A method for fast retrieval of data containing meteor spectrum information

Abstract

The present application relates to a kind of fast retrieval methods comprising meteor spectrum information data, comprising: step 1. reading meteor spectrum video;Step 2. remove background noise;Step 3. extract meteor signal;Step 4. isolated noise point is filtered using convolution denoising method;Step 5. morphological repair of meteor signal image;Step 6. local annular filtering method denoising;Step 7. meteor signal enhancement based on hough transform;Step 8. efficient noise filtering and signal enhancement are realized by connected region labeling method;Step 9. detect meteor spectrum feature.The present application combines local annular filtering method with image processing and data analysis algorithm, and accurately identifies weak meteor spectrum feature, realizes the automatic screening of spectral signal in meteor data, improves the accuracy of data screening and data processing efficiency.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of data information processing, and is especially suitable for screening of meteor spectrum information data. BACKGROUND

[0002] Traditional meteor data processing methods, especially for extraction and analysis of meteor spectrum information, have a series of problems, mainly reflected in low recognition efficiency, complex processing, large errors and heavy repetitive labor, etc. These problems seriously restrict the efficient use of meteor spectrum information, especially in the case of a large amount of video data and spectrum signal is relatively scarce, which highlights the shortcomings of traditional methods.

[0003] The processing of meteor spectrum mainly includes selecting meteor events with spectrum information from a large amount of video data. However, since the meteor data containing spectrum information is relatively small, most of the data only contains position information or brightness information of light points without spectrum characteristics. Therefore, a large amount of time is consumed for screening and preprocessing when processing such data, the repetitive labor is huge, and manual selection is easily affected by fatigue, which is prone to misjudgment and omission, especially when detecting weak spectrum signals, the error rate is high, which further affects the accuracy of subsequent analysis. At present, the common meteor video screening method still relies on manual selection, and the average screening time of one video is about 1 minute. In the case of a large amount of video data (such as 1500 videos), manual selection needs to spend 3 to 4 days. Moreover, manual selection is not only inefficient, but also has a certain error risk, especially when the spectrum signal is weak, it is easy to miss key data, resulting in damaged data quality. These problems bring many challenges to the subsequent processing and in-depth analysis of data.

[0004] Therefore, it is particularly important to develop a data screening method that can quickly and accurately identify meteor spectrum information. This method should be able to automatically and efficiently identify meteor events containing spectrum information from a large amount of video data, reduce manual intervention, and improve the accuracy and efficiency of screening. This is of great significance to scientific research such as detection of meteors, trajectory analysis, spectrum characteristic research, and analysis of physical properties of meteor bodies, especially in large-scale meteor monitoring systems, its automatic screening capability will greatly improve the efficiency and accuracy of research results. SUMMARY

[0005] In view of the deficiencies of the prior art, a method for quickly retrieving meteor spectrum information data is provided, which can automatically detect and optimize a large number of weak meteor spectrum signals. By combining background subtraction, inter-frame difference, morphological processing, connected region analysis and adaptive noise suppression and other technologies, the sensitivity, signal quality and noise resistance of the meteor spectrum detection are significantly improved, and the false detection rate is effectively reduced. These optimization methods not only improve the efficiency and accuracy of data processing, but also ensure the integrity of the meteor trajectory, providing a reliable data basis for subsequent chemical composition analysis.

[0006] The key of the present application lies in that for weak meteor spectrum signals, a series of optimization steps are adopted, including convolution denoising method, local ring filtering method, Hough transform enhancement and other technologies, combined with advanced image processing methods, to automatically extract data containing spectrum information from a large number of meteor videos quickly and efficiently. This method greatly improves the screening efficiency of meteor data and can effectively avoid the misjudgment and omission in the traditional manual selection process.

[0007] In a first aspect, the present application provides a method for quickly retrieving meteor spectrum information data, comprising the following steps:

[0008] Step 1. Reading meteor spectrum video

[0009] The meteor spectrum video to be retrieved is read.

[0010] Step 2. Remove background noise

[0011] 2.1 Convert the rgb image to a grayscale image;

[0012] 2.2 Calculate the image background using formula (I) exponential moving average method:

[0013] (I)

[0014] Wherein, α is the weighting coefficient, α=0.5; Pic 0 is the image sequence of the meteor video, the size is M * N * n , n is the frame number of the video, M is the video height, and N is the video width.

[0015] In an embodiment of the present application, M =576, N =720.

[0016] 2.3 Remove background noise using formula (II):

[0017] (II)

[0018] 2.4 Smooth the image of each frame using median filter;

[0019] 2.5 Remove time and lens information, i.e. set the pixel values of 550-576 rows of each frame to 0;

[0020] 2.6 Extract the maximum value of each pixel in all frames to obtain the image of complete meteor trail after removing background noise.

[0021] Step 3. Extract meteor signal

[0022] 3.1 Perform frame-by-frame difference detection on the video signal of step 2, and calculate the difference between each frame (except the first frame) and the previous frame.

[0023] 3.2 Determine the preset threshold (T) T diff )

[0024] Set the initial threshold to 0.01, and gradually increase the threshold to process the image frames without meteors, and gradually remove the pixel values below the current threshold until the number of non-zero pixels in the image is reduced to 50, to obtain the preset threshold (T) T diff ).

[0025] 3.3 Determine the position of significant change of meteor signal, and generate a binary signal image: compare the difference between the current frame and the previous frame frame by frame, if the difference exceeds the preset threshold (T) T diff ), it is considered that the signal changes significantly at this position, and the pixel value at this position is set to 1, otherwise it is set to 0, to obtain a complete meteor binary trajectory signal image.

[0026] Step 4. Filter isolated noise points using convolution denoising method

[0027] In the process of extracting meteor trail, video frame difference may introduce a large number of isolated noise points. The present application uses a convolution denoising method to effectively remove these noise while preserving weak meteor signals.

[0028] Specifically, by using a convolution kernel with a size of 20x20, the total number of non-zero pixels in the neighborhood of each pixel position of the complete meteor trail image obtained in step 3 is calculated. Subsequently, only the pixels with more than 3 non-zero pixels in the neighborhood are retained to filter out isolated noise points, thereby ensuring that the remaining signal region has sufficient significance and meets the neighborhood pixel number condition.

[0029] Step 5. Morphological repair of meteor signal image

[0030] In order to improve the quality of meteor signal image, morphological closing operation is used to repair the meteor signal image. The morphological closing operation is an image processing technology combining dilation and erosion, and the operation sequence is dilation first and then erosion, filling small holes and connecting adjacent objects. The dilation refers to restoring the target area, and using a structure element to dilate the image. The erosion refers to eliminating small noise points, and using a structure element to erode the image.

[0031] In the specific repair process, using a disc-shaped structure element can effectively smooth the image boundary, fill small pores and broken areas. After this processing, the meteor signal image is generated. This step significantly improves the image quality, makes the meteor signal, especially the weak signal, more continuous and complete, effectively solves the technical problems of signal loss, distortion, error and the like caused by inaccurate extraction of part of the weak signal. This step uses morphological closing operation to mainly eliminate small noise points, fill small gaps, and maintain the overall shape of the meteor signal, so that the edge is smoother and the influence of nonlinear transformation on subsequent processing is reduced. This step does not introduce Hough transform too early, reduces the loss of local features, and avoids that the short or weak meteor signal is ignored. In addition, in the case of more noise, directly applying Hough transform may detect false straight lines, affecting the effectiveness of subsequent enhancement operation. Therefore, the morphological repair first helps to retain the signal integrity, and provides more accurate input for the Hough transform of step 7.

[0032] Step 6: Local annular filtering denoising

[0033] In order to effectively remove noise while retaining weak meteor signals, the present application proposes a local annular filtering method, which detects signals in the annular region to determine whether to remove noise in the core region, and is applied to the denoising processing of the repaired image in step 5 to achieve precise denoising.

[0034] The local annular filtering method includes generating grid coordinates in the Cartesian coordinate system, creating annular and core region masks, and signal detection and core region updating:

[0035] 6.1 Generate grid coordinates in the Cartesian coordinate system X , Y

[0036] Set the core region radius r 1 and the annular region outer radius r 2, in the discrete Cartesian coordinate system with grid range from - r 2 to r 2, the horizontal and vertical coordinates x i and y j As shown in formula (III-a):

[0037] (I-a)

[0038] Grid X , Y Each element of the grid represents the coordinates of the corresponding position:

[0039] (I-b).

[0040] 6.2 Creating the annular mask and core region mask

[0041] The Euclidean distance d of each point to the center is calculated using formula (IV):

[0042] (IV)

[0043] The annular mask is created using formula (V) Mask ring and the core region mask Mask core :

[0044] (V).

[0045] 6.3 Signal monitoring and updating

[0046] For each pixel point of the image obtained in step 5, a local window is defined W ( x , y ) whose range is a region centered at x,y with a radius of r 2 The window W ( x , y ) is composed of the annular mask Mask ring and the core region mask Mask core For any window, condition (VI) is executed through matching the local window with the mask:

[0047] (VI)

[0048] By detecting the signal intensity of the annular mask Mask ring around each pixel and whether it is less than 20, it is determined whether to set the core region mask Mask core where the pixel is located to zero, that is Mask ring the signal intensity is less than 20, set Mask core to zero, thereby achieving noise removal.

[0049] The denoised image generated by the local annular filtering method of the present application not only effectively removes noise, but also maximally retains weak meteor spectral signals, thereby avoiding the possible deletion error in traditional methods. The advantage of the local annular filtering method lies in its targeted processing of signal changes in the local area, rather than simply uniform processing of the entire image.

[0050] By creating a local annular region and performing signal detection, the present application can accurately identify and retain weak meteor signals in the image, while effectively filtering out irrelevant interference noise. Compared with traditional image denoising methods such as mean filtering and median filtering, the annular filtering method pays more attention to the signal changes of local features in the image, and can process signals more finely. In mean filtering, all pixel points in the image are averaged, which can easily lead to the loss of weak signals; while median filtering can better remove isolated noise, but is less sensitive to weak signals that change continuously. On the contrary, the annular filtering method dynamically adjusts the processing method of the local region to ensure the integrity and accuracy of the signal, and is particularly suitable for processing images containing weak signals.

[0051] Therefore, the local annular filtering method of the present application significantly improves the retention effect of weak meteor signals while denoising, avoids the deletion error of key signals in traditional methods, and improves the accuracy and reliability of signal extraction, which has important application value for accurate analysis of meteor spectral signals.

[0052] Step 7. Meteor signal enhancement based on Hough transform

[0053] The present application uses Hough transform to detect and locate straight lines in the image, combines morphological operations to repair breaks or defects along the straight line direction, uses morphological closing operation to repair and enhance meteor signals, and further improves the continuity and integrity of meteor signals. The specific steps are as follows:

[0054] 7.1 Detecting straight lines: perform Hough transform on the image obtained in step 6 to detect and locate straight lines in it, including their position, direction and length;

[0055] 7.2 Constructing a structure element: according to the direction and length of the detected straight lines, construct a structure element for morphological dilation operation, so that the dilation operation is performed along the straight line direction to enhance the continuity of the straight line;

[0056] 7.3 Morphological closing operation: use a disc-shaped structure element to perform morphological closing operation to further repair and enhance meteor signals;

[0057] 7.4 Repeat processing: to ensure the continuity of the straight line and the repair effect, repeat the above dilation and closing operation steps 3-10 times until the desired effect is achieved.

[0058] The closed operation based on the Hough transform of the application can repair the direction characteristics of the meteor signal, enhance the connectivity along the meteor trajectory direction, effectively enhance the continuity of the meteor signal, repair the breakage or defects along the straight line direction, and improve the accuracy of signal detection. At the same time, the noise inconsistent with the meteor trajectory is removed.

[0059] Step 8. Efficient noise filtering and signal enhancement realized by connected region labeling method

[0060] In the signal enhancement process of step 7, the noise region may be accidentally enlarged, so further denoising is needed to improve the accuracy of the results. The application processes the image of step 7 by the connected region labeling method, and the specific steps are as follows:

[0061] 8.1 Identify all connected regions in the image by the connected region labeling method, and analyze their characteristics one by one;

[0062] 8.2 For each connected region, calculate the sum of the pixel intensity. If the intensity sum is lower than the specified threshold ( T area = 50), set all pixel values in the region to 0 to remove the region that does not meet the condition;

[0063] 8.3 Keep the region with a pixel intensity sum higher than the threshold, so as to ensure that only the effective region with large signal intensity is kept. After this step, the reliability and effectiveness of the signal are further improved.

[0064] Step 9. Detect whether it contains meteor spectral characteristics: including Hough transform to detect straight line, extract meteor signal coordinates, linear fitting step, identify the number of connected regions by connected region labeling method, combine connected region analysis, linear fitting parameters, fitting residual and error to determine the meteor spectral signal. After all the denoising and signal enhancement processing is completed, the detection of whether the video contains meteor spectral signal can be performed. The specific process is as follows:

[0065] 9.1 Search for the positions with pixel value of 1 in the image obtained in step 8, and calculate the number of these positions. If the number of detected pixel points is less than or equal to 100, it means that no meteor signal is detected; if the number is greater than 100, continue with the subsequent processing;

[0066] 9.2 Detect the straight line (i.e. the meteor trajectory) by Hough transform, and crop the image according to the longitudinal position of the straight line to generate a sub-image containing the meteor trajectory region, so as to improve the accuracy of subsequent processing;

[0067] 9.3 Extract the coordinates of the pixel value of 1 in the sub-image containing the meteor trajectory region as the coordinates of the meteor signal, wherein x,y x and​y These are the horizontal and vertical coordinates of the image, respectively;

[0068] 9.4 Using the least squares method to determine the signal pixel coordinates ( x,y Perform linear fitting y = ax + b To obtain the fitting parameters and correlation coefficients r Fitting residuals R and error δ The fitting parameters include a and b. The theoretical fitting value for each data point is calculated using the fitting parameters, and the fitting error / residual is calculated accordingly.

[0069] 9.5 Identify the number of connected components in the cropped image from step 9.2 using the connected component labeling method. N ;

[0070] 9.6 Define the objective function as shown in formula (VII):

[0071] (VII)

[0072] in, , , σ and σ are the fitting residuals, respectively. R The mean and standard deviation; X is the design matrix, X=[1,1,…1;x1,x2,…,x] n ] T x i This represents the horizontal coordinate of the signal pixel, and diag() means taking the diagonal elements of the matrix.

[0073] In one embodiment of the present invention, parameters are set. N min =2, N max =20, r max =0.7, σ min =50, δ min =50.

[0074] Calculate the objective function J To determine whether the video data contains meteor spectral features.

[0075] Only if the objective function J =0 or J When =1, the video data contains meteor spectral features.

[0076] Furthermore, when JWhen the correlation coefficient is 0, it indicates that the video data contains meteor spectral features, and the following conditions must be met simultaneously: ① Number of connected components 2 ≤ N < 20; ② Correlation coefficient r <0.7; ③ The standard deviation of the fitting residual σ≥10; ④ Error δ ≥10, video data may contain zero-order spectral features.

[0077] The zero-order spectrum is the emission trajectory of the meteor itself. Generally, the distance between the spectrum obtained by grating dispersion and the meteor trajectory is relatively large. The zero-order spectrum can be used to determine the direction of the meteor spectrum, facilitating subsequent work to extract meteor spectral information.

[0078] when J When =1, it represents the number of connected components. N =1, although the video data still contains meteor spectral features, it does not contain zero-order spectral features at this point. The specification limit on the number of connected regions is to control noise or non-meteor spectral interference.

[0079] In the process of identifying the spectral features of meteors, this invention constructs a self-designed objective function J, which comprehensively considers factors such as connected component analysis, linear fitting coefficients, fitting residuals, and errors. By combining various image processing techniques, including Hough transform, connected component labeling, and least squares fitting, with statistical characteristic analysis, this determination mechanism can efficiently detect spectral signals, including weak signals. This enables effective and accurate detection of data containing weak spectral signals, solving the technical problem of large errors in weak signal identification.

[0080] Secondly, the method for rapidly retrieving meteor spectral information data provided by this invention can be used in fields such as meteor monitoring, meteor shower research, and solar system formation and evolution research.

[0081] The beneficial effects of this invention are:

[0082] 1. This invention combines local ring filtering with image processing and data analysis algorithms. Through a series of efficient data processing algorithms, it automatically filters out meteor data with spectral information and accurately identifies the spectral characteristics of meteors. This method realizes automatic filtering of spectral signals in meteor data, which not only improves the accuracy of data filtering, but also greatly reduces the time and workload of manual processing. The entire retrieval process takes only 2-6 seconds, which significantly improves the efficiency of data processing.

[0083] 2. The local ring filtering method of the present application can accurately denoise the core area of the image and effectively preserve weak meteor signals. The method has the advantage of being flexible in adapting to local feature changes in the image, avoiding the problem of weak signal loss caused by excessive smoothing in traditional mean filtering and median filtering methods. In addition, by combining Hough transform to detect the straight line features in the image and morphological dilation and closing operation to repair the missing or broken parts along the straight line direction, the integrity and continuity of the meteor signal are significantly improved, thereby ensuring the accurate extraction of weak signals. This technical breakthrough not only has important significance in meteor monitoring, but also provides a reliable technical means for the extraction and analysis of weak spectral signals.

[0084] 3. In the process of identifying the spectral characteristics of meteors, the present application constructs a target function J, which comprehensively considers factors such as connected region analysis, linear fitting coefficient, fitting residual and error. By combining various image processing techniques such as Hough transform, connected region labeling and least squares fitting with statistical characteristic analysis, a high-efficiency determination mechanism is formed, which is helpful for accurate detection of spectral signals.

[0085] 4. The identification method of the present application preserves weak spectral / meteor signals by denoising a small number of times to avoid loss and false deletion of key signals. This method has strong universality for weak signal data and can be applied to different types and sources of meteor data after appropriate adjustment and optimization. At the same time, it has high automation degree and small error, and is suitable for wide promotion and application.

[0086] 5. The present application can be widely applied in the fields of meteor monitoring, meteor shower research, solar system formation and evolution research, etc., providing more efficient and accurate data support for related scientific research and promoting further exploration of meteors and their spectral characteristics in the scientific community. Especially in large-scale meteor monitoring systems and meteor spectral analysis projects, the application of this technology can significantly improve research efficiency and promote the development of meteorology, space physics and other related disciplines. BRIEF DESCRIPTION OF DRAWINGS

[0087] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0088] Fig. 1 The original image of the meteor sample in Example 1, wherein (a) the original image containing complete meteor trajectory; (b) the image containing complete trajectory after removing background noise;

[0089] Fig. 2The image containing the meteor signal is a binary image extracted in Example 1;

[0090] Fig. 3 The meteor trajectory diagram after filtering isolated noise points in Example 1 ( FilteredImage );

[0091] Fig. 4 The image shows the meteor trajectory after morphological restoration in Example 1. RepairedSignalMap );

[0092] Fig. 5 The image after denoising using the circular local circular filtering method in Example 1 ( FilteredSignalMap );

[0093] Fig. 6 The enhanced meteor signal image of Example 1 ( EnhancedSignalMap );

[0094] Fig. 7 The image after denoising the connected regions in Example 1 ( CleanedSignalMap );

[0095] Fig. 8 The results of linear fitting of the meteor signal region in Example 1;

[0096] Fig. 9 The image shows the process and results of spectral information recognition for meteor (20191220_210801) in Example 2. (a) is the original image, (b) is the image after removing background noise, (c) is the image of extracted meteor signal, (d) is the image of noise point filtering, (e) is the image of morphological restoration, (f) is the image of local filtering and denoising, (g) is the image of enhanced meteor signal, (h) is the image of denoising with connected region marking, and (i) is the result of linear fitting of meteor signal region.

[0097] Fig. 10 The image shows the process and results of spectral information recognition for meteor (20210211_165334) in Example 3. (a) is the original image, (b) is the image after removing background noise, (c) is the image of extracted meteor signal, (d) is the image of noise point filtering, (e) is the image of morphological restoration, (f) is the image of local filtering and denoising, (g) is the image of enhanced meteor signal, (h) is the image of denoising with connected region marking, and (i) is the result of linear fitting of meteor signal region.

[0098] Fig. 11Figures of the spectral information recognition process and results of the meteor (20230129_173228) of Example 4, where (a) is the original image, (b) is the image after removing background noise, (c) is the meteor signal extraction image, (d) is the noise point filtering image, (e) is the morphological repair image, (f) is the local filter denoising image, (g) is the enhanced meteor signal image, (h) is the connected region label denoising image, and (i) is the linear fitting result of the meteor signal region.

[0099] Fig. 12 Figures of the spectral information recognition process and results of the meteor (20210912_173638) of Example 5, where (a) is the original image, (b) is the image after removing background noise, (c) is the meteor signal extraction image, (d) is the noise point filtering image, (e) is the morphological repair image, (f) is the local filter denoising image, (g) is the enhanced meteor signal image, (h) is the connected region label denoising image, and (i) is the linear fitting result of the meteor signal region.

[0100] Fig. 13 Figures of the spectral information recognition process and results of the meteor (20220102_153646) of Example 6, where (a) is the original image, (b) is the image after removing background noise, (c) is the meteor signal extraction image, (d) is the noise point filtering image, (e) is the morphological repair image, (f) is the local filter denoising image, (g) is the enhanced meteor signal image, (h) is the connected region label denoising image, and (i) is the linear fitting result of the meteor signal region.

[0101] Fig. 14 Figures of the spectral information recognition process and results of the meteor (20231010_204157) of Example 7, where (a) is the original image, (b) is the image after removing background noise, (c) is the meteor signal extraction image, (d) is the noise point filtering image, (e) is the morphological repair image, (f) is the local filter denoising image, (g) is the enhanced meteor signal image, (h) is the connected region label denoising image, and (i) is the linear fitting result of the meteor signal region. DETAILED DESCRIPTION

[0102] In order to make the objects, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0103] Example 1

[0104] Step 1. Read video data: Read the meteor spectrum video to be retrieved.

[0105] Step 2. Remove background noise interference

[0106] 2.1 Convert the rgb image to a grayscale image;

[0107] 2.2 Calculate the image background using the exponential moving average method:

[0108] (I)

[0109] wherein a is the weighting coefficient, a = 0.5; Pic 0 is the image sequence of the meteor video, with a size of M * N * n ( M = 576, N = 720, n is the frame number of the video).

[0110] 2.3 Subtract the background noise:

[0111] (II)

[0112] 2.4 Smooth each frame image using a median filter with a window size of [2, 2].

[0113] 2.5 Remove the time and lens information, i.e. set the pixel values of the 550-576th row of each frame image to 0. By extracting the maximum value of each pixel point in all frames, the complete meteor trajectory image can be obtained, as shown in FIG. 1.

[0114] Step 3. Extract the meteor signal

[0115] Perform frame-by-frame difference detection on the video signal, and calculate the difference between each frame (except the first frame) and the previous frame. For pixel positions with a difference exceeding a preset threshold (T = 50), T diff mark them as 1, indicating positions with significant signal changes, thereby generating a binary signal image. T diff The determination of T is achieved by gradually increasing the threshold, processing image frames without meteors (such as the tenth frame of the video), and gradually removing pixel values below the current threshold until the number of non-zero pixels in the image decreases to 50. In all processed frames, extract the maximum signal value of each pixel position to generate an image containing the maximum signal intensity MaxSignalMap ( M × N ), as shown in FIG. 2. Fig. 2

[0116] ​Step 4. Use convolutional denoising to filter isolated noise points.

[0117] During meteor trajectory extraction, video frame differencing may introduce numerous isolated noise points. To effectively remove this noise while preserving the faint meteor signal, a convolutional denoising method is employed: a 20×20 convolution kernel is used to calculate an image containing the complete meteor trajectory. MaxSignalMap The total number of non-zero pixels in the neighborhood of each pixel location. Then, only pixels with a non-zero neighborhood count greater than 3 are retained to filter out isolated noise points, thus ensuring that the retained signal region is sufficiently salient and meets the neighborhood pixel count requirement. The image after this denoising process is as follows: Fig. 3 As shown, denoted as FilteredImage .

[0118] Step 5. Morphological restoration of meteor signal images

[0119] To improve meteor signal images FilteredImage The image quality was improved by using morphological closing operations for restoration. Morphological closing is an image processing technique that combines dilation and erosion, with the operation sequence being dilation followed by erosion. The restoration process uses a disk-shaped structuring element with a radius of 6, which effectively smooths image boundaries and fills small gaps and broken areas. The resulting restored image significantly improves image quality, making the meteor signal, especially weak signals, more continuous and complete. Fig. 4 As shown, denoted as RepairedSignalMap .

[0120] Step 6: Local Ring Filtering for Noise Reduction

[0121] To effectively remove noise while preserving the weak meteor signal, a local ring filter method was employed and applied to image restoration. RepairedSignalMap The method involves noise reduction. It detects signals within a ring-shaped region to determine whether noise in the core region needs to be removed, achieving precise noise reduction. Specific steps include generating grid coordinates in Cartesian coordinates, creating masks for the ring and core regions, signal detection, and core region updating, as described below:

[0122] 6.1 Generating grid coordinates in Cartesian coordinate system X , Y

[0123] Set the radius of the core area r 1=2 and the outer radius of the annular region r 2=8, within the grid range from - r 2 to r In a discrete Cartesian coordinate system of 2, the x and y coordinates x i and yj is:

[0124] (III-a)

[0125] Then, each element of the grid X , Y represents the coordinates of the corresponding position as:

[0126] (III-b).

[0127] 6.2 Creating the annular mask and the core region mask

[0128] Calculate the Euclidean distance of each point to the center:

[0129] (IV)

[0130] Create the annular mask Mask ring and the core region mask Mask core :

[0131] (V).

[0132] 6.3 Signal monitoring and updating

[0133] Traverse each pixel point of the image RepairedSignalMap , define a local window W ( x , y ) whose range is the area centered at (x, y) with a radius of r2, and the window W ( x , y ) is composed of the annular mask Mask ring and the core region mask Mask core . Match the local window with the mask, for any window, execute condition (VI):

[0134] (VI)

[0135] This step detects the signal intensity of the annular mask Mask ring around each pixel and whether it is less than 20, and judges whether to set the core region mask Mask core where the pixel is to zero, so as to realize noise removal. After local annular filtering processing, the generated denoising image is as shown in Fig. 5 , denoted as FilteredSignalMapThis method can not only effectively remove noise, but also preserve weak spectral signals to the maximum extent, avoiding accidental deletion.

[0136] Step 7. Meteor signal enhancement based on Hough transform

[0137] To improve the coherence and integrity of meteor signals, Hough transform is used to detect straight lines in the image, and morphological operations are combined to repair breaks or defects along the straight line direction. The specific steps are as follows:

[0138] 7.1 Line detection: For the image FilteredSignalMap Apply the Hough transform to detect and locate the straight lines, including their position, direction, and length;

[0139] 7.2 Structural element construction: Based on the detected straight line direction and length, construct the corresponding structural elements for the morphological dilation operation, so that the dilation operation is carried out along the straight line direction, thereby enhancing the continuity of the straight line;

[0140] 7.3 Morphological Closure Operation: Morphological closure operation is performed using a disk-shaped structuring element with a radius of 8 to further repair and enhance the meteor signal;

[0141] 7.4 Repeat processing: To ensure the continuity of the line and the repair effect, repeat the above expansion and closing operations until the expected effect is achieved.

[0142] The image after this processing is denoted as EnhancedSignalMap ,like Fig. 6 As shown, this method effectively enhances the continuity of meteor signals, repairs breaks or defects along straight lines, and improves the accuracy of signal detection.

[0143] Step 8. Efficient noise filtering and signal enhancement achieved through connected component labeling.

[0144] During signal enhancement, noisy regions may be unintentionally amplified, thus requiring further denoising to improve the accuracy of the results. Connected component labeling is used to denoise the image. EnhancedSignalMap The specific steps for processing are as follows:

[0145] 8.1 Identify all connected regions in an image using the connected region labeling method, and analyze their characteristics one by one;

[0146] 8.2 For each connected region, calculate the sum of its pixel intensities. If the sum of intensities is lower than a specified threshold ( T area If the value is 50, then all pixel values ​​in that region are set to 0 to remove regions that do not meet the criteria.

[0147] 8.3 Only regions with a total pixel intensity higher than a threshold are retained, thus ensuring that only effective regions with high signal strength are retained.

[0148] After this step, the generated denoised image is labeled as CleanedSignalMap ,like Fig. 7 As shown, this further improves the reliability and effectiveness of the signal.

[0149] Step 9. Detect whether spectral features are present.

[0150] After completing all denoising and signal enhancement processing, the presence of spectral signals in the video can be detected. The specific process is as follows:

[0151] 9.1 First, retrieve the image. CleanedSignalMap The system identifies the locations where the pixel value is 1 and counts the number of these locations. If the number of detected pixels is less than or equal to 100, it indicates that no meteor signal was detected; if the number is greater than 100, further processing continues.

[0152] 9.2 The Hough transform is used to detect straight lines (i.e., meteor trails), and the image is cropped according to the vertical position of the straight lines to generate sub-images containing the meteor trail regions. MeteorTrackRegion To improve the accuracy of subsequent processing;

[0153] 9.3 Image Extraction MeteorTrackRegion The coordinates of the pixel value 1 ( x,y ), used as the coordinates of the meteor signal, where x and y These are the horizontal and vertical coordinates of the image, respectively;

[0154] 9.4 Using the least squares method to determine the signal pixel coordinates ( x,y Perform linear fitting y = ax + b To obtain the fitting parameters and correlation coefficients r Fitting residuals R and error δ The fitting parameters include a and b. The theoretical fitting value for each data point is calculated using the fitting parameters, and the fitting error / residual is calculated accordingly.

[0155] 9.5 Image Recognition Using Connected Component Labeling MeteorTrackRegion Number of connected regions N ;

[0156] 9.6 Define the objective function J As shown in formula (VII):

[0157] (VII)

[0158] in, , , and σ are the mean and standard deviation of the fitting residuals, respectively. R , X is the design matrix, X = [1, 1, … 1; x1, x2, …, x n ] T , x i represents the coordinate of the signal pixel in the horizontal direction, diag() represents taking the diagonal elements of the matrix. Set the parameters N min = 2, N max = 20, r max = 0.7, σ min = 50, δ min = 50.

[0159] The fitting results of this embodiment are shown in Fig. 8 J = 0, the video data contains meteor spectral characteristics. Synchronous verification is performed by manual selection, which takes 2-3 days, and the results of the two methods are consistent, verifying the accuracy of the spectral data processing method of this embodiment.

[0160] Embodiment 2

[0161] The meteor video case 20191220_210801 is processed according to steps 1-9 of embodiment 1, and the spectral recognition process and result graph are shown in Fig. 9 , in which (a) is the original image obtained in step 1, (b) is the image obtained after removing background noise in step 2, (c) is the meteor signal extraction image in step 3, (d) is the noise point filtering image obtained in step 4, (e) is the morphological repair image in step 5, (f) is the local filter denoising image in step 6, (g) is the enhanced meteor signal image in step 7, (h) is the connected region marking denoising image in step 8; (i) is the linear fitting result of the meteor signal region in step 9.

[0162] The fitting results of this embodiment show J = 0, the video data contains meteor spectral characteristics, and may contain zero-order spectrum;

[0163] Synchronous verification is performed by manual selection, which takes 2-3 days, and the results of the two methods are consistent, verifying the accuracy of the spectral data processing method of this embodiment.

[0164] Embodiment 3

[0165] The meteor video case 20210211_165334 is processed according to steps 1-9 of embodiment 1, and the spectral recognition process and result graph are shown in Fig. 10 ​​As shown, (a) is the original image obtained in step 1, (b) is the image obtained after removing background noise in step 2, (c) is the meteor signal extraction image in step 3, (d) is the noise point filtered image obtained in step 4, (e) is the morphological restoration image in step 5, (f) is the local filtering and denoising image in step 6, (g) is the enhanced meteor signal image in step 7, (h) is the connected region labeling and denoising image in step 8; and (i) is the linear fitting result of the meteor signal region in step 9.

[0166] The fitting results of this embodiment show J =0, the video data contains meteor spectral features and may contain zero-order spectra.

[0167] Simultaneously, manual selection was used for verification, which takes 2-3 days. The results of the two methods were consistent, verifying the accuracy of the spectral data processing method in this embodiment.

[0168] Example 4

[0169] Following steps 1-9 of Example 1, the meteor video case 20230129_173228 was processed. The spectral recognition process and results are shown in the figure below. Fig. 11 As shown, (a) is the original image obtained in step 1, (b) is the image obtained after removing background noise in step 2, (c) is the meteor signal extraction image in step 3, (d) is the noise point filtered image obtained in step 4, (e) is the morphological restoration image in step 5, (f) is the local filtering and denoising image in step 6, (g) is the enhanced meteor signal image in step 7, (h) is the connected region labeling and denoising image in step 8; and (i) is the linear fitting result of the meteor signal region in step 9.

[0170] The fitting results of this embodiment show J =1, indicating the number of connected components. N =1, although the video data still contains meteor spectral features, it does not contain zero-order spectral features at this time. Simultaneous manual selection was used for verification, which takes 2-3 days. The results of both methods were consistent, verifying the accuracy of the spectral data processing method in this embodiment.

[0171] Example 5

[0172] Following steps 1-9 of Example 1, the meteor video case 20210912_173638 was processed. The spectral recognition process and results are shown in the figure below. Fig. 12The spectral recognition process and result image are shown in FIG. 6, wherein (a) is the original image obtained in step 1, (b) is the image obtained after removing background noise in step 2, (c) is the meteor signal extraction image in step 3, (d) is the noise point filtering image obtained in step 4, (e) is the morphological repair image in step 5, (f) is the local filter denoising image in step 6, (g) is the meteor signal enhancement image in step 7, (h) is the connected region marking denoising image in step 8, and (i) is the linear fitting result of the meteor signal region in step 9.

[0173] The fitting result of this embodiment shows that J=2.73, and the video data does not contain meteor spectral characteristics. Verification is simultaneously performed by manual selection, which takes 2-3 days. The results of the two methods are consistent, verifying the accuracy of the spectral data processing method of this embodiment.

[0174] Embodiment 6

[0175] The meteor video case 20231010_204157 is processed according to steps 1-9 of embodiment 1, and the spectral recognition process and result image are shown in FIG. 8. Fig. 13 The spectral recognition process and result image are shown in FIG. 8, wherein (a) is the original image obtained in step 1, (b) is the image obtained after removing background noise in step 2, (c) is the meteor signal extraction image in step 3, (d) is the noise point filtering image obtained in step 4, (e) is the morphological repair image in step 5, (f) is the local filter denoising image in step 6, (g) is the meteor signal enhancement image in step 7, (h) is the connected region marking denoising image in step 8, and (i) is the linear fitting result of the meteor signal region in step 9.

[0176] The fitting result of this embodiment shows that J=34.23, and the video data does not contain meteor spectral characteristics. Verification is simultaneously performed by manual selection, which takes 2-3 days. The results of the two methods are consistent, verifying the accuracy of the spectral data processing method of this embodiment.

[0177] Embodiment 7

[0178] The meteor video case 20231010_204157 is processed according to steps 1-9 of embodiment 1, and the spectral recognition process and result image are shown in FIG. 8. Fig. 14 The spectral recognition process and result image are shown in FIG. 8, wherein (a) is the original image obtained in step 1, (b) is the image obtained after removing background noise in step 2, (c) is the meteor signal extraction image in step 3, (d) is the noise point filtering image obtained in step 4, (e) is the morphological repair image in step 5, (f) is the local filter denoising image in step 6, (g) is the meteor signal enhancement image in step 7, (h) is the connected region marking denoising image in step 8, and (i) is the linear fitting result of the meteor signal region in step 9.

[0179] The fitting result of the embodiment shows that J=99.98, and the video data does not contain meteor spectral characteristics. The synchronization is verified by manual selection, which takes 2-3 days, and the results of the two methods are consistent, verifying the accuracy of the spectral data processing method of the embodiment.

[0180] In summary, the present application adopts an innovative local annular filtering method, combines image processing and data analysis algorithms, and automatically screens meteor data with spectral information with the help of a series of efficient data processing algorithms, significantly improving the efficiency and accuracy of meteor spectrum screening. The local annular filtering method of the present application can flexibly cope with the local changes of meteor signals, accurately remove the interference noise in the image, and retain the weak meteor spectral signal. Compared with traditional denoising methods, this method retains important signals while avoiding signal loss caused by excessive smoothing, especially for the extraction of weak signals.

[0181] In addition, the present application combines image processing techniques such as Hough transform, morphological operation and connected region analysis, effectively improving the continuity and integrity of meteor tracks. These technical means not only enhance the recognizability of meteor signals, but also reduce the problem of data loss caused by signal breakage or loss. Through these optimizations, the present application can accurately identify meteor signals in complex environments and provide reliable basic data for subsequent scientific analysis.

[0182] It is worth mentioning that the entire retrieval process of the present application only takes 2-6 seconds, which is much faster than the traditional manual screening method, greatly reducing the complexity and labor intensity of manual operation. At the same time, the automatic screening system can eliminate errors in human operation, greatly improving the accuracy of data screening, especially in large-scale data processing, showing more significant advantages. For scientific research institutions and meteor research projects that need to process a large amount of meteor monitoring video or image data, this technology not only improves work efficiency, but also effectively saves manpower and time cost.

[0183] Overall, the application of the present application in the field of meteor spectrum screening not only provides an efficient and accurate solution for meteor monitoring, but also provides strong data support for related scientific research. Its efficient automatic screening, strong signal retention capability and simple operation process make meteor data processing more intelligent, and promote the research process of meteorology and other astronomical fields.

[0184] The above-described embodiments are merely specific embodiments of the present application, which are used to illustrate the technical solutions of the present application, but not to limit the present application. The protection scope of the present application is not limited to this. Although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily think of changes to the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features therein, within the technical scope disclosed by the present application. The modifications, changes or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for fast retrieval of data containing meteoric spectral information, characterized by, Comprising the following steps: Step 1. Read the meteor spectrum video; Step 2. Remove background noise; Step 3. Extract meteor signal; Step 4. Filter isolated noise points using convolution denoising method; Step 5. Repair meteor signal image using morphological closing operation; Step 6. Local ring filter method denoising: including generating grid coordinates in Cartesian coordinate system, creating ring and core area mask, and signal detection and core area updating; 6.1 Generating grid coordinates in Cartesian coordinate system X , Y : Setting the core region radius r 1 and the outer annular region radius r 2, in a discrete Cartesian coordinate system where the horizontal and vertical coordinates r 2 range from r 2 to x i and y j as shown in equation (III-a): (III-a) grid each element of the grid represents the coordinates of the corresponding position as formula (III-b): (III-b) 6.2 Create ring mask and core area mask: Calculate the Euclidean distance d of each point to the center using formula (IV): (IV) Creating a ring mask using equation (V) Mask ring and core area mask Mask core : （V） 6.3 Signal monitoring and updating: Define a local window for each pixel in the image obtained in step 5. W ( x , y Its scope is ( x,y With ) as the center and radius as r 2 area, window W ( x , y ) by ring mask Mask ring and core area mask Mask core Composition; Matching a local window to a mask; for any window, executing condition (VI): (VI) Detecting a ring-shaped mask around each pixel Mask ring The signal intensity and whether it is less than 20 are judged to determine whether the core region mask where the pixel is located is set to zero to achieve noise removal. Mask core The signal intensity and whether it is less than 20 are judged to determine whether the core region mask where the pixel is located is set to zero to achieve noise removal. Step 7. Meteor signal enhancement based on Hough transform: detect and locate straight lines in the image through Hough transform, repair broken or defective areas along the straight line direction through morphological operation, and repair and enhance meteor signal through morphological closing operation; 7.1 Detect straight lines: perform Hough transform on the image obtained in step 6 to detect and locate straight lines in it, including their position, direction and length; 7.2 Construct structure element: according to the direction and length of the detected straight line, construct the structure element of morphological dilation operation, so that the dilation operation is carried out along the straight line direction, and the coherence of the straight line is enhanced; 7.3 Morphological closing operation: use disc-shaped structure element to perform morphological closing operation to further repair and enhance meteor signal; 7.4 Repeat processing: to ensure the coherence of the straight line and the repair effect, repeat the above dilation and closing operation steps 3-10 times until the desired effect is achieved; Step 8. Efficient noise filtering and signal enhancement through connected region labeling method; Step 9. Detect meteor spectrum characteristics: including Hough transform to detect straight lines, extract meteor signal coordinates, linear fitting steps, and identify the number of connected regions through connected region labeling method, combined with connected region analysis, linear fitting parameters, fitting residual and error to determine meteor spectrum signal.

2. The method of claim 1, wherein step 2 comprises the following steps: 2.1 Convert the rgb image to a grayscale image; 2.2 Calculate the image background using formula (I) exponential moving average method: （I） Where α is the weighting coefficient, and Pic0 is the image sequence of the meteor video; 2.3 Subtract the background noise using formula (II): (I) 2.4 Use a median filter to smooth each frame of image; 2.5 Remove time and lens information; 2.6 Extract the maximum value of each pixel point in all frames to obtain the meteor trajectory image with complete trajectory after removing background noise.

3. The method of claim 1, wherein, Step 3 comprises the following steps: 3.1 Perform frame-by-frame difference detection on the video signal of step 2 to calculate the difference between each frame and the previous frame; 3.2 Determine the preset threshold T diff : Set the initial threshold to 0.01, gradually increase the threshold, process the image frame without meteor, gradually remove the pixel value below the current threshold, until the number of non-zero pixels in the image is reduced to 50, get the preset threshold T diff ; 3.3 Determine the position of meteor signal change significantly, generate binary signal image: compare the difference between the current frame and the last frame, if the difference exceeds the preset threshold T diff , consider that the position signal changes significantly, set the position pixel value to 1, otherwise set to 0, get the complete meteor binary trajectory signal image.

4. The method of claim 1, wherein, Step 4. Calculate the total number of non-zero pixels in the neighborhood of each pixel position in the image obtained in step 3 using a convolution kernel of size 20x20, and only keep the pixels with more than 3 non-zero pixels in the neighborhood to filter out isolated noise points.

5. The method of claim 1, wherein, Step 5 uses the operation sequence of dilation followed by erosion to effectively smooth the image boundary using a disc-shaped structure element to fill small pores and broken areas.

6. The method of claim 1, wherein, Step 8 comprises: 8.1 Identify all connected regions in the image by connected region labeling method, and analyze one by one; 8.2 For each connected region, calculate the sum of pixel intensity; if the intensity sum is lower than a specified threshold, set all pixel values of the region to 0, and remove the region that does not meet the condition; 8.3 Keep the region whose pixel intensity sum is higher than the specified threshold.

7. The method of claim 1, wherein, Step 9 includes: 9.1 Retrieve the positions of pixel values equal to 1 in the image obtained in step 8, and calculate the number of said positions: if the number of detected pixel points is less than or equal to 100, it indicates that no meteor signal is detected; if the number is greater than 100, continue with the subsequent processing; 9.2 Detect straight lines using Hough transform, and crop the image according to the longitudinal position of the straight lines to generate a sub-image containing the meteor trajectory region; 9.3 Extracting coordinates of pixels with value 1 in the sub-image containing the area of the meteor trajectory (x, y) x, y as coordinates of the meteor signal, where x and y are the horizontal and vertical coordinates of the image, respectively. 9.4 Linear fitting of signal pixel coordinates (x, y) using least squares method x, y = a + b x y = a + b x ax = a + b x b = a + b x r = a + b x R = a + b x d = a + b x where the fitted parameters include a and b, the fitted theoretical value of each data point is calculated by the fitted parameters, and the fitting error / residual is calculated accordingly; 9.5 Identify the number of connected regions of the cropped image in step 9.2 using connected region labeling N ; 9.6 Calculate the objective function J according to formula (VII) to determine whether the video data contains meteor spectral characteristics: (VII) wherein, , , and are the mean and standard deviation of the fitting residuals R respectively. , X is a design matrix, X = [1, 1,... 1; x1, x2,..., x n ] T , x i represents the coordinate of the signal pixel in the horizontal direction, and diag() represents taking the diagonal elements of the matrix; When the objective function J = 0 or J = 1, the video data contains meteoric spectral features.

Images