generator. The error and variance between the matched plume velocity inversion results and the anemometer data were calculated to verify the results.

[0052] Preferably, high-precision emission flux quantification is achieved by combining plume velocity inversion results with hyperspectral remote sensing imaging, including:

[0053] Before observation, it is ensured that the acquisition angle of the image acquisition camera in the improved dense optical flow method Farneback is consistent with the observation angle of the hyperspectral remote sensing system. Spatial resolution matching is achieved by setting the camera's spatial resolution to be higher than that of the ground-based hyperspectral remote sensing system. Temporal resolution matching is also achieved by integrating the acquired images with the integration time of the ground-based hyperspectral remote sensing system. Based on this, the plume velocity inversion results are combined with hyperspectral remote sensing imaging to achieve high-precision emission flux quantification.

[0054] Preferably, high-precision emission flux quantification is achieved by combining plume velocity inversion results with hyperspectral remote sensing imaging, including:

[0055]

[0056] Where M refers to the molecular mass of the component for which the flux is to be calculated, and N... A It is Avogadro's constant, SCD i,j It is the column concentration value in the j-th column and i-th row, v i,j It is the velocity value in the j-th column and i-th row, N refers to the number of pixels in the j-th column containing the plume, h i,j This refers to the actual height represented by the height of each pixel in the results of a ground-based hyperspectral remote sensing system.

[0057] Compared with the prior art, the beneficial effects of the present invention include at least the following:

[0058] This invention improves the Farneback dense optical flow method. Based on the improved Farneback, plume velocity inversion measurement can be performed, which can improve the accuracy and robustness of plume velocity inversion and does not depend on training data. On this basis, a verification scheme is proposed to verify the effect of Farneback in plume velocity inversion measurement. Furthermore, the plume velocity inversion results are combined with hyperspectral remote sensing imaging to achieve high-precision emission flux quantification, further improving the accuracy of emission flux calculation for organized industrial emissions. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a flowchart of the high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging provided in the embodiments;

[0061] Figure 2This is a flowchart of the improved dense optical flow method Farneback provided in the embodiments;

[0062] Figure 3 This is a schematic diagram of the plume velocity inversion results of the improved dense optical flow method Farneback provided in the embodiment;

[0063] Figure 4 This is the verification effect of the original optical flow method and model data provided in the embodiment;

[0064] Figure 5 This is the verification effect of the improved dense optical flow method Farneback and model data provided in the embodiment. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in 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 invention and do not limit the scope of protection of this invention.

[0066] The inventive concept of this invention is as follows: This invention provides a high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging, which includes three improvements: first, using the improved Farneback dense optical flow method for plume velocity inversion measurement; second, proposing a verification method for the plume velocity inversion measurement results of the improved Farneback dense optical flow method; and third, achieving high-precision emission flux quantification by combining a ground-based hyperspectral remote sensing system for rapid high-resolution imaging of multi-component atmospheric trace components with the plume velocity inversion measurement results.

[0067] like Figure 1 As shown in the embodiment, a high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging includes the following steps:

[0068] S1. The improved dense optical flow method Farneback was used to perform plume velocity inversion measurement to obtain the plume velocity inversion results.

[0069] In the embodiments, improvements to the dense optical flow method Farneback include: extracting plume features using a clustering algorithm, adaptive parameter determination based on image texture, temporal consistency correction, and weighted total variation denoising based on fusion weight factors. Figure 2 As shown, the improved dense optical flow method Farneback is used to perform plume velocity inversion measurement, which includes the following steps:

[0070] S1-1, a clustering algorithm is used to extract plume features from the image and filter the main plume body.

[0071] In plume velocity measurement based on Farneback optical flow, traditional methods directly calculate the dense optical flow field of the entire image. However, due to the complex interaction between the plume and the background (such as turbulent diffusion, illumination changes, and background interference), it leads to problems such as severe noise interference, blurred target area, and high computational redundancy. To solve this problem, this invention uses cluster analysis to analyze the spatial distribution and motion consistency of optical flow vectors, which can separate plume-related vectors (clusters) from background noise (outliers), significantly improving the signal-to-noise ratio.

[0072] First, the clustering degree is represented by the sum of squared errors, and the optimal number of clusters k is determined:

[0073]

[0074] Among them, C i It is the i-th cluster, x j Let μ represent the j-th sample. i It is the i-th cluster center.

[0075] Clustering algorithms are used to cluster pixels to extract plume features. Since plumes are usually the brightest areas, the cluster with the largest channel value is selected as the main body of the plume.

[0076] S1-2, calculate the image texture intensity based on the plume body and determine the optical flow parameters corresponding to the image texture intensity.

[0077] In plume velocity inversion measurement, traditional fixed-parameter optical flow methods have technical drawbacks. For example, setting the window too small can lead to optical flow breaks, while setting it too large can cause motion blur. The same problem exists with the number of pyramid layers and iterations: blindly increasing these parameters leads to computational resource redundancy and overfitting risks, while blindly decreasing them can prevent convergence in high-gradient regions. Therefore, for plumes in different scenes or at different times, adaptive adjustments to the optical flow parameters are needed based on the texture complexity of the plume in the current frame.

[0078] First, let M be the plume body obtained in S1-1, and calculate its gradient fields M in the horizontal and vertical directions. x and M y :

[0079] M x =K x *M

[0080] M y =K y *M

[0081] Among them, K x With K y It is the Sobel operator kernel:

[0082]

[0083] For the (i,j)th pixel, based on the gradient fields in two directions and Calculate the gradient magnitude statistic for:

[0084]

[0085] Gradient magnitude statistics G of all pixels image The standard deviation std is used to represent the image texture intensity σ:

[0086] σ=std(G image )

[0087] The optical flow parameters corresponding to the image texture intensity are determined based on the correspondence between image texture intensity and optical flow parameters, as shown in Table 1:

[0088] Table 1. Correspondence between image texture intensity and optical flow parameters

[0089]

[0090] S1-3: Calculate the forward optical flow based on optical flow parameters, introduce a consistency error field based on bidirectional optical flow, and perform time consistency correction on the forward optical flow based on the consistency error field.

[0091] In this embodiment, the bidirectional optical flow is first calculated, and then the consistency error field E used for time consistency correction is calculated:

[0092] E = |F f +F b |

[0093] Among them, F f For the normally calculated forward optical flow (i.e., the pixel motion vector field from time t to t+1), F b This is the reverse optical flow (i.e., the pixel motion vector field from time t+1 to t) obtained through reverse calculation. Ideally, the forward and reverse optical flows should be opposite matrices, with E = 0. In reality, complex motions such as turbulence and rotation of the plume can lead to mismatch in the motion model, and light reflection / absorption during plume diffusion can cause brightness to deviate from the constant assumption, so the consistency error field is not zero. The physical meaning of this inconsistency is a measure of the local unreliability of optical flow estimation. To avoid errors caused by abnormal optical flow vectors, the forward optical flow F should be... f Time consistency is corrected:

[0094] F = F f -λ·E

[0095] Where λ is the correction weighting factor, which is set to λ = 0.2 based on experience, and F represents the corrected forward optical flow.

[0096] S1-4: Determine the fusion weight factor based on the gradient of the forward optical flow, and perform weighted total variation denoising optimization on the forward optical flow based on the fusion weight factor.

[0097] The significance of total variation denoising lies in preserving edges while smoothing optical flow by minimizing the total variation energy function. By fusing gradient information from optical flow and the image, a fusion weighting factor is calculated, which can adaptively control the smoothing intensity, balancing the needs of denoising and edge preservation in motion estimation.

[0098] First, calculate the magnitude gradient G of the forward optical flow. flow The corrected forward optical flow obtained from S1-3 is discretized and differentiated in the horizontal and vertical directions:

[0099]

[0100] Then there is the magnitude gradient G of the positive optical flow. flow for:

[0101]

[0102] The linear combination gradient method is used based on the magnitude gradient G of the forward optical flow. flow and gradient magnitude statistic G image Obtain the fusion weight factor w:

[0103] w=α·G flow +(1-α)·G image

[0104] Here, α represents the weight of the optical flow amplitude gradient, which is generally set to 0.7.

[0105] After obtaining the fusion weighting factor, weighted total variation denoising optimization is performed on the forward optical flow. The optimization objective of weighted total variation denoising is to make the optical flow smoother while maintaining sharp edges. Therefore, an energy function is constructed by combining the weighting factor, and the energy functional E(F) is minimized:

[0106] E(F)=∫w·G flow (F)dxdy

[0107] G flow (F) represents the gradient of the forward optical flow. Minimizing the energy functional E(F) minimizes the overall weighted gradient, allowing the image to achieve optimal smoothness while maintaining clear edges.

[0108] After performing variational analysis on the energy functional E(F), we obtain the divergence operator. The divergence reflects the energy gradient and serves as a tool for optimizing the energy functional. The discretized gradient divergence is calculated as follows:

[0109]

[0110] ∈ is a minimal quantity, avoiding division by zero errors. It is the weight update term, (i,j) represents the coordinate position, ΔF x (i,j) and ΔF y (i,j) represents the change in gradient divergence in the horizontal x-direction and the vertical y-direction. Gradient descent optimizes the energy functional through divergence. The formula for iteratively updating the optical flow field by gradient descent is as follows:

[0111]

[0112] η is the learning rate of gradient descent, set to 1. The update direction is the negative energy gradient. The optical flow after each update will make the energy functional smaller. k represents the update round.

[0113] S1-5, after weighted total variation denoising optimization, the flow velocity is calculated to obtain the plume velocity inversion result.

[0114] The calculation result of the forward optical flow actually records the displacement vector of each pixel in each optical flow module. The forward optical flow after iterative optimization is denoted as F′, which is the gradient divergence F including the horizontal x-direction and the vertical y-direction. x ′ (i,j) and F y ′ (i,j), then we have:

[0115]

[0116] m(i,j) refers to the magnitude of the displacement vector of the (i,j)th pixel in a frame, i.e., the pixel frame velocity, in pixels per frame. The actual velocity v(i,j) of each pixel is:

[0117] v(i,j)=m(i,j)×pm×FPS

[0118] The direction θ(x,y) of the true velocity is:

[0119] θ(x,y)=tan -1 (F x ′ (i,j),F y ′ (i,j))

[0120] Where pm is the conversion coefficient between pixels and actual physical relationships, obtained by a rangefinder combined with geometric calculations, and FPS refers to the video frame rate.

[0121] The actual velocity (speed) of each pixel and its corresponding direction constitute the plume velocity inversion result, such as... Figure 3 As shown.

[0122] S2 is used to verify the plume velocity inversion results.

[0123] Traditional machine vision-based methods for calculating plume velocity are typically limited to verification using gas distribution devices and simple models. Existing verification methods usually obtain a continuous velocity field (such as a velocity vector diagram or streamline diagram) by directly solving the Navier-Stokes equations (CFD module). However, optical flow methods rely on brightness variations in the image, and directly using the CFD velocity field cannot effectively simulate noise, particle occlusion, and image sampling rate issues in the actual imaging process, potentially leading to overly idealized verification results. To address this issue, this invention employs two methods to test the adaptability of the improved dense optical flow method Farneback to conditions such as uneven particle concentration, image noise, and dynamic range variations.

[0124] Verification Method 1

[0125] COMSOL software provides a plume simulation case based on a particle tracking model. Using this case, a 10-second, 50-frame-rate visualization model was generated as the base data source for validation. After exporting the model data, plume velocity data containing 120,000 rows of height information (each height is divided 484 times horizontally) and 500 columns of time information was obtained. To verify the accuracy of the improved dense optical flow Farneback method, the model data was processed as follows:

[0126] (1) Perform low-speed filtering on the model data, that is, filter out data points with speeds less than 0.1 m / s, in order to filter out regions in the model where there is no particle motion or the particle motion is sparse.

[0127] (2) The average value of the velocity data at different heights in the same frame is calculated, that is, the average value of the data at different heights in the same frame is calculated to obtain 500 data points for verification as true data;

[0128] The plume velocity inversion data obtained using the optical flow method and the improved dense optical flow method Farneback were processed in the same way, and correlation verification was performed with the ground truth data. The results are as follows: Figure 4 , 5 As shown. Analysis Figure 4 and Figure 5The results show that the correlation between the Farneback calculation results and the true data obtained by the improved dense optical flow method increased from 0.76 to 0.91, and the error decreased from 15% to 9%. This indicates that the accuracy of the improved Farneback method is significantly better than that of the original method.

[0129] Verification Method Two

[0130] Existing field verification methods use gas mixing devices, but these methods rely excessively on the readings of the gas mixing device and cannot obtain real-time data on the gas within the diffused plume. This invention utilizes a pulsed smoke generator in conjunction with an anemometer to achieve accurate field verification of the improved dense optical flow method, Farneback. The specific steps are as follows:

[0131] Turn on the pulsed smoke generator and record data for 10 seconds at a frame rate of 60. Place the anemometer horizontally at a position where the smoke plume diffusion radius is similar to the radius of the anemometer's impeller. Turn on the pulsed smoke generator and record the anemometer reading for 10 seconds. Repeat this process multiple times to collect anemometer data.

[0132] Simultaneously, the plume velocity inversion results are matched for pulse interval and time resolution relative to the pulsed smoke generator. Specifically, for pulse interval matching, the plume velocity inversion data obtained by the improved dense optical flow method Farneback needs to be processed as follows:

[0133] V k =max(V k·n V k·n+1 ,…,V k·(n+1) )

[0134] Among them, V k This represents the k-th data point used for verification in the improved optical flow method, and the number of frames captured during the n-th pulse. Since the pulsed smoke generator does not produce smoke uniformly, and the anemometer's plume velocity is limited by time resolution, it is necessary to take the maximum value of the plume velocity per n frames as the plume velocity captured by the improved dense optical flow method Farneback during each pulse.

[0135] To address the issue of matching time resolution, the improved dense optical flow method Farneback needs to match the time resolution of the anemometer. This means that the plume velocity inversion data after matching pulse interval processing needs to be resampled according to the time resolution of the anemometer.

[0136] Then, the error and variance between the matched plume velocity inversion results and the anemometer data are calculated to verify the results.

[0137] S3 combines plume velocity inversion results with hyperspectral remote sensing imaging to achieve high-precision emission flux quantification.

[0138] A ground-based hyperspectral remote sensing system for rapid high-resolution imaging of multi-component atmospheric trace components can achieve push-broom hyperspectral remote sensing imaging of multiple components, acquiring concentration information of a single plume stream simultaneously. Compared to the camera snapshot imaging used in the improved dense optical flow method Farneback, this ground-based hyperspectral remote sensing system for rapid high-resolution imaging of multi-component atmospheric trace components offers advantages such as eliminating cross-interference, high accuracy, and no calibration required. The steps for achieving high-precision emission flux quantification using the improved dense optical flow method Farneback and the ground-based hyperspectral remote sensing system for rapid high-resolution imaging of multi-component atmospheric trace components are as follows:

[0139] Before observation, ensure that the acquisition angle of the image acquisition camera in the improved dense optical flow method Farneback is consistent with the observation angle of the hyperspectral remote sensing system, and start the observation at the same time.

[0140] Spatial resolution matching is performed by setting the camera's spatial resolution, which is typically higher than that of the ground-based hyperspectral remote sensing system. Therefore, the number of pixels merging required by the camera, α, is calculated.

[0141]

[0142] Where, N c N represents the number of pixels occupied by the plume in the camera's field of view. h,L This represents the number of pixels occupied by the plume in a ground-based hyperspectral remote sensing system at a distance of L nanometers. It should be noted that this calculation needs to be performed separately in the horizontal and vertical directions.

[0143] Temporal resolution matching is also performed. Specifically, the acquired images are matched with the integration time of the ground-based hyperspectral remote sensing system. The typical integration time of the ground-based hyperspectral remote sensing system is 1500ms. Therefore, the camera needs to average the 90 frames of data acquired within 1500ms.

[0144] Based on this, the plume velocity inversion results are combined with hyperspectral remote sensing imaging to achieve high-precision emission flux quantification. Due to the large plume thickness near the chimney outlet, light sometimes has difficulty penetrating the plume, leading to an underestimation of concentration. Furthermore, the boundary at the plume tip is difficult to define, resulting in an underestimation. Therefore, the column used for flux calculation is the maximum value calculated column by column.

[0145]

[0146] Where M refers to the molecular mass of the component for which the flux is to be calculated, and N... A It is Avogadro's constant, SCD i,jIt is the column concentration value in the j-th column and i-th row, v i,j It is the velocity value in the j-th column and i-th row, N refers to the number of pixels in the j-th column containing the plume, h i,j This refers to the actual height represented by the height of each pixel in the results of a ground-based hyperspectral remote sensing system.

[0147] The above method can provide a high-precision, robust plume velocity calculation method and an accuracy verification method that does not depend on training data. Furthermore, it can be combined with a ground-based hyperspectral remote sensing system that uses rapid high-resolution imaging of multi-component atmospheric trace components to achieve high-precision emission flux quantification, thereby further improving the accuracy of emission flux calculation for organized industrial emissions.

[0148] The specific embodiments described above illustrate the technical solution and beneficial effects of the present invention in detail. It should be understood that the above description is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging, characterized in that, Includes the following steps: In the improved Farneback method for plume velocity inversion measurement, a clustering algorithm is used to extract plume features from images and filter plume entities. Based on the plume entities, image texture intensity is calculated, and the corresponding optical flow parameters are determined. The forward optical flow is then calculated based on these parameters. Finally, a consistency error field based on bidirectional optical flow is introduced, and time consistency correction is applied to the forward optical flow based on this field. The fusion weighting factor is determined based on the gradient of the forward optical flow, and the flow velocity is calculated after weighted total variation denoising based on the fusion weighting factor to obtain the plume flow velocity inversion result, specifically including: Calculate the magnitude gradient of the forward optical flow The linear combination gradient method is used based on the magnitude gradient of the forward optical flow. and gradient magnitude statistics Obtain the fusion weight factor : in, This indicates the weight of the optical flow magnitude gradient; Weighted total variation denoising of the forward optical flow is performed based on a fusion weighting factor. Specifically, the optimization objective of weighted total variation denoising is to make the optical flow smoother while maintaining sharp edges. This is achieved by constructing an energy function based on the weighting factor and minimizing the energy functional. : The gradient representing the forward optical flow, with respect to the energy functional. After performing variational analysis, we obtain the divergence operator. The divergence reflects the energy gradient and serves as a tool for optimizing energy functionals. The discretized gradient divergence is calculated as follows: It is a very small amount. It is a weight update item. Indicates coordinate position, and This represents the change in gradient divergence in the horizontal x-direction and the vertical y-direction. Gradient descent optimizes the energy functional through divergence. The formula for iteratively updating the optical flow field in gradient descent is as follows: It's the learning rate for gradient descent, and the update direction is the negative energy gradient. Indicates the update round; The forward optical flow after iterative optimization is denoted as That is, the gradient divergence includes the horizontal x-direction and the vertical y-direction. and Then we have: It refers to the first The magnitude of the displacement vector of a pixel in a frame, i.e., the pixel frame velocity magnitude, the true velocity of each pixel. Size: Direction of true speed for: in, This represents the conversion coefficient between pixels and their actual physical relationship. Refers to the frame rate of the video; The results of plume velocity inversion were verified, and the results of plume velocity inversion were combined with hyperspectral remote sensing imaging to achieve high-precision emission flux quantification.

2. The high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging according to claim 1, characterized in that, Clustering algorithms are used to extract plume features from images and filter plume entities, including: Clustering algorithms are used to cluster pixels to extract plume features. The plume is considered to be the area with the highest brightness, and the cluster with the largest channel value is selected as the main body of the plume.

3. The high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging according to claim 1, characterized in that, Calculate the image texture intensity based on the plume body and determine the corresponding optical flow parameters, including: The gradient fields in the horizontal and vertical directions of the feather body are calculated, and the gradient magnitude statistics of each pixel are statistically analyzed based on the gradient fields in the two directions. The standard deviation of the gradient magnitude statistics is used to represent the image texture intensity. The optical flow parameters corresponding to the image texture intensity are determined according to the correspondence between the image texture intensity and the optical flow parameters.

4. The high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging according to claim 3, is characterized in that, The correspondence is as follows: When image texture intensity When the texture level is in the range [0, 5), the corresponding pyramid scaling factor is 0.2, the number of pyramid layers is 2, the window size is 7, and the number of iterations is 2. When image texture intensity When the texture level is in the range [5, 10), the corresponding pyramid scaling factor is 0.25, the number of pyramid layers is 3, the window size is 11, and the number of iterations is 3. When image texture intensity When the texture level is [10, 20), the corresponding pyramid scaling factor is 0.4, the number of pyramid layers is 4, the window size is 15, and the number of iterations is 4. When image texture intensity When the texture level is [20, 30), the corresponding pyramid scaling factor is 0.5, the number of pyramid layers is 5, the window size is 21, and the number of iterations is 5. When image texture intensity When the texture level is in the range [30, +∞), the corresponding pyramid scaling factor is 0.6, the number of pyramid layers is 6, the window size is 25, and the number of iterations is 6.

5. The high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging according to claim 1, characterized in that, The forward optical flow is calculated based on optical flow parameters. A consistency error field based on bidirectional optical flow is introduced, and time consistency correction of the forward optical flow is performed based on the consistency error field, including: Farneback, a dense optical flow method, calculates the forward optical flow based on optical flow parameters. Introducing a uniformity error field based on bidirectional optical flow Represented as: in, This is the reverse optical flow obtained through reverse calculation. Time consistency correction of the forward optical flow based on the consistency error field is expressed as: in, For the corrected weighting factor, This indicates the corrected forward optical flow.

6. The high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging according to claim 1, characterized in that, The results of the plume velocity inversion were verified, including: Based on the plume simulation case based on the particle tracking model provided in COMSOL software, plume velocity data is exported as model data. Low velocity is filtered out from the model data and the average value of velocity data at different heights in the same frame is used as the ground truth data. After filtering out low velocities and averaging the velocities at different heights in the same frame, the plume velocity inversion data were compared with the ground truth data for correlation verification.

7. The high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging according to claim 1, characterized in that, The results of the plume velocity inversion were verified, including: The field verification of plume velocity inversion results was achieved using a pulsed smoke generator in conjunction with an anemometer, specifically including: An anemometer was used to collect anemometer data using a pulsed smoke generator. At the same time, the plume velocity inversion results were matched with the pulse interval and time resolution of the pulsed smoke generator. The error and variance between the matched plume velocity inversion results and the anemometer data were calculated to verify the results.

8. The high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging according to claim 1, characterized in that, High-precision emission flux quantification is achieved by combining plume velocity inversion results with hyperspectral remote sensing imaging, including: Before observation, it is ensured that the acquisition angle of the image acquisition camera in the improved dense optical flow method Farneback is consistent with the observation angle of the hyperspectral remote sensing system. Spatial resolution matching is achieved by setting the camera's spatial resolution to be higher than that of the ground-based hyperspectral remote sensing system. Temporal resolution matching is also achieved by integrating the acquired images with the integration time of the ground-based hyperspectral remote sensing system. Based on this, the plume velocity inversion results are combined with hyperspectral remote sensing imaging to achieve high-precision emission flux quantification.

9. The high-precision emission flux quantification method for plume velocity inversion, verification, and fusion imaging according to claim 1 or 8, characterized in that, High-precision emission flux quantification is achieved by combining plume velocity inversion results with hyperspectral remote sensing imaging, including: in, This refers to the molecular mass of the component for which the flux is being sought. It is Avogadro's constant. It is the first Liede The column concentration value of the row, It is the first Liede The speed value of the line, N refers to the speed value of the line. The number of pixels containing the feather. This refers to the actual height represented by the height of each pixel in the results of a ground-based hyperspectral remote sensing system.

Images