Photon number statistical method and system for solar blind ultraviolet imager

Abstract

The invention relates to the technical field of ultraviolet imaging, in particular to a photon number statistical method and system for a solar-blind ultraviolet imager, and the method comprises the steps: obtaining an ultraviolet image, and carrying out the multi-scale decomposition of the ultraviolet image, so as to separate a noise object from a photon object; noise objects in the ultraviolet image are filtered and eliminated, and photon objects are reserved; carrying out binarization processing on the filtered ultraviolet image to obtain a binarized image; a communication relationship between each pixel point and adjacent pixel points in the binarized image is analyzed, a plurality of connected domains containing a plurality of foreground pixels are established based on an analysis result, and the communication relationship comprises position communication and attribute communication; and counting the number of foreground pixels in each connected domain to calculate the total area of all connected domains, obtaining the pixel area of a pre-calibrated single photon object, and converting the equivalent photon number based on the pixel area and the total area. The method and the device have the effect of reducing photon counting underestimation caused by photon overlapping.

Description

Technical Field

[0001] The present application relates to the technical field of ultraviolet imaging, and in particular to a photon counting method and system for a solar-blind ultraviolet imager. Background Art

[0002] After a long period of operation in the atmosphere, high-voltage equipment in substations and transmission lines will have some structural defects, or the insulation performance will decrease with the increase of surface dirt and the influence of environmental humidity, thus generating corona or partial discharge. Corona is a high-voltage pulse discharge that will cause chemical reactions in the air, producing substances such as ozone and nitrogen oxides, causing the insulation performance of electrical equipment to decrease, and thus causing equipment accidents. Therefore, timely and accurate detection of the corona discharge level is crucial to ensure the safe operation of the power grid.

[0003] During the discharge process, electrons in the air continuously gain and release energy, and when electrons release energy, they also release ultraviolet light. Ultraviolet imaging technology uses this principle to receive ultraviolet light generated by high-voltage electrical equipment during discharge, to determine the location and intensity of the corona, and thus provide a reliable basis for evaluating the operation of the equipment.

[0004] The number of photons in the UV imager can indicate the discharge intensity of the corona and plays an important role in assessing the severity of the disaster at the discharge location. However, the current UV imager directly counts the number of connected domains of the binary image when counting the number of photons, but ignores some key issues:

[0005] Photons have a superposition effect, that is, during high-density discharge, multiple electrons will be superimposed in the same connected domain. Traditional technical methods will regard photons superimposed in the same area as being classified in the same connected domain, which will seriously underestimate the number of real photons. At the same time, the connected domains in existing technologies all use a fixed area (such as fixed deletion of connected domains with less than 5 pixels), which will mistakenly filter out some weak photon signals. Summary of the invention

[0006] In order to reduce the underestimation of photon counting caused by photon overlap, the present application provides a photon counting method and system for a solar-blind ultraviolet imager.

[0007] In the first aspect, the present application provides a method for counting photon numbers of a solar-blind ultraviolet imager, which adopts the following technical solution:

[0008] A method for counting photons of a solar-blind ultraviolet imager comprises the following steps:

[0009] Acquire UV images and perform multi-scale decomposition to separate noise objects from photon objects;

[0010] filtering and eliminating the noise object in the ultraviolet image and retaining the photon object;

[0011] Perform binarization processing on the filtered ultraviolet image to obtain a binary image;

[0012] Analyze the connectivity relationship between each pixel point and its adjacent pixel points in the binary image, and establish a number of connected regions containing a number of foreground pixels based on the analysis results, where the connectivity relationship includes position connectivity and attribute connectivity;

[0013] Count the number of foreground pixels in each connected region to calculate the total area of all connected regions, obtain the pixel area of a single photon object calibrated in advance, and calculate the equivalent number of photons based on the pixel area and the total area.

[0014] In some embodiments, obtaining an ultraviolet image and performing multi-scale decomposition to separate noise objects and photon objects includes the following steps:

[0015] Obtain a scale factor and a translation factor, and perform wavelet transform on the ultraviolet image in combination with the Haar wavelet function, where,

[0016] Analyze the low-frequency features and high-frequency features in the ultraviolet image by changing the scale factor to obtain a frequency distribution, and control the position of the Haar wavelet function on the time axis by changing the translation factor to obtain a position distribution;

[0017] Generate high-frequency components of feature points based on the frequency distribution and the position distribution;

[0018] Define the feature points corresponding to the high-frequency components with concentrated position distribution and stable frequency distribution as the photon objects, and define the feature points corresponding to the high-frequency components with discrete position distribution and fluctuating frequency distribution as the noise objects.

[0019] In some embodiments, filtering and eliminating the noise objects and retaining the photon objects includes the following steps:

[0020] After performing the wavelet transform, obtain wavelet coefficients, extract high-frequency subband coefficients from the wavelet coefficients, calculate the median absolute deviation of each high-frequency subband coefficient and calculate the noise standard deviation in combination with a preset distribution constant;

[0021] Calculate the total number of pixels of the ultraviolet image, and calculate a dynamic threshold in combination with the noise standard deviation;

[0022] Compare each wavelet coefficient with the dynamic threshold, and filter all wavelet coefficients smaller than the dynamic threshold to complete the filtering of the ultraviolet image.

[0023] In some of these embodiments, after calculating the dynamic threshold, the following steps are further included:

[0024] Perform region segmentation on the ultraviolet image based on the total number of pixels to obtain a number of sub-images;

[0025] Calculate the sub-wavelet coefficients corresponding to each of the sub-images, and statistically analyze the probability distribution of the sub-wavelet coefficients with each value;

[0026] Calculate the characteristic entropy corresponding to each of the sub-images based on the probability distribution, and calculate the average entropy corresponding to the number of sub-images;

[0027] Establish a linear mapping function based on the characteristic entropy and the average entropy to calculate a threshold adjustment coefficient, and optimize the dynamic threshold based on the threshold adjustment coefficient.

[0028] In some of these embodiments, performing binarization processing on the filtered ultraviolet image to obtain a binarized image includes the following steps:

[0029] Obtain the pixel intensity of each pixel point in the filtered ultraviolet image, and calculate the global mean value of the image as a reference value;

[0030] Assign a value of 1 to the pixel points whose pixel intensity is higher than the reference value to define them as foreground pixels, and assign a value of 0 to the pixel points whose pixel intensity is lower than the reference value to define them as background pixels;

[0031] Eliminate the background pixels through morphological closing operations of dilation first and then erosion to obtain the binarized image.

[0032] In some of these embodiments, analyzing the connectivity relationship between each pixel point and its adjacent pixel points in the binarized image, and establishing a number of connected regions containing a number of foreground pixels based on the analysis results includes the following steps:

[0033] Traverse all the pixel points in the binarized image and mark the passed foreground pixels to add labels;

[0034] When marking, based on the 8-neighborhood algorithm, determine whether there are adjacent foreground pixels with labels around. If so, select the label with the smallest value that conforms to the connectivity relationship for marking. If not, recursively generate a new label based on the largest value among the existing labels in the current binarized image;

[0035] Combine a number of foreground pixels with the same label to form a connected region.

[0036] In some of these embodiments, the method for calculating the total area of all the connected components by counting the number of foreground pixels in each connected component, obtaining the pixel area of a single photon object calibrated in advance, and calculating the equivalent number of photons based on the pixel area and the total area includes the following steps:

[0037] Define the pixel area of the single photon object calibrated in advance as the standard area;

[0038] Calculate the sub-area based on the number of foreground pixels included in each connected component;

[0039] Obtain the maximum aspect ratio of each connected component and generate a first coefficient;

[0040] Obtain the position distribution of the hole pixels in each connected component in the connected component to generate a second coefficient, where the hole pixels are characterized as foreground pixels that do not all have a connected relationship in the 8-neighborhood;

[0041] Obtain the ratio between the sub-area of each connected component and the overall area of the ultraviolet image to generate a third coefficient;

[0042] Adjust the standard area based on the first coefficient, the second coefficient, and the third coefficient to obtain the final pixel area;

[0043] Divide each sub-area by the pixel area to obtain the number of photons per unit, and sum several numbers of photons per unit to calculate the equivalent number of photons.

[0044] In some of these embodiments, calculating the equivalent number of photons based on the pixel area and the total area further includes the following steps:

[0045] Label the connected components with a sub-area smaller than the pixel area of the photon object as objects to be verified;

[0046] Calculate the minimum pixel distance between the objects to be verified and other connected components;

[0047] When the minimum pixel distance is less than a preset value, set the number of photons per unit calculated for the object to be verified to 1;

[0048] When the minimum pixel distance is greater than the preset value, the object to be verified does not participate in the calculation of the number of photons per unit.

[0049] In some of these embodiments, the method further includes the following steps:

[0050] Obtain the measurement distance of the ultraviolet image, and correct the equivalent number of photons based on the attenuation model. Specifically, ,

[0051] wherein, is characterized as the corrected photon number, is characterized as the equivalent photon number, d is characterized as the measurement distance, and D is characterized as the constant parameter of the ultraviolet imager.

[0052] In a second aspect, the present application provides a solar-blind ultraviolet imager photon number statistics system, adopting the following technical solution:

[0053] A solar-blind ultraviolet imager photon number statistics system is used to implement the above method.

[0054] The technical solution provided by the embodiments of the present application has the following technical effects:

[0055] Firstly, a three-stage hybrid denoising architecture is adopted to separate noise and photon signals, and filtering and binarization processing are performed. At the same time, the number of connected components in the binarized image is counted. The area of the connected components is not specified for defense, but the overall area is determined based on connected components of different sizes, and the method of dividing by the pixel area of a calibrated single photon is used to convert the equivalent photon number, ensuring that even if multiple photon signals overlap in the same area, their total contribution will be correctly counted, thus avoiding the problem of underestimating the photon number in traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a step schematic diagram of a solar-blind ultraviolet imager photon number statistics method provided by this embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] To understand the purpose, technical solution, and advantages of the present application more clearly, the present application will be described and explained below with reference to the drawings and embodiments. However, those of ordinary skill in the art should understand that the present application can be implemented without these details. In some cases, to avoid unnecessary descriptions from making aspects of the present application obscure, well-known methods, processes, systems, components, and / or circuits that have been described at a higher level will not be elaborated too much. For those of ordinary skill in the art, it is obvious that various changes can be made to the disclosed embodiments of the present application, and without departing from the principles and scope of the present application, the general principles defined in the present application can be applied to other embodiments and application scenarios. Therefore, the present application is not limited to the shown embodiments, but conforms to the broadest scope consistent with the scope claimed in the present application.

[0058] It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation to the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0059] In the description of the present application, the meaning of "several" is one or more, the meaning of "multiple" is more than two, and understandings such as "greater than", "less than", "exceeding", etc. do not include the base number, and understandings such as "above", "below", "within", etc. include the base number. If there is a description of "first" and "second", it is only used to distinguish technical features and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence relationship of the indicated technical features.

[0060] In the description of the present application, the description with reference to terms such as "one embodiment", "some embodiments", "schematic embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic descriptions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a combined manner.

[0061] As Figure 1 shown, the embodiments of the present application disclose a method for photon number statistics of a solar-blind ultraviolet imager, including the following steps:

[0062] S100, acquire an ultraviolet image and perform multi-scale decomposition to separate noise objects and photon objects.

[0063] In the actual use of the ultraviolet imager, it will acquire one or more frames of ultraviolet images at the target position, and the normally acquired ultraviolet images contain photon pulses corresponding to, for example, partial discharge or corona, and also include other noises corresponding to the environment or other heat sources.

[0064] Then, in order to accurately retain photon objects when measuring the photon number subsequently, it is necessary to first preprocess the ultraviolet image to separate noise objects and photon objects, facilitating subsequent processing steps such as filtering and binarization.

[0065] First of all, it is necessary to perform multi-scale decomposition on the ultraviolet image. The function of multi-scale decomposition is applicable to decomposing the information in the image into different frequency components. The high-frequency information obtained after multi-scale decomposition will contain the detailed information of the image. Then, according to the distribution and positional relationship of high-frequency and low-frequency in its frequency components, the noise objects and photon objects in the ultraviolet image can be effectively distinguished.

[0066] S200, filter and eliminate the noise objects in the ultraviolet image and retain the photon objects.

[0067] Secondly, filter the noise objects separated after multi-scale decomposition to effectively remove the noise in the ultraviolet image while retaining and restoring the photon pulse characteristics.

[0068] Filtering and eliminating the noise can effectively increase the photon signal intensity in the image, making subsequent photon statistics more accurate and reducing the possibility of misjudgment and missed judgment.

[0069] S300, perform binarization on the filtered ultraviolet image to obtain a binary image.

[0070] After filtering, perform binarization on the current ultraviolet image. Binarization is based on a required threshold to distinguish between photon pixels and noise pixels in the ultraviolet image, and represent the photon pixels and noise pixels with different binary values and further eliminate discrete noise points.

[0071] At the same time, the binary image obtained after binarization is also convenient for subsequent selection of connected regions to count the number of photons.

[0072] S400, analyze the connectivity relationship between each pixel point and adjacent pixel points in the binary image, and establish several connected regions containing several foreground pixels based on the analysis results. The connectivity relationship includes position connectivity and attribute connectivity.

[0073] Perform a comprehensive scan analysis on the binary image. The role of the analysis is to judge whether each pixel point in the image conforms to the same connectivity relationship with adjacent pixel points based on the neighborhood algorithm. If it conforms, it means that the pixel point and adjacent pixel points belong to the same connected region.

[0074] Among them, the connectivity relationship includes position connectivity and attribute connectivity. Position connectivity is characterized by the adjacency between a pixel point and another pixel point, and the adjacency relationship varies according to different neighborhood algorithms. In the 4-neighborhood algorithm, position connectivity is characterized by the presence of pixel points adjacent to the left, right, top, and bottom of a pixel point. In the 8-neighborhood algorithm, the position connectivity relationship is characterized by the presence of pixel points adjacent to the left, right, top, bottom, upper left, lower left, upper right, and lower right of a pixel point. Attribute connectivity means that pixel points that meet the position connectivity are all characterized as photon pixels. Among them, photon pixels are defined as foreground pixels, while irrelevant noise pixels are characterized as background pixels.

[0075] After traversing all pixel points, several different connected regions can be obtained. Each connected region is composed of photon pixels characterized as photon objects, and each connected region contains at least one photon pixel.

[0076] S500, count the number of foreground pixels in each connected region to calculate the total area of all connected regions, obtain the pixel area of a single pre-calibrated photon object, and calculate the equivalent number of photons based on the pixel area and the total area.

[0077] In the existing technical solution, it is default that each connected component corresponds to a photon number. However, in the actual environment, due to reasons such as different discharge faults, different acquisition positions, and different acquisition angles, there will be a problem of multiple photons overlapping. Then, in this case, a connected component often contains multiple overlapping photons. Therefore, the existing technical method will seriously underestimate the photon number when photons overlap, resulting in multiple overlapping pixels being misrecognized as a single photon signal, leading to poor final detection error.

[0078] In this application, to solve this problem, first, the area of the region corresponding to each connected component is calculated according to the number of foreground pixels, and at the same time, the pixel area of a calibrated and dynamic single photon is obtained.

[0079] Dividing the area of the region of each connected component by the pixel area of a single photon can calculate the theoretical photon number of each connected component. This is because when multiple photons overlap, the photon energy increases, which will cause the pixel area corresponding to the photon signal in the image to increase. When multiple photons overlap, some of the energy between the overlapping photons may be superimposed in the binary image. However, compared with the traditional method, this method can effectively make the contribution of the overlapping photons be statistically relatively correctly, improving the accuracy of the statistics.

[0080] Finally, by accumulating the photon numbers calculated for each connected component, the total equivalent photon number in the ultraviolet image can be obtained.

[0081] It should be noted that in the equivalent conversion of the above areas, when the result has a decimal, the rounding method is used to obtain an integer equivalent photon number.

[0082] Through the above method, first, a three-level hybrid denoising architecture is adopted to separate noise and photon signals, and filtering and binarization processing are carried out. At the same time, the number of connected components in the binary image is counted, and no defense is set for the area of the connected component. Instead, based on connected components of different sizes, the overall area is determined, and the method of dividing by the pixel area of a calibrated single photon is used to convert the equivalent photon number, ensuring that even if multiple photon signals overlap in the same region, their total contribution will be correctly statistically, thus avoiding the problem of the traditional method underestimating the photon number.

[0083] Obtain an ultraviolet image and perform multi-scale decomposition to separate noise objects and photon objects, including the following steps:

[0084] S110, obtain a scale factor and a translation factor, and perform wavelet transform on the ultraviolet image in combination with the Haar wavelet function.

[0085] First, the formula for multi-scale decomposition in the embodiments of this application is as follows: ,

[0086] Among them, a represents the scale factor, and b represents the translation factor. represents the mother wavelet function. In the embodiments of the present application, the Haar wavelet function is adopted. t represents the current time. represents the wavelet coefficient.

[0087] S120. Analyze the low-frequency features and high-frequency features in the ultraviolet image by changing the scale factor to obtain the frequency distribution, and control the position of the Haar wavelet function on the time axis by changing the translation factor to obtain the position distribution.

[0088] The scale factor is used to stretch or scale the wavelet function at multiple scales. When the scale factor increases, the wavelet function is stretched, and it will analyze the features of larger scales (low frequencies) in the signal; when the scale factor decreases, the wavelet function is compressed, and it can focus on the features of smaller sizes (high frequencies) in the signal. Therefore, the low-frequency features and high-frequency features of each signal in the ultraviolet image can be analyzed through the scale factor to analyze the frequency distribution.

[0089] The translation factor can control the position of the wavelet function on the time axis, and analyze the signal by continuously changing the size of the translation factor to adjust different positions, so as to analyze the position distribution of the signal.

[0090] The Haar wavelet has a rectangular waveform. In signal analysis, it can effectively capture the mutation points and local features of the signal. Regarding the ultraviolet image as a signal f(t) and performing discrete wavelet transform on it, the image can be decomposed into components of different frequencies. Among them, the high-frequency components contain the detailed information of the image, such as the edges, textures and other features of the objects in the image.

[0091] S130. Generate the high-frequency components of the feature points based on the frequency distribution and the position distribution.

[0092] The high-frequency components are characterized as the partial frequency components in the image that reflect the details and high-speed change features after wavelet transform, and they reflect the fine structure and change situation of the signal in the local area.

[0093] S140. Define the feature points corresponding to the high-frequency components with concentrated position distribution and stable frequency distribution as photon objects, and define the feature points corresponding to the high-frequency components with discrete position distribution and fluctuating frequency distribution as noise objects.

[0094] In solar-blind ultraviolet imaging, photon signals and random noise present different features in the high-frequency components.

[0095] First, for photon signals, they exhibit strong spatial concentration and relatively stable intensity. Therefore, in the image, the high-frequency components corresponding to photon signals will gather together in spatial positions, and the high-frequency distribution part is relatively stable. For example, when an electrical device discharges to generate photons, the signals formed by these photons on the image are concentrated in a certain local area, and their intensity changes little in a short time.

[0096] Secondly, for the noise signals corresponding to random noise, they exhibit uniform distribution, and the intensity change of the high-frequency component part fluctuates greatly. Random noise may appear at various positions in the image, without an obvious concentrated area, and its intensity may change greatly at any time, sometimes large and sometimes small. For example, the noise generated by electromagnetic interference during the imaging process will be evenly distributed throughout the image.

[0097] Therefore, by analyzing the position distribution and frequency distribution in the high-frequency components, the corresponding photon objects and noise objects can be separated.

[0098] In some other embodiments, filtering and eliminating the noise objects and retaining the photon objects include the following steps:

[0099] S210, after performing wavelet transform to obtain wavelet coefficients, extract the high-frequency subband coefficients from the wavelet coefficients, calculate the median absolute deviation of each high-frequency subband coefficient, and calculate the noise standard deviation in combination with a preset distribution constant.

[0100] After performing wavelet transform, wavelet coefficients are calculated. The wavelet coefficients are a coefficient matrix, which contains different frequency bands, including LL (low frequency - low frequency), LH (low frequency - high frequency), HL (high frequency - low frequency), HH (high frequency - high frequency). Among them, the high-frequency subband coefficients include LH, HL, and HH. The high-frequency subband coefficients contain the detailed information of the image and can be used to identify the edges and features of photon information.

[0101] When filtering the noise objects, the soft threshold filtering method needs to be adopted. Then, first, a threshold for distinguishing noise objects and photon objects needs to be set. At the same time, because different image precisions, different image sizes, etc. all affect the precision of the filtering process, the threshold needs to be further dynamically processed so that it can be applicable to more scenarios.

[0102] Specifically, first extract all the high-frequency subband coefficients in the wavelet coefficients, and calculate the median absolute deviation of all the coefficients. The specific formula is:

[0103] Among them, is characterized as the high-frequency subband coefficient.

[0104] After calculating the median absolute deviation, the noise standard deviation is calculated by the following formula.

[0105] ,

[0106] where is a constant derived based on the Gaussian noise distribution.

[0107] By calculating the noise standard deviation through the above method, the noise standard deviation is a key indicator of the noise intensity in the horizontal image.

[0108] S220, calculate the total number of pixels in the ultraviolet image, and calculate the dynamic threshold in combination with the noise standard deviation.

[0109] Secondly, obtain the total number of pixels N in the ultraviolet image, and calculate the dynamic threshold based on the following formula: ,

[0110] In the related art, generally, after calculating the noise standard deviation, noise filtering is directly performed based on this value, which seriously depends on linear filters such as the median value. Such methods will cause signal distortion in a complex electromagnetic interference environment. At the same time, the fixed median filtering threshold is difficult to adapt to dynamic noise (daylight fluctuations), and it will blur the photon pulse transformation.

[0111] Therefore, in the embodiments of the present application, while calculating the noise standard deviation, the total number of pixels in the image is collected. Generally speaking, there is no direct numerical relationship between the noise standard deviation and the total number of pixels. However, in actual applications, the total number of pixels will affect the accuracy of the noise standard deviation. When the total number of pixels is large, the number of samples used to calculate the median absolute deviation is large, and the estimated value of the noise standard deviation obtained will be more accurate and stable. On the contrary, when the total number of pixels is small, the error of the estimated value may be large.

[0112] Therefore, by collecting the total number of pixels, the dynamic weight of the noise standard deviation can be further dynamically verified. When dealing with different images and noise situations, the filtering threshold calculated by the noise standard deviation can be dynamically optimized based on the weight corresponding to the total number of pixels, and the dynamic change relationship of the threshold under different relationships between the noise quantity and the pixel quantity can be further accurately determined.

[0113] At the same time, by the dynamically adjusted threshold, it is also possible to avoid the loss of some weak photon signals caused by over-segmentation in some cases where the imaging effect of the image is poor.

[0114] S230, compare each wavelet coefficient with the dynamic threshold, and filter all wavelet coefficients smaller than the dynamic threshold to complete the filtering of the ultraviolet image.

[0115] Set the filtering threshold through dynamic threshold setting. For each wavelet coefficient, if its value is greater than the dynamic threshold, it is considered that this coefficient is very likely to represent a photon signal and is retained. Conversely, if it is less than the dynamic threshold, it is considered that it may represent a noise signal and is discarded to complete the filtering.

[0116] The basis for filtering is as follows: Since the intensity of photon signals is relatively stable and concentrated, the high-frequency coefficients corresponding to most photon signals will be greater than the threshold, while the intensity of random noise fluctuates greatly, and the high-frequency coefficients corresponding to many noises will be less than the threshold, so that effective separation of photon signals and noises can be achieved.

[0117] Since the intensity of photon signals is relatively stable and concentrated, the high-frequency coefficients corresponding to most photon signals will be greater than the threshold, while the intensity of random noise fluctuates greatly, and the high-frequency coefficients corresponding to many noises will be less than the threshold, so that effective separation of photon signals and noises can be achieved.

[0118] In some other embodiments, after calculating the dynamic threshold, in order to further optimize the size of the dynamic threshold, the following steps are further included:

[0119] S240, perform regional segmentation on the ultraviolet image based on the total number of pixels to obtain several sub-images.

[0120] The local feature entropy of an image can reflect the complexity and information richness of the region. In an image, the region where photon signals are concentrated usually has a higher feature entropy, while the feature entropy of the noise region is relatively low. By calculating the feature entropy of the local region of the image, the threshold coefficient can be dynamically adjusted according to the complexity of different regions, so as to more accurately separate photon signals and noises.

[0121] Therefore, first perform segmentation of small windows on the ultraviolet image to obtain several sub-images, and the segmentation size can be adjusted according to the specific size, resolution, detection environment, etc. of the image.

[0122] S250, calculate the sub-wavelet coefficients corresponding to each sub-image, and statistically analyze the probability distribution of the sub-wavelet coefficients with each value.

[0123] After segmentation, it is necessary to calculate the sub-wavelet coefficients corresponding to each sub-image respectively, and calculate the feature entropy of each local image based on the analysis of the sub-wavelet coefficients.

[0124] Specifically, statistically analyze the distribution of wavelet coefficients in each sub-image, calculate the probability of the coefficient value of each wavelet coefficient appearing, and thus obtain its probability distribution.

[0125] Such as coefficient The number of occurrences is count ( ), then its probability is .

[0126] Wherein, m and n respectively represent the horizontal distance and vertical distance of the sub-image.

[0127] S260. Calculate the feature entropy corresponding to each sub-image based on the probability distribution, and calculate the average entropy corresponding to several sub-images.

[0128] After calculating the probability distribution content of each wavelet coefficient, calculate the feature entropy by combining the probability distribution based on the definition formula of the feature entropy. Specifically: ,

[0129] The summation here is performed on all different wavelet coefficients within each sub-image.

[0130] After calculating the feature entropy corresponding to each sub-image, it is also necessary to calculate the average entropy corresponding to all sub-images. The calculation process of the average entropy is to accumulate the feature entropy of all sub-images and divide by the number of sub-images to calculate.

[0131] The feature entropy represents the richness of photon signals in each independent local region, and the average entropy also represents the average value of the richness of photon signals in the entire image.

[0132] S270. Establish a linear mapping function based on the feature entropy and the average entropy to calculate the threshold adjustment coefficient, and optimize the dynamic threshold based on the threshold adjustment coefficient.

[0133] Finally, it is necessary to determine the corresponding coefficient for adjusting the final dynamic threshold through the feature entropy and the average entropy.

[0134] First, retrieve a mapping function to establish a linear mapping relationship, and further map the feature entropy to a threshold adjustment factor.

[0135] The linear mapping function in this application is: ,

[0136] Wherein, represents the threshold adjustment coefficient, H represents the feature entropy of the current sub-image, represents the average entropy, and k is an adjustable constant parameter, which is determined through experiments and generally takes -0.1.

[0137] When the feature entropy is greater than the average entropy, it indicates that the photon signal features in this sub-image are more complex compared to the overall average, and it may contain more detailed photon information. At this time, the threshold adjustment coefficient is less than 1, which means that the dynamic threshold needs to be dynamically reduced to retain more photon signals. When the feature entropy is less than the average value, it indicates that the photon signal features in this sub-image are relatively simple compared to the overall average and may contain more random noise. At this time, the threshold adjustment coefficient will be greater than 1, which means that the dynamic threshold needs to be dynamically increased to filter out more noise signals to enhance the suppression of noise.

[0138] Then, after finally optimizing the dynamic threshold, the new dynamic threshold is: ,

[0139] Through the above dynamic threshold optimization scheme based on feature entropy, the dynamic threshold can be adaptively adjusted according to the feature complexity of the local area of the image, more accurately separating photon signals and noise in different regions, and improving the denoising effect and the accuracy of photon number statistics.

[0140] In some other embodiments, the filtered ultraviolet image is binarized to obtain a binary image, including the following steps:

[0141] S310, obtain the pixel intensity of each pixel point in the filtered ultraviolet image, and calculate the global mean value of the image as a reference value.

[0142] After filtering the ultraviolet image, it is necessary to provide an image basis for subsequent steps such as connected component search and marking, and it is also necessary to binarize the ultraviolet image.

[0143] During binarization, first obtain the pixel intensity of each pixel point in the ultraviolet image, that is, the gray value size of each pixel point on the image, recorded as I(x, y).

[0144] Calculate the global mean value of the ultraviolet image. The global mean value is the result obtained by dividing the sum of the pixel intensities of all pixel points in the image by the total number of pixel points, recorded as . The global mean value is used as an adaptive threshold, that is, a reference value for determining whether a pixel point is 0 or 1.

[0145] S320, assign a value of 1 to the pixel points with pixel intensity higher than the reference value to define them as foreground pixels, and assign a value of 0 to the pixel points with pixel intensity lower than the reference value to define them as background pixels.

[0146] Establish a binarization process system, specifically: ,

[0147] Characterized as a binarization result, when the pixel intensity is higher than the reference value, the corresponding pixel is assigned a value of 1, and at the same time it is defined as a foreground pixel (photon). When the pixel intensity is lower than the preset value, the corresponding pixel is assigned a value of 0, and at the same time it is defined as a background pixel (background noise).

[0148] S330, eliminate background pixels through morphological closing operation of dilation first and then erosion to obtain a binarized image.

[0149] After binarization processing, it is also necessary to eliminate discrete noise through morphological closing operation. Among them, the dilation operation is characterized by first expanding the foreground area in the image to fill the holes in the foreground area; the erosion operation is to shrink the foreground area to remove noise and small discrete points, while retaining the morphology and boundary features of important photon signal pulses.

[0150] Through binarization conversion and morphological closing operation, the clarity and distinguishability of the signal are enhanced (the binarized image only contains 0 and 1). At the same time, discrete noise is removed while retaining the integrity of the photon signal, thus providing a more accurate basis for subsequent connected component analysis and photon number statistics.

[0151] In some other embodiments, analyze the connectivity relationship between each pixel point and adjacent pixel points in the binarized image, and establish several connected components containing several foreground pixels based on the analysis results, including the following steps:

[0152] S410, traverse all pixel points in the binarized image and mark the foreground pixels passed through to add labels.

[0153] S420, when marking, based on the 8-neighborhood algorithm, judge whether there are foreground pixels with adjacent positions and existing labels around. If so, select the label with the smallest value that conforms to the connectivity relationship for marking. If not, recursively generate a new label based on the largest value among the existing labels in the current binarized image.

[0154] S430, jointly form a connected component with several foreground pixels with the same label.

[0155] First, perform the search for connected components:

[0156] Use DFS to search for connected components in the binarized image. During the search process, start from each unmarked pixel and check its adjacent pixels through 8-neighborhood. All a group of pixels adjacent to the current pixel and with the same attribute will be regarded as the same connected component. It should be noted that for the search of connected components, the search objects are all foreground pixels.

[0157] Secondly, perform the marking of connected components:

[0158] For each detected connected component, all pixels within that component are marked with a unique label. By using a recursive method, the same label is assigned to all pixels belonging to the same connected component, and the marking information of each pixel is stored in a two-dimensional array.

[0159] During the marking process, first, in the first scan, starting from the upper left corner of the image, each pixel is scanned row by row and column by column. When a foreground pixel is encountered and it is not marked, a new connected component is started and a label is provided, and all adjacent foreground pixels will also be marked as belonging to the same connected component and provided with the same label.

[0160] When a foreground pixel is scanned and it is found that there is more than one label, that is, there is a conflict between the labels, the label with the smallest value within the 8-neighborhood is selected.

[0161] After that, a second scan is performed to check all unmarked foreground pixels and assign them the corresponding connected component labels.

[0162] By the above method, all connected components in the binary image are effectively identified and marked, providing basic data for subsequent area calculation and photon number statistics.

[0163] In some other embodiments, to calculate the total area of all connected components by counting the number of foreground pixels in each connected component, obtain the pixel area of a single pre-calibrated photon object, and convert the equivalent photon number based on the pixel area and the total area, the following steps are included:

[0164] S510, Define the pixel area of a single pre-calibrated photon object as the standard area.

[0165] After counting and marking all connected components in the image, first obtain the pixel area of a single pre-calibrated photon object, which is obtained after the experiment and represents the pixel area that an ideal photon object occupies in the image, and it is mainly also determined by the resolution of the detector.

[0166] S520, Calculate the sub-area based on the number of foreground pixels included in each connected component.

[0167] Judge the number of foreground pixels included in each connected component, and combine the resolution of the image to obtain the size of each pixel point, and thus calculate the sub-area of the entire connected component.

[0168] Normally, only by calculating the ratio of the area of each connected component to the area of a single photon can the number of photons N included in all connected components be calculated, as shown in the following formula: ,

[0169] Among them, M represents the total number of connected components, Characterized as the sub-area corresponding to the i-th connected component, Characterized as the pixel area corresponding to a single photon pixel.

[0170] However, in fact, since the binary image only reflects the difference between foreground pixels and background pixels and cannot intuitively reflect the pixel intensity of photon pixels in each pixel point, when there is overlap between multiple photons, it will cause the photon information of partial overlap to be fused in one pixel grid after binarization, which makes the binary image actually slightly ignore the area of some superimposed photons.

[0171] Therefore, in order to optimize the ignored photon area and improve the accuracy of the final photon count, it is also necessary to dynamically adjust the fixed standard area corresponding to the unit photon pixel in this application to make it an adjustable dynamic pixel area based on the environment and scenario.

[0172] Specifically,

[0173] S530, obtain the maximum and minimum aspect ratios of each connected component and generate a first coefficient.

[0174] First, determine the maximum and minimum aspect ratios of each connected component. Since normal photon signals without overlap will appear approximately circular or rectangular in actual imaging and the boundaries are relatively clear, when multiple photon signals overlap, if they overlap horizontally, it will cause the horizontal pixel distance of the connected component corresponding to the overlapping photons to be greater than its vertical pixel distance; if they overlap vertically, it will cause the vertical pixel distance of the connected component corresponding to the overlapping photons to be greater than its horizontal pixel distance; and when multiple photons overlap randomly, it will cause the shape of the corresponding connected component to be extremely irregular, and the ratio difference between the longest width and the shortest length, and between the longest length and the shortest width is relatively large.

[0175] Therefore, by calculating the maximum and minimum aspect ratios such as maximum length - maximum width, minimum length - minimum width, maximum length - minimum width, and minimum length - maximum width in each connected component, the shape distribution of each connected component can be analyzed.

[0176] Ideally, when there are no overlapping photons, the maximum and minimum values of each aspect ratio should be close to 1. When any aspect ratio is abnormal, the larger the abnormal value, the more overlapping photons may exist. At this time, more overlapping photon information may be ignored by the binary image. Therefore, in order to reduce the measurement error, the first coefficient can be appropriately reduced so that the finally calculated number of photons increases to make up for the error of the ignored overlapping photons.

[0177] S540, obtain the position distribution of the hole pixels in each connected component to generate a second coefficient, where the hole pixels are characterized as foreground pixels that do not all have a connected relationship in the 8-neighborhood.

[0178] The presentation form of photon signals is generally high and dense in the center and low and dispersed around. Then, when there is overlap between photons, due to the position difference between two or more overlapping photons, certain voids will appear in the binary image.

[0179] Then it is necessary to scan and screen the void pixels in the adjacent connected domain. A void pixel is characterized in that not all of the 8-neighborhood directions of a foreground pixel are adjacent to other foreground pixels, but there are some adjacent background pixels. In this case, it is determined as a void pixel.

[0180] Generally speaking, the pixels on the boundary of photon signals are all void pixels. Since the pixel energy at the center position of the photon is relatively high, void pixels generally do not appear at the center position. When multiple photons overlap, if the edge parts of two or more photons overlap, a large number of void pixels will appear in the central part of the connected domain of a region (because the energy of the edge part of the photon is relatively low, which is characterized as relatively sparse foreground pixels at the edge of the photon in the binary image). When the central parts of two or more photons overlap, there will be no void pixels in the central part of the connected domain (because the overlapping part is located at the center positions of several photons, and the energy at the photon center is relatively high, which is characterized as relatively dense foreground images at the photon center in the binary image).

[0181] Therefore, the overlap situation between multiple photons can be intuitively analyzed through the positional distribution relationship of void pixels in a connected domain. If there are a large number of void pixels in the central part of the connected domain, it may be that the edge parts of multiple photons overlap, and the number of binary pixels lost due to the overlap is relatively small. Then the second coefficient can be relatively high to maintain the pixel area of the standard photon object. If there are not a large number of void pixels in the central part of the connected domain, it may be that the central parts of multiple photons overlap, and the number of binary pixels lost due to the overlap is relatively large. Then, in order to avoid excessive loss of pixels that cannot be reflected through binary at the overlapping position, the second coefficient can be appropriately reduced to reduce the pixel area of the photon object so that the finally calculated number of photons increases dynamically.

[0182] S550, obtain the ratio between the sub-area of each connected domain and the overall area of the ultraviolet image to generate a third coefficient.

[0183] Finally, it is also necessary to judge the ratio of the sub-area of each connected domain to the overall area of the ultraviolet image. If the proportion of the sub-area in the overall area is relatively large, it indicates that the proportion of photons in the ultraviolet image is relatively large. On the one hand, it may be determined by the size of the shooting frame, or it may be determined by a relatively high level of discharge, corona, etc.

[0184] In theory, the larger the size of the photon object in the image, the greater the number of photons and the possibility of photon overlap. In order to compensate for the impact of pixel loss caused by overlap, the size of the third coefficient can be appropriately reduced to dynamically increase the final calculated number of photons.

[0185] When the proportion of the sub-area in the overall area is relatively small, which is characterized by the smaller proportion of photons in the ultraviolet image, the number of photons and the possibility of photon overlap are both smaller. At this time, the size of the third coefficient can be increased.

[0186] S560: Adjust the standard area based on the first coefficient, the second coefficient, and the third coefficient to obtain a final pixel area.

[0187] By using the above-mentioned analysis indicators of image information to generate several dynamic coefficients, the standard area of ​​a single photon pixel is dynamically adjusted based on different scenarios and images to optimize the content loss caused by photon overlap. The final calculated photon pixel area can be closer to actual needs based on the actual scene.

[0188] S570, dividing each sub-area by the pixel area to obtain a unit photon number, and adding a number of unit photon numbers to calculate an equivalent photon number.

[0189] Each sub-area is divided by the pixel area to obtain the unit photon number calculated for each connected domain, and the photon numbers corresponding to several connected domains are accumulated to obtain the equivalent photon number contained in the final entire ultraviolet image.

[0190] In some other embodiments, the equivalent photon number is calculated based on the pixel area and the total area, further comprising the following steps:

[0191] S571, marking a connected domain whose sub-area is smaller than the pixel area of ​​the photon object as an object to be verified.

[0192] In some cases, the area of ​​some connected domains may be smaller than the pixel area of ​​a single photon. There are two possibilities for this situation. One is that the weak signal at the boundary of the photon object is mistakenly eliminated during the filtering, binarization, and morphological closing operations, resulting in the area of ​​the binary connected domain presented by the single photon signal being smaller than the pixel area of ​​a normal single photon. The other possibility is that some background noise is not processed correctly and is mistakenly converted into a photon object.

[0193] If these small connected domains are not specially processed, there will be certain errors in the final calculated number of photons, which may be too large or too small.

[0194] Therefore, these small connected domains are first selected and marked as objects to be verified for further analysis and processing.

[0195] S572. Calculate the minimum pixel distance between the object to be verified and other connected regions.

[0196] S573. When the minimum pixel distance is less than the preset value, set the calculated number of photons per unit of the object to be verified to 1.

[0197] S574. When the minimum pixel distance is greater than the preset value, the object to be verified does not participate in the calculation of the number of photons per unit.

[0198] Analyze the minimum pixel distance between the edge of the object to be verified and other connected regions, and this distance represents the minimum distance between the object to be verified and other connected regions.

[0199] When the minimum pixel distance is greater than the preset value, it is considered that the object to be verified is far from other large-area regions with a high probability of being photon objects. In the actual application scenario of an ultraviolet imager, there will not be a situation where a single fault causes only one photon signal and the distance from the large-area photon region corresponding to the actual fault is extremely far, which will result in a low correlation between this small region and other large regions. Therefore, at this time, it is considered that the object to be verified is probably background noise misidentified as a photon object. Then, when calculating the number of photons per unit, the connected domain sub-area corresponding to the object to be verified is not calculated.

[0200] When the minimum pixel distance is less than the preset value, it is very likely that it is actually the photon pixel of the same photon object as the shortest adjacent connected region. It's just that during signal processing, some photon pixels are erroneously eliminated and cannot be connected to other regions. Therefore, at this time, the calculated number of photons per unit less than 1 for the object to be verified can be directly changed to 1, defaulting that it corresponds to one photon number.

[0201] Through the above optimization, the area of photon objects erroneously eliminated or the area of noise objects erroneously identified is effectively analyzed and processed, optimizing the numerical accuracy of the finally calculated equivalent number of photons.

[0202] In some other embodiments, the following steps are further included:

[0203] S600. Obtain the measurement distance of the ultraviolet image and correct the equivalent number of photons based on the attenuation model. Specifically, ,

[0204] where, represents the corrected number of photons, represents the equivalent number of photons, d represents the measurement distance, and D represents the constant parameter of the ultraviolet imager.

[0205] Further correction is performed according to the attenuation characteristics of the photon number with distance. When ultraviolet photons propagate in the atmosphere, as the distance increases, the influence of scattering and absorption will also increase correspondingly, and the calculated photon number will show a certain attenuation law.

[0206] Traditional photon attenuation models usually use a simple inverse-square attenuation law. However, in practical applications, due to factors such as the atmospheric environment, equipment characteristics, and measurement distance, the attenuation of photons is not just a simple inverse-square relationship. Therefore, a more refined correction method is needed.

[0207] Therefore, considering the distance attenuation characteristics, the measured photon number is corrected to improve the accuracy of long-distance measurement.

[0208] Among them, D can be obtained through experimental calibration and is characterized as a distance attenuation parameter. By measuring the photon number at different known distances and using these experimental data for curve fitting, the attenuation correction factor D can be obtained, enabling this correction model to adapt to different environmental and equipment characteristics.

[0209] This model can effectively compensate for the photon number attenuation error caused by distance. Especially in long-distance measurements (such as over 50 meters), the photon number may be significantly underestimated due to attenuation. After correction using a quadratic function, the measured photon number can be closer to the actual photon number.

[0210] Furthermore, when ultraviolet photons propagate in the atmosphere, high humidity and high haze conditions will also exacerbate photon attenuation. Then, according to the relationship between humidity, haze concentration, etc. and photon attenuation, a corresponding humidity correction factor k1 and haze attenuation factor k2 can be established, and the correction formula can be optimized to obtain: ,

[0211] This application also discloses a photon number statistical system for a solar-blind ultraviolet imager to implement the above method.

[0212] Through the above steps, first, a three-stage hybrid denoising architecture is adopted to separate noise and photon signals, and filtering and binarization processing are performed. At the same time, the number of connected components in the binarized image is counted. The area of the connected components is not specified for defense, but the overall area is determined based on connected components of different sizes, and the equivalent photon number is calculated by dividing by the pixel area of a calibrated single photon, ensuring that even if multiple photon signals overlap in the same area, their total contribution will be correctly counted, thus avoiding the problem of underestimating the photon number in traditional methods.

[0213] It should be understood that although the steps in the flowchart of the accompanying drawings are shown sequentially according to the indication of the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless there is a clear indication in this document, there is no strict order restriction for the execution of these steps, and they can be executed in other orders.

[0214] The above are all preferred embodiments of the present application, and do not limit the protection scope of the present application accordingly. Therefore, all equivalent changes made according to the structure, shape, and principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method for counting photon numbers of a solar-blind ultraviolet imager, characterized in that: The following steps are involved: Acquire UV images and perform multi-scale decomposition to separate noise objects from photon objects; filtering and eliminating the noise object in the ultraviolet image and retaining the photon object; Binarizing the filtered ultraviolet image to obtain a binary image; Analyzing the connectivity relationship between each pixel point in the binary image and the adjacent pixel points, and establishing a plurality of connected domains including a plurality of foreground pixels based on the analysis result, wherein the connectivity relationship includes position connectivity and attribute connectivity; The number of the foreground pixels in each of the connected domains is counted to calculate the total area of ​​all the connected domains, the pixel area of ​​a single photon object calibrated in advance is obtained, and the equivalent number of photons is converted based on the pixel area and the total area.

2. The day-blind ultraviolet imager photon number statistical method according to claim 1, wherein Acquire a UV image and perform multi-scale decomposition to separate noise objects from photon objects, including the following steps: The scale factor and the translation factor are obtained, and the ultraviolet image is subjected to wavelet transform in combination with the Haar wavelet function, wherein: By changing the scale factor, low-frequency features and high-frequency features in the ultraviolet image are analyzed to obtain frequency distribution, and by changing the translation factor, the position of the Haar wavelet function on the time axis is controlled to obtain position distribution; generating a high frequency component of a feature point based on the frequency distribution and the position distribution; The feature points corresponding to the high-frequency components whose position distribution is concentrated and whose frequency distribution is stable are defined as the photon objects, and the feature points corresponding to the high-frequency components whose position distribution is discrete and whose frequency distribution is fluctuating are defined as the noise objects.

3. The day-blind ultraviolet imager photon number statistical method according to claim 2, wherein Filtering and eliminating the noise object and retaining the photon object comprises the following steps: After performing the wavelet transform, wavelet coefficients are obtained, high-frequency sub-band coefficients are extracted from the wavelet coefficients, the median absolute deviation of each high-frequency sub-band coefficient is calculated, and the noise standard deviation is calculated in combination with a preset distribution constant; Calculating the total number of pixels of the ultraviolet image, and calculating a dynamic threshold value in combination with the noise standard deviation; Each of the wavelet coefficients is compared with the dynamic threshold, and all the wavelet coefficients that are less than the dynamic threshold are filtered to complete filtering of the ultraviolet image.

4. The day-blind ultraviolet imager photon number statistical method according to claim 3, wherein After calculating the dynamic threshold, the following steps are also included: Performing region segmentation on the ultraviolet image based on the total number of pixels to obtain a plurality of sub-images; Calculating the sub-wavelet coefficients corresponding to each of the sub-images, and statistically analyzing the probability distribution of the sub-wavelet coefficients of each value; Calculating the feature entropy corresponding to each of the sub-images based on the probability distribution, and calculating the average entropy corresponding to a number of the sub-images; A linear mapping function is established based on the characteristic entropy and the average entropy to calculate a threshold adjustment coefficient, and the dynamic threshold is optimized based on the threshold adjustment coefficient.

5. The day-blind ultraviolet imager photon number statistical method according to claim 1, wherein Binarization processing is performed on the filtered ultraviolet image to obtain a binary image, comprising the following steps: Obtaining the pixel intensity of each pixel in the filtered ultraviolet image, and calculating the global mean of the image as a reference value; The pixel point whose pixel intensity is higher than the reference value is assigned a value of 1 to define it as a foreground pixel, and the pixel point whose pixel intensity is lower than the reference value is assigned a value of 1 to define it as a background pixel; The background pixels are eliminated by a morphological closing operation of first dilation and then erosion to obtain the binary image.

6. The method for counting photons of a solar-blind ultraviolet imager according to claim 5, wherein: Analyzing the connectivity relationship between each pixel point in the binary image and the adjacent pixel points, and establishing a plurality of connected domains including a plurality of foreground pixels based on the analysis result, comprising the following steps: Traversing all the pixel points in the binary image and marking the foreground pixels passed through to add labels; When marking, it is determined whether there are adjacent foreground pixels with labels around based on the 8-neighborhood algorithm. If so, the label with the smallest value in the connectivity relationship is selected for marking. If not, a new label is recursively generated based on the maximum value of the labels in the current binary image. A plurality of foreground pixels having the same label together form a connected region.

7. The method for counting photons of a solar-blind ultraviolet imager according to claim 1, wherein: Counting the number of the foreground pixels in each of the connected domains to calculate the total area of ​​all the connected domains, obtaining the pixel area of ​​a single photon object calibrated in advance, and converting the equivalent number of photons based on the pixel area and the total area, comprising the following steps: defining the pre-calibrated pixel area of ​​the single photon object as a standard area; Calculating a sub-area based on the number of foreground pixels contained in each of the connected domains; Obtaining the maximum aspect ratio of each of the connected domains and generating a first coefficient; Obtaining the position distribution of the hole pixels in each of the connected domains in the connected domain to generate a second coefficient, wherein the hole pixels are characterized as foreground pixels that do not all have a connected relationship in the 8-neighborhood; Obtaining a ratio between the sub-area of ​​each of the connected domains and the overall area of ​​the ultraviolet image to generate a third coefficient; Adjusting the standard area based on the first coefficient, the second coefficient and the third coefficient to obtain the final pixel area; Each of the sub-areas is divided by the pixel area to obtain a unit photon number, and a number of the unit photon numbers are added together to calculate the equivalent photon number.

8. The solar-blind ultraviolet imager photon number statistical method according to claim 7, wherein: Calculating the equivalent number of photons based on the pixel area and the total area also includes the following steps: Marking the connected domain whose sub-area is smaller than the pixel area of ​​the photon object as an object to be verified; Calculate the minimum pixel distance between the object to be verified and the other connected domains; When the minimum pixel distance is less than a preset value, the unit photon number calculated for the object to be verified is set to 1; When the minimum pixel distance is greater than a preset value, the object to be verified does not participate in the calculation of the unit photon number.

9. The method for counting photons of a solar-blind ultraviolet imager according to claim 1, wherein: The following steps are also included: The measured distance of the ultraviolet image is obtained, and the equivalent number of photons is corrected based on an attenuation model. Specifically, ， in, Characterized as the corrected photon number, It is characterized by the equivalent photon number, d is characterized by the measurement distance, and D is characterized by the constant parameter of the ultraviolet imager.

10. A photon counting system for solar-blind ultraviolet imagers, characterized in that: Used to implement the method according to any one of claims 1 to 9.

Images