Multispectral photoelectric measurement method and system based on smoke screen interference condition

Abstract

The invention discloses a multispectral photoelectric measurement method and system based on a smoke screen interference condition, and relates to the technical field of photoelectric measurement, and the method comprises the steps: obtaining a smoke screen type and smoke screen concentration; adjusting a multispectral camera based on the smoke screen type and the smoke screen concentration to obtain a multispectral photoelectric image; constructing a multispectral denoising model based on the generative adversarial network model; performing denoising processing on the multispectral photoelectric image through a multispectral denoising model to obtain a denoised multispectral photoelectric image; performing image enhancement processing on the de-noised multispectral photoelectric image to obtain an enhanced multispectral photoelectric image; and performing target identification on the enhanced multispectral photoelectric image through a target identification algorithm to obtain a multispectral photoelectric measurement result. According to the method, the smoke screen type and the smoke screen concentration are dynamically obtained, and the multispectral camera parameters are adjusted according to the smoke screen type and the smoke screen concentration, so that the reliability and precision of multispectral photoelectric measurement in a complex environment are improved.

Description

Technical Field

[0001] The present invention relates to the field of optoelectronic measurement technology, and in particular to a multispectral optoelectronic measurement method and system under the condition of smoke screen interference. Background Art

[0002] Multispectral imaging technology has been widely applied in recent years in fields such as environmental monitoring, military reconnaissance, agricultural monitoring, and resource exploration. By acquiring image information in multiple bands, this technology can provide richer data content than traditional monochromatic or color imaging. However, in complex environments, especially in the presence of smoke screen interference, traditional multispectral optoelectronic measurement methods face numerous challenges. As a common natural or man-made phenomenon, a smoke screen can significantly reduce the transmittance in the visible and near-infrared bands, leading to a decrease in target recognition accuracy and affecting the quality of multispectral images. To overcome this problem, researchers have continuously explored new technologies and methods, such as correction algorithms based on physical models and image restoration technologies driven by deep learning, to improve the multispectral image processing ability in harsh environments.

[0003] Although the above technologies have improved the quality of multispectral images under smoke screen conditions to a certain extent, the existing solutions still have obvious deficiencies. First, most existing methods lack dynamic adaptability to the type and concentration of the smoke screen. Since different types of smoke screens (such as natural fog, industrial emissions, or military smoke screens) have different spectral characteristics, correction algorithms with fixed parameters are difficult to meet diverse actual needs. Second, current denoising models usually rely on a large amount of labeled data for training, which not only increases the complexity of the preliminary preparation work but also limits the generalization ability of the model. Especially when facing newly emerged or unseen samples, the model performance often drops significantly, and it is impossible to ensure a stable and reliable improvement in image quality. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a multispectral optoelectronic measurement method under the condition of smoke screen interference, which solves the problems of insufficient dynamic adaptability to smoke screen characteristics and the dependence of the denoising model on a large amount of labeled data in the prior art.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a multi-spectral optoelectronic measurement method under smoke interference conditions, which includes: obtaining the smoke type and smoke concentration; adjusting a multi-spectral camera based on the smoke type and smoke concentration to obtain a multi-spectral optoelectronic image; constructing a multi-spectral denoising model based on a generative adversarial network model; performing denoising processing on the multi-spectral optoelectronic image through the multi-spectral denoising model to obtain a denoised multi-spectral optoelectronic image; performing image enhancement processing on the denoised multi-spectral optoelectronic image to obtain an enhanced multi-spectral optoelectronic image; and performing target recognition on the enhanced multi-spectral optoelectronic image through a target recognition algorithm to obtain a multi-spectral optoelectronic measurement result.

[0008] As a preferred embodiment of the multi-spectral optoelectronic measurement method under smoke interference conditions according to the present invention, wherein: the steps of obtaining the smoke type and smoke concentration are as follows:

[0009] Training a CNN model based on historical environmental spectral data;

[0010] Analyzing the environmental spectral data through the trained CNN model to obtain the smoke type;

[0011] Analyzing the environmental spectral data through the light scattering principle to obtain the smoke concentration.

[0012] As a preferred embodiment of the multi-spectral optoelectronic measurement method under smoke interference conditions according to the present invention, wherein: the steps of adjusting the multi-spectral camera based on the smoke type and smoke concentration to obtain a multi-spectral optoelectronic image are as follows:

[0013] Updating the weight vector using the recursive least squares method based on the smoke type and smoke concentration;

[0014] Adjusting the central wavelength of the built-in filter of the multi-spectral camera based on the updated weight vector;

[0015] Adjusting the bandwidth of the built-in filter of the multi-spectral camera based on the updated weight vector;

[0016] The multi-spectral camera obtains a multi-spectral optoelectronic image based on the adjusted central wavelength and bandwidth.

[0017] As a preferred embodiment of the multi-spectral optoelectronic measurement method under smoke interference conditions according to the present invention, wherein: the steps of constructing a multi-spectral denoising model based on the generative adversarial network model are as follows:

[0018] Using the generative adversarial network model as the basic model;

[0019] The input layer of the generator receives historical multi-spectral optoelectronic images;

[0020] The encoder gradually extracts the features of the historical multi-spectral optoelectronic image and compresses the spatial resolution through multiple convolutional layers and pooling layers;

[0021] The skip connection connects the intermediate features of the encoder to the corresponding layer of the decoder through feature splicing, enhancing the detailed information of historical multi-spectral optoelectronic images;

[0022] The decoder gradually restores the resolution of the historical multi-spectral optoelectronic image through the transposed convolution layer for detail reconstruction;

[0023] Through the last convolutional layer of the decoder, the denoising processing result is output;

[0024] The discriminator discriminates the denoising processing result;

[0025] Finally, a multi-spectral denoising model is constructed.

[0026] As a preferred solution of the multi-spectral optoelectronic measurement method based on the smoke interference condition of the present invention, wherein: the multi-spectral optoelectronic image is denoised by the multi-spectral denoising model to obtain a denoised multi-spectral optoelectronic image, and the specific steps are as follows:

[0027] Input the multi-spectral optoelectronic image into the multi-spectral denoising model, and the multi-spectral optoelectronic image is denoised by the generator in the multi-spectral denoising model;

[0028] The discriminator in the multi-spectral denoising model crops the sub-region image of the denoised multi-spectral image in a sliding window manner;

[0029] The local features of the sub-region image are gradually extracted through multiple convolutional layers;

[0030] The pooling layer captures the abstract features of the local features;

[0031] The fully connected layer flattens the abstract features;

[0032] The output layer obtains the noise-free probability value P through the Sigmoid activation function;

[0033] Based on the statistical characteristics of the noise-free probability value P, a threshold T is set to discriminate the denoising effect of the denoised multi-spectral optoelectronic image;

[0034] When P > T, it indicates that the denoising effect of the denoised multi-spectral optoelectronic image is good, and it is directly output;

[0035] When P ≤ T, it indicates that the denoising effect of the denoised multi-spectral optoelectronic image is poor, and it is denoised again until the denoising effect P > T is satisfied;

[0036] Through the discrimination result, the denoised multi-spectral optoelectronic image is obtained.

[0037] As a preferred embodiment of the multi-spectral optoelectronic measurement method under the condition of smoke interference according to the present invention, the denoised multi-spectral optoelectronic image is processed by image enhancement to obtain an enhanced multi-spectral optoelectronic image. The specific steps are as follows:

[0038] Use the Laplace operator to enhance the edges and details of the denoised multi-spectral optoelectronic image;

[0039] Perform frequency domain processing through Fourier transform and inverse Fourier transform;

[0040] Perform color balance and color enhancement on the denoised multi-spectral optoelectronic image through white balance and color enhancement;

[0041] Finally, obtain the enhanced multi-spectral optoelectronic image.

[0042] As a preferred embodiment of the multi-spectral optoelectronic measurement method under the condition of smoke interference according to the present invention, the enhanced multi-spectral optoelectronic image is subjected to target recognition through a target recognition algorithm to obtain multi-spectral optoelectronic measurement results. The specific steps are as follows:

[0043] Perform edge detection on the enhanced multi-spectral optoelectronic image through an edge detection algorithm to extract the edge features of the enhanced multi-spectral optoelectronic image;

[0044] Based on the edge detection results, use connected component analysis to extract candidate regions;

[0045] Based on each candidate region, extract the spectral features of the enhanced multi-spectral optoelectronic image;

[0046] Match the extracted spectral features with the spectral feature library, and calculate the similarity S through the Euclidean distance;

[0047] Set a threshold N based on the similarity distribution of targets and non-targets in the spectral feature library;

[0048] Based on the similarity S and the threshold N, determine the target region corresponding to the candidate region;

[0049] When S≥N, the candidate region is the target region;

[0050] When S<N, the candidate region is the non-target region;

[0051] Based on the target region, obtain multi-spectral optoelectronic measurement results through the multi-spectral optoelectronic spatial characteristics and multi-spectral optoelectronic spectral characteristics.

[0052] Second aspect, the present invention provides a multi-spectral optoelectronic measurement system under smoke interference conditions, including: a smoke analysis module, a camera adjustment module, a model construction module, an image denoising module, an image enhancement module, and a target recognition module; the smoke analysis module is used to obtain the smoke type and smoke concentration; the camera adjustment module is used to adjust the multi-spectral camera based on the smoke type and smoke concentration to obtain a multi-spectral optoelectronic image; the model construction module is used to construct a multi-spectral denoising model based on a generative adversarial network model; the image denoising module is used to perform denoising processing on the multi-spectral optoelectronic image through the multi-spectral denoising model to obtain a denoised multi-spectral optoelectronic image; the image enhancement module is used to perform image enhancement processing on the denoised multi-spectral optoelectronic image to obtain an enhanced multi-spectral optoelectronic image; the target recognition module is used to perform target recognition on the enhanced multi-spectral optoelectronic image through a target recognition algorithm to obtain a multi-spectral optoelectronic measurement result.

[0053] Third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the multi-spectral optoelectronic measurement method under smoke interference conditions as described in the first aspect of the present invention is implemented.

[0054] Fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the multi-spectral optoelectronic measurement method under smoke interference conditions as described in the first aspect of the present invention is implemented.

[0055] The beneficial effects of the present invention are as follows: By dynamically obtaining the smoke type and smoke concentration and adjusting the multi-spectral camera parameters accordingly, the present invention realizes the adaptive response to complex environments and ensures the optimization of image acquisition; at the same time, an efficient multi-spectral denoising model is constructed based on the generative adversarial network model, and excellent denoising effects can be achieved without a large amount of labeled data, significantly improving the image quality and the accuracy of target recognition, thus greatly optimizing the entire process from image acquisition to target recognition, and ultimately enhancing the reliability and accuracy of multi-spectral optoelectronic measurement in complex environments. Description of the Drawings

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for description in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0057] Figure 1 It is a flowchart of the multi-spectral optoelectronic measurement method under smoke interference conditions in Embodiment 1.

[0058] Figure 2 It is a schematic diagram of the multi-spectral optoelectronic measurement system under the condition of smoke interference in Embodiment 1. Specific implementation manner

[0059] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the specific implementation manners of the present invention will be described in detail below with reference to the accompanying drawings of the specification.

[0060] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0061] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure or characteristic that may be included in at least one implementation manner of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it an individual or alternative embodiment that excludes other embodiments.

[0062] Embodiment 1, referring to Figure 1 and Figure 2 , which is the first embodiment of the present invention. This embodiment provides a multi-spectral optoelectronic measurement method under the condition of smoke interference, including the following steps:

[0063] S1: Obtain the smoke type and smoke concentration.

[0064] S1.1: Train the CNN model based on historical environmental spectral data.

[0065] Specifically, the historical environmental spectral data is divided into a training set, a validation set and a test set by means of random division or stratified sampling;

[0066] The convolutional neural network (CNN) model is trained using the training set, the performance of the convolutional neural network (CNN) model is monitored using the validation set to prevent overfitting, and hyperparameter adjustment is performed if necessary. The performance of the convolutional neural network (CNN) model is evaluated using the test set, and metrics such as accuracy, recall rate, and F1 score are calculated to obtain the trained convolutional neural network (CNN) model.

[0067] S1.2: Analyze the environmental spectral data through the trained CNN model to obtain the smoke type.

[0068] Specifically, the environmental spectral data is converted into a two-dimensional image form, where the light intensity value of each wavelength is represented as a pixel of the image. These two-dimensional image forms are input into a trained convolutional neural network (CNN) model. The CNN model extracts local features through multiple convolutional layers, and then reduces the spatial dimension of the feature map through the pooling layer to enhance the robustness of the features. The feature map is flattened and further processed through the fully connected layer, and finally the probability distribution of each smoke screen type is output through the softmax layer. The category with the highest probability is selected as the final smoke screen type;

[0069] It should be noted that the environmental spectral data is collected by a spectrometer, including spectral intensity data and environmental parameter data.

[0070] S1.3: Analyze the environmental spectral data through the light scattering principle to obtain the smoke screen concentration.

[0071] Specifically, according to the Mie scattering theory, the spectral intensity is proportional to the smoke screen concentration, and the proportionality constant is determined by calibration in a standard sample with a known concentration;

[0072] The spectral intensity in the actual measurement environment is measured, and the smoke screen concentration is calculated using the formula. The expression is:

[0073]

[0074] where D is the smoke screen concentration, is the proportionality constant, and J is the spectral intensity.

[0075] S2: Based on the smoke screen type and the smoke screen concentration, adjust the multispectral camera to obtain a multispectral optoelectronic image.

[0076] S2.1: Based on the smoke screen type and the smoke screen concentration, use the recursive least squares method to update the weight vector.

[0077] It should be noted that the specific expression is:

[0078]

[0079] where w(n) is the updated weight vector at the current time step n, w(n - 1) is the weight vector at the previous time step n - 1, P(n - 1) is the covariance matrix at the previous time step n - 1, s(n) is the smoke screen type vector at the current time step n, D(n) is the smoke screen concentration vector at the current time step n, e(n) is the error at the current time step n, Θ is a constant, s H (n) is the conjugate transpose of s(n), and n is the current time step;

[0080] It should also be noted that the center wavelength and bandwidth of the filter continuously update the weight vector and the covariance matrix according to the real-time error feedback to optimize the signal-to-noise ratio and signal intensity of the image;

[0081] It should be noted that the covariance matrix update formula:

[0082]

[0083] where P(n) is the covariance matrix at the current time n, P(n - 1) is the covariance matrix at the previous time (n - 1), x(n) is the smoke type vector at the current time n, and x H (n) is the conjugate transpose operation of the smoke type vector at the current time n, H is the conjugate transpose, and θ is a constant.

[0084] S2.2: Adjust the central wavelength of the built-in filter of the multispectral camera based on the updated weight vector.

[0085] It should be noted that the specific expression is:

[0086] λ(n) = λ0 + Δλ·w1(n);

[0087] where λ(n) is the adjusted central wavelength at the current time n, λ0 is the initial central wavelength, Δλ is the wavelength adjustment step, and w1(n) is the first element of the weight vector at the current time n.

[0088] S2.3: Adjust the bandwidth of the built-in filter of the multispectral camera based on the updated weight vector.

[0089] It should be noted that the specific expression is:

[0090] B(n) = B0 + ΔB·w2(n);

[0091] where B(n) is the bandwidth at the current time n, B0 is the initial bandwidth, ΔB is the bandwidth adjustment step, and w2(n) is the second element of the weight vector at the current time n.

[0092] S2.4: The multispectral camera obtains a multispectral optoelectronic image based on the adjusted central wavelength and bandwidth.

[0093] S3: Construct a multispectral denoising model based on the generative adversarial network model.

[0094] S3.1: Use the generative adversarial network model as the basic model;

[0095] The input layer of the generator receives historical multispectral optoelectronic images;

[0096] The encoder gradually extracts the features of the historical multispectral optoelectronic images and compresses the spatial resolution through multiple convolutional layers and pooling layers;

[0097] The skip connection connects the intermediate features of the encoder to the corresponding layers of the decoder through feature splicing to enhance the detailed information of the historical multispectral optoelectronic images;

[0098] The decoder gradually restores the resolution of the historical multi-spectral optoelectronic image through the deconvolution layer for detail reconstruction;

[0099] Through the last convolutional layer of the decoder, the denoising result is output;

[0100] The discriminator discriminates the denoising result;

[0101] Finally, a multi-spectral denoising model is constructed.

[0102] It should be noted that the reason for constructing the multi-spectral denoising model based on the generative adversarial network (GAN) model is that it can efficiently learn complex image features with a small amount of labeled data, significantly improve the denoising effect through the adversarial training mechanism of the generator and the discriminator, and enhance the ability to restore image details, thereby effectively improving the quality of multi-spectral images and the accuracy of target recognition.

[0103] S4: The multi-spectral optoelectronic image is denoised by the multi-spectral denoising model to obtain the denoised multi-spectral optoelectronic image.

[0104] S4.1: The multi-spectral optoelectronic image is input into the multi-spectral denoising model, and the multi-spectral optoelectronic image is denoised by the generator in the multi-spectral denoising model.

[0105] It should be explained that the specific expression is:

[0106]

[0107] Where, I is the denoised multi-spectral optoelectronic image, W is the deconvolution kernel of the output layer of the decoder, V′ l is the deconvolution kernel of the l-th layer of the decoder, F l is the output feature of the l-th layer of the encoder, U l is the output feature of the l-th layer of the decoder, Q′ l is the bias term of the l-th layer of the decoder, b is the bias term of the output layer of the decoder, R is the activation function, is the feature splicing operation, l is the index of the current layer, and L is the total number of network layers.

[0108] It should also be explained that the expression represents the denoising process of the multi-spectral optoelectronic image, which is as follows:

[0109] The multi-spectral optoelectronic image passes through each layer of convolution kernels and feature splicing operations, combines the output features of the encoder and the output features of the decoder, as well as the bias term, undergoes weighted summation and activation function processing, and then passes through the linear combination of the weight matrix and the bias term, and finally outputs the denoised multi-spectral optoelectronic image through the hyperbolic tangent tanh activation function.

[0110] S4.2: The discriminator in the multi-spectral denoising model crops the sub-region image of the denoised multi-spectral image in a sliding window manner.

[0111] S4.3: Gradually extract the local features of the sub-region image through multiple convolutional layers.

[0112] S4.4: The pooling layer captures the abstract features of the local features.

[0113] S4.5: The fully connected layer flattens the abstract features.

[0114] S4.6: The output layer will obtain the noise-free probability value through the Sigmoid activation function.

[0115] It should be noted that the specific expression is:

[0116] P = σ(K f ·Z + a f )

[0117] where P is the noise-free probability value, σ is the activation function, K f is the weight matrix of the fully connected layer, Z is the flattened feature vector, and a f is the bias term of the fully connected layer.

[0118] S4.7: Set the threshold T based on the statistical characteristics of the noise-free probability value P to judge the denoising effect of the denoised multi-spectral optoelectronic image;

[0119] When P > T, it indicates that the denoising effect of the denoised multi-spectral optoelectronic image is good, and it is directly output;

[0120] When P ≤ T, it indicates that the denoising effect of the denoised multi-spectral optoelectronic image is poor, and it is denoised again until the denoising effect P > T is satisfied.

[0121] It should be noted that the threshold is set by analyzing the distribution of the noise-free probability value through statistical analysis methods.

[0122] S4.8: Obtain the denoised multi-spectral optoelectronic image through the discrimination result.

[0123] S5: The denoised multi-spectral optoelectronic image is processed by image enhancement to obtain the enhanced multi-spectral optoelectronic image.

[0124] S5.1: Use the Laplacian operator to enhance the edges and details of the denoised multi-spectral optoelectronic image.

[0125] Specifically, apply the Laplacian operator to the denoised multi-spectral optoelectronic image to detect and highlight the edge and detail information in the denoised multi-spectral optoelectronic image. By superimposing the original denoised multi-spectral optoelectronic image and the high-frequency components obtained after the Laplacian operator processing, enhance the edge sharpness and detail texture of the denoised multi-spectral optoelectronic image.

[0126] S5.2: Perform frequency domain processing through Fourier transform and inverse Fourier transform.

[0127] Specifically, apply Fourier transform to the denoised multispectral optoelectronic image to convert it from the spatial domain to the frequency domain, so as to separate and process different frequency components. In the frequency domain, enhance or suppress specific frequency components by adjusting the amplitude of specific frequency components, and convert the processed frequency domain denoised multispectral optoelectronic image back to the spatial domain through inverse Fourier transform.

[0128] S5.3: Perform color balance and color enhancement on the denoised multispectral optoelectronic image through white balance and color enhancement.

[0129] Specifically, apply the white balance algorithm to adjust the color temperature of the denoised multispectral optoelectronic image, ensure that the relative brightness between bands conforms to natural lighting conditions, eliminate color cast, and use color enhancement techniques (such as histogram stretching or enhancement based on the hue-saturation-intensity (HSI) model) to enhance the saturation and contrast of colors in the denoised multispectral optoelectronic image, making the colors more vivid and realistic.

[0130] S5.4: Finally, obtain the enhanced multispectral optoelectronic image.

[0131] S6: Perform target recognition on the enhanced multispectral optoelectronic image through the target recognition algorithm to obtain the multispectral optoelectronic measurement result.

[0132] S6.1: Perform edge detection on the enhanced multispectral optoelectronic image through the edge detection algorithm to extract the edge features of the enhanced multispectral optoelectronic image.

[0133] S6.2: Based on the edge detection result, use connected component analysis to extract candidate regions.

[0134] S6.3: Based on each candidate region, extract the spectral features of the enhanced multispectral optoelectronic image.

[0135] It should be noted that the specific steps for extracting the spectral features of the enhanced multispectral optoelectronic image are as follows:

[0136] Calculate the mean and standard deviation of each band;

[0137] Collect the average reflectance values of each band within the candidate region, arrange them in the order of wavelength, and draw a curve of reflectance versus wavelength;

[0138] Use the combination of two or more bands to construct new indices;

[0139] Integrate the various spectral feature means, standard deviations, reflectance curves, and spectral indices calculated above into a comprehensive feature vector to obtain the spectral features of the enhanced multispectral optoelectronic image.

[0140] S6.4: Match the extracted spectral features with the spectral feature library, and calculate the similarity through the Euclidean distance.

[0141] It should be noted that the specific expression is:

[0142]

[0143] Among them, S is the similarity, M is the extracted spectral feature, and Y is the target spectrum in the spectral feature library.

[0144] S6.5: Set the threshold N based on the similarity distribution of targets and non-targets in the spectral feature library;

[0145] Judge the target area corresponding to the candidate area based on the similarity S and the threshold N;

[0146] When S≥N, the candidate area is the target area;

[0147] When S<N, the candidate area is the non-target area.

[0148] It should be noted that the target in the spectral feature library refers to the spectral features of the specific object to be recognized, and the non-target in the spectral feature library refers to the spectral features of all other irrelevant background or interference objects;

[0149] Analyze the similarity distribution of targets and non-targets through statistical analysis methods and set the threshold.

[0150] S6.6: Based on the target area, obtain the multispectral optoelectronic measurement results through the multispectral optoelectronic spatial characteristics and multispectral optoelectronic spectral characteristics.

[0151] It should be noted that the multispectral optoelectronic spatial characteristics are obtained through the description of the target area area, perimeter and shape. The specific process is as follows:

[0152] By counting the number of all pixels in the target area, multiplying the total number of pixels by the ground area corresponding to a single pixel (provided by the image metadata), the actual area of the target area is obtained;

[0153] Precisely define the positions of the boundary pixels through the edge detection algorithm, and then calculate the Euclidean distance between adjacent pixels point by point along this path. Finally, add up the distances between all adjacent pixels to obtain the perimeter of the target area;

[0154] Obtain the compactness by multiplying four times pi by the area and then dividing by the square of the perimeter;

[0155] Obtain the rectangularity by dividing the target area by the area of its minimum bounding rectangle;

[0156] Obtain the aspect ratio by calculating the ratio of the length of the major axis of the target area;

[0157] The centroid coordinates are obtained by calculating the average value of all pixel coordinates;

[0158] The bounding box is obtained by determining the position and size of the smallest rectangle enclosing the target area;

[0159] The geometric morphological features of the target area are comprehensively quantified through compactness, rectangularity, aspect ratio, centroid coordinates, and bounding box, thereby obtaining a comprehensive shape description;

[0160] By analyzing the area, perimeter, and shape description of the target area, its geometric morphological features are quantified, thereby obtaining the spatial characteristics of the multispectral optoelectronic image.

[0161] It should also be noted that the multispectral optoelectronic spectral characteristics are obtained through band statistics, reflectance curves, and characteristic bands. The specific steps are as follows:

[0162] Calculate the mean, standard deviation, maximum value, and minimum value of each band to obtain band statistics;

[0163] Collect the reflectance values of the target area in each band and draw a curve of reflectance versus wavelength to obtain the reflectance curve;

[0164] According to the research purpose or application requirements, select the band that best represents a specific land cover type or phenomenon as the characteristic band;

[0165] By analyzing the band statistics and reflectance curve to identify specific patterns and changes in the characteristic band, the optoelectronic characteristics of the ground objects in the multispectral image are analyzed.

[0166] This embodiment also provides a multispectral optoelectronic measurement system under smoke interference conditions, including: a smoke analysis module, a camera adjustment module, a model construction module, an image denoising module, an image enhancement module, and a target recognition module;

[0167] The smoke analysis module is used to obtain the smoke type and smoke concentration;

[0168] The camera adjustment module is used to adjust the multispectral camera based on the smoke type and smoke concentration to obtain a multispectral optoelectronic image;

[0169] The model construction module is used to construct a multispectral denoising model based on the generative adversarial network model;

[0170] The image denoising module is used to perform denoising processing on the multispectral optoelectronic image through the multispectral denoising model to obtain a denoised multispectral optoelectronic image;

[0171] The image enhancement module is used to perform image enhancement processing on the denoised multispectral optoelectronic image to obtain an enhanced multispectral optoelectronic image;

[0172] A target recognition module, which is used to perform target recognition on the enhanced multispectral optoelectronic image through a target recognition algorithm to obtain multispectral optoelectronic measurement results.

[0173] This embodiment also provides a computer device applicable to the case of a multispectral optoelectronic measurement method under smoke interference conditions, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the multispectral optoelectronic measurement method under smoke interference conditions proposed in the above embodiment.

[0174] This computer device may be a terminal, and this computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of this computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of this computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of this computer device may be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0175] This embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the multispectral optoelectronic measurement method under smoke interference conditions proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (Static Random Access Memory, abbreviated as SRAM), an electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), an erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), a programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), a read-only memory (Read-Only Memory, abbreviated as ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0176] In summary, the present invention realizes the adaptive response to complex environments and ensures the optimization of image acquisition by dynamically obtaining the smoke screen type and concentration and adjusting the multi-spectral camera parameters accordingly. At the same time, an efficient multi-spectral denoising model is constructed based on the generative adversarial network model, achieving excellent denoising effects without a large amount of labeled data, significantly improving the image quality and the accuracy of target recognition, thus greatly optimizing the entire process from image acquisition to target recognition, and ultimately enhancing the reliability and accuracy of multi-spectral optoelectronic measurement in complex environments.

[0177] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A multi-spectral photoelectric measurement method under smoke interference conditions, characterized in that: include: Get the smoke type and smoke density; Based on the type and density of smoke, the multispectral camera is adjusted to obtain a multispectral photoelectric image; Based on the generative adversarial network model, a multi-spectral denoising model is constructed; De-noising the multispectral photoelectric image by using a multispectral de-noising model to obtain a de-noised multispectral photoelectric image; The denoised multispectral photoelectric image is processed by image enhancement to obtain an enhanced multispectral photoelectric image; Through the target recognition algorithm, the enhanced multispectral photoelectric image is recognized and the multispectral photoelectric measurement results are obtained.

2. The multi-spectral photoelectric measurement method based on smoke interference conditions as claimed in claim 1, characterized in that: The specific steps of obtaining the smoke type and smoke density are as follows: Train the CNN model based on historical environmental spectral data; The environmental spectrum data is analyzed through the trained CNN model to obtain the type of smoke screen; The environmental spectrum data is analyzed through the principle of light scattering to obtain the smoke concentration.

3. The multi-spectral photoelectric measurement method based on smoke interference conditions as claimed in claim 2 is characterized in that: The multispectral camera is adjusted based on the smoke type and smoke density to obtain a multispectral photoelectric image. The specific steps are as follows: Based on the smoke type and smoke density, the weight vector is updated using recursive least squares method; adjusting the center wavelength of the built-in filter of the multispectral camera based on the updated weight vector; adjusting the built-in filter bandwidth of the multispectral camera based on the updated weight vector; The multispectral camera acquires a multispectral optoelectronic image based on the adjusted central wavelength and bandwidth.

4. The multi-spectral photoelectric measurement method based on smoke interference conditions as claimed in claim 3 is characterized in that: The multispectral denoising model is constructed based on the generative adversarial network model, and the specific steps are as follows: The basic model is based on the generative adversarial network model; The generator input layer receives historical multispectral electro-optical images; The encoder gradually extracts historical multispectral photoelectric image features and compresses spatial resolution through multiple convolutional layers and pooling layers; The jump connection connects the intermediate features of the encoder to the corresponding layer of the decoder through feature concatenation to enhance the detail information of the historical multispectral optoelectronic image; The decoder gradually restores the historical multispectral optoelectronic image resolution through the deconvolution layer and reconstructs the details; Through the last convolutional layer of the decoder, the denoising result is output; The denoising result is judged by a discriminator; Finally, a multispectral denoising model was constructed.

5. The multi-spectral photoelectric measurement method based on smoke interference conditions as claimed in claim 4 is characterized in that: The multispectral photoelectric image is denoised by using a multispectral denoising model to obtain a denoised multispectral photoelectric image. The specific steps are as follows: The multispectral photoelectric image is input into the multispectral denoising model, and the multispectral photoelectric image is denoised by the generator in the multispectral denoising model; The discriminator in the multispectral denoising model crops the sub-region image of the denoised multispectral image by means of a sliding window; The local features of the sub-region image are gradually extracted through multiple convolutional layers; The pooling layer captures the abstract features of local features; The fully connected layer flattens the abstract features; The output layer obtains the noise-free probability value P through the Sigmoid activation function; The threshold T is set based on the statistical characteristics of the noise-free probability value P to determine the denoising effect of the denoised multispectral photoelectric image; When P>T, it means that the denoising effect of the denoised multispectral photoelectric image is good and it is directly output; When P≤T, it means that the denoising effect of the denoised multispectral photoelectric image is poor, and denoising is performed again until the denoising effect P>T is satisfied; Through the discrimination results, a denoised multispectral photoelectric image is obtained.

6. The multi-spectral photoelectric measurement method based on smoke interference conditions as claimed in claim 5, characterized in that: The denoised multispectral photoelectric image is processed by image enhancement to obtain an enhanced multispectral photoelectric image, and the specific steps are as follows: Enhance the edges and details of denoised multispectral optoelectronic images using the Laplacian operator; Frequency domain processing via Fourier transform and inverse Fourier transform; Color balance and color enhancement of denoised multispectral photoelectric images through white balance and color enhancement; Finally, enhanced multispectral optoelectronic images are obtained.

7. The multi-spectral photoelectric measurement method based on smoke interference conditions as claimed in claim 6 is characterized in that: The target recognition algorithm is used to perform target recognition on the enhanced multispectral photoelectric image to obtain the multispectral photoelectric measurement result. The specific steps are as follows: The edge detection algorithm is used to detect the edge of the enhanced multispectral photoelectric image and extract the edge features of the enhanced multispectral photoelectric image; Based on the edge detection results, the candidate regions are extracted using connected domain analysis; Based on each candidate region, the spectral features of the enhanced multispectral optoelectronic image are extracted; Match the extracted spectral features with the spectral feature library and calculate the similarity S through the Euclidean distance; The threshold N is set based on the similarity distribution between the target and non-target in the spectral feature library; Based on the similarity S and the threshold N, determine whether the candidate area corresponds to the target area; When S ≥ N, the candidate region is the target region; When S<N, the candidate area is a non-target area; Based on the target area, the multispectral photoelectric measurement results are obtained through the multispectral photoelectric spatial characteristics and the multispectral photoelectric spectral characteristics.

8. A multi-spectral photoelectric measurement system under smoke interference conditions, based on the multi-spectral photoelectric measurement method under smoke interference conditions according to any one of claims 1 to 7, characterized in that: include: Smoke analysis module, camera adjustment module, model building module, image denoising module, image enhancement module and target recognition module; Smoke analysis module, used to obtain smoke type and smoke density; A camera adjustment module, used to adjust the multispectral camera based on the smoke type and smoke density to obtain a multispectral photoelectric image; A model building module is used to build a multi-spectral denoising model based on a generative adversarial network model; An image denoising module is used to denoise a multispectral photoelectric image by using a multispectral denoising model to obtain a denoised multispectral photoelectric image; An image enhancement module is used to denoise a multispectral photoelectric image and obtain an enhanced multispectral photoelectric image through image enhancement processing; The target recognition module is used to perform target recognition on the enhanced multispectral photoelectric image through a target recognition algorithm to obtain multispectral photoelectric measurement results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the multi-spectral photoelectric measurement method under smoke interference conditions described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the multi-spectral photoelectric measurement method under smoke interference conditions described in any one of claims 1 to 7 are implemented.

Images