A method for extracting radiation features of moving weak targets from high frame rate time series images

Abstract

A method for extracting radiation features of moving weak targets from high-frame-rate time-series images comprises the following steps: performing background estimation using a time-domain filtering method on cached historical frames of images to obtain a background estimation value; performing differential statistics on two adjacent frames of images based on the obtained background estimation value to obtain an average noise level estimation value; performing differential processing between the current frame image and the obtained background estimation value to obtain a target slice differential map, and processing the target slice differential map to determine target spot slice information; performing statistics on target spot grayscale accumulation; calculating target spot radiation shielding information; and calculating target radiation inversion results using the absolute radiation calibration coefficient, target spot grayscale accumulation, and target spot radiation shielding information.

Description

Technical Field

[0001] The invention relates to a method for extracting radiation characteristics of moving weak point targets in high frame rate time sequence images, and belongs to the field of high frame rate time sequence image processing. Background Art

[0002] Commonly used methods for extracting weak target radiation signatures use a historical frame as the background image. The presence of flashes or other moving targets in the background image can lead to significant deviations in the background estimation. Methods that directly differenciate the current frame image from the background image, then use a threshold to select the target spot area and accumulate the radiation in that target area are significantly affected by the threshold. Inaccurate target spot extraction leads to low precision in weak target extraction. Occlusion of the background by small targets is ignored; the smaller the target energy, the greater the impact of background occlusion. Summary of the Invention

[0003] The technical problem solved by the present invention is: in view of the problem that the target is blocked and the background is affected in the weak target radiation feature extraction scheme in the current existing technology, a method for extracting the radiation features of moving weak targets in high frame rate time series images is proposed.

[0004] The present invention solves the above technical problems by the following technical solutions:

[0005] A method for extracting radiation features of moving weak point targets from high frame rate time series images, comprising:

[0006] The background estimation is performed by using a time domain filtering method through the cached historical frames of images to obtain a background estimation value;

[0007] Based on the obtained background estimation value, the average noise level estimation value is obtained by performing difference statistics on two adjacent frames of images;

[0008] Performing differential processing on the current frame image and the obtained background estimation value to obtain a target slice differential map, and processing the target slice differential map to determine target spot slice information;

[0009] Count the grayscale accumulation of target spots;

[0010] Calculate the target spot radiation shielding information;

[0011] The target radiation inversion result is calculated using the absolute radiation calibration coefficient, target spot grayscale accumulation, and target spot radiation shielding information.

[0012] The specific method of using the cached historical image frames to perform background estimation using the time domain filtering method is as follows:

[0013] Taking the current moment as moment t, select the t-Nm frame image to the t-Nm+N frame image as the historical frame image, Nm is the historical frame image count to be cached, N is the time domain filter parameter, and the median filter is used to calculate the median of the image data at each pixel position in each historical frame image as the background estimation value.

[0014] The calculation method for the historical frame image count that needs to be cached is:

[0015] According to the target's motion speed v and the pixel ground resolution GSD, the frame count M required for the target to cross the pixel is calculated as M = GSD / v;

[0016] Measure the diffuse spot diameter D of the target in the historical frame image;

[0017] Calculate the number of historical frame images that need to be cached:

[0018] Nm=M*(D+1)+N.

[0019] The two adjacent frames of images are specifically the image at time t-1 and the image at time t-2;

[0020] The method for determining the estimated value of the average noise level is as follows:

[0021] Divide the image at time t-1 and the image at time t-2 into Np*Np sub-blocks;

[0022] Calculate the standard deviation of each sub-fast data separately;

[0023] The median of all standard deviation data was calculated as an estimate of the average noise level.

[0024] The specific method for obtaining the target slice difference map is:

[0025] With the target center of mass as the center point, double is the diameter, intercept the target slice difference map, and P is the number of target spot pixels.

[0026] The current frame image is the image at time t. Obtain the target slice difference map between the image at time t and the background estimation value. Based on the target slice difference map, with the known target centroid position as the center, obtain the radiation information caused by the target in the centroid position and its surroundings. The specific steps are as follows:

[0027] Target spots are extracted from the target slice difference image, and the grayscale threshold is set to 3 times the estimated value of the camera average noise level;

[0028] Mark the locations in the target slice difference image where the grayscale value is greater than 3 times the estimated value of the camera's average noise level;

[0029] Perform morphological dilation processing on one pixel in the determined position image, and mark the position after morphological dilation processing;

[0030] Only the positions after morphological expansion processing are retained, the grayscale of the remaining positions in the target slice difference image is set to zero, and the modified target spot slice information is saved.

[0031] The specific method for counting the grayscale accumulation of target spots is:

[0032] For the modified target spot slice, the grayscale of the position marked as 1 whose grayscale is less than the camera average noise ns is set to 0, and then the grayscale of the slice information is summed to obtain the target grayscale cumulative value DNS.

[0033] The specific steps for calculating the target spot radiation shielding information are as follows:

[0034] Extract the grayscale information G of the target spot centroid position in the background estimation image;

[0035] Calculate the ratio R of the projected area of ​​the target vertical camera angle of view to the projected area of ​​the earth projected by one pixel;

[0036] The radiation shielding information of the target spot, ie, the radiation shielding grayscale P of the target spot, is calculated according to the formula P=G*R.

[0037] The absolute radiation calibration coefficient is obtained by performing absolute radiation calibration on the camera, and includes a radiation calibration coefficient K and a radiation calibration coefficient C.

[0038] The calculation method of the target radiation inversion result W is:

[0039] W=K*(DNS+P)+C.

[0040] The calculation method of the ratio R of the projected area of ​​the target vertical camera angle of view to the projected area of ​​the earth projected by one pixel is:

[0041] R=MS / GSD 2 .

[0042] The advantages of the present invention compared with the prior art are:

[0043] The present invention provides a method for extracting radiation features of moving weak targets in high-frame-rate time-series images. The method first improves the accuracy of background estimation through a time-domain filtering method, and then obtains high-precision target spot radiation information by reasonably determining the size of the target spot and the radiation information collection method of the target spot. Compared with the existing technology, the accuracy is improved. The method of using 3 times the noise as a threshold for segmentation, morphological expansion of the target area, and extraction of radiation information greater than 1 times the noise improves the accuracy of target spot energy collection, compensates for the energy of background radiation blocked by the target spot, improves the accuracy of target spot energy extraction, and is more efficient in extracting radiation features of weak targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Flowchart of the method for extracting radiation characteristics of moving weak point targets provided by the invention; DETAILED DESCRIPTION

[0045] A method for extracting radiation features of moving weak targets from high-frame-rate time-series images includes background estimation, noise estimation, target spot slice generation, target spot grayscale accumulation, target spot occlusion background compensation, and target spot radiation information inversion. The specific steps of the method are as follows:

[0046] The background estimation is performed by using a time domain filtering method through the cached historical frames of images to obtain a background estimation value;

[0047] Based on the obtained background estimation value, the average noise level estimation value is obtained by performing difference statistics on two adjacent frames of images;

[0048] Performing differential processing on the current frame image and the obtained background estimation value to obtain a target slice differential map, and processing the target slice differential map to determine target spot slice information;

[0049] Count the grayscale accumulation of target spots;

[0050] Calculate the target spot radiation shielding information;

[0051] The target radiation inversion result is calculated using the absolute radiation calibration coefficient, target spot grayscale accumulation, and target spot radiation shielding information.

[0052] The specific method for estimating the background using the time domain filtering method using the cached historical frames is as follows:

[0053] Taking the current moment as time t, select the t-Nm frame image to the t-Nm+N frame image as the historical frame image, Nm is the historical frame image count to be cached, N is the time domain filter parameter, and the median value of the image data at each pixel position in each historical frame image is calculated through the median filter as the background estimation value;

[0054] The calculation method for the historical frame image count that needs to be cached is:

[0055] According to the target's motion speed v and the pixel ground resolution GSD, the frame count M required for the target to cross the pixel is calculated as M = GSD / v;

[0056] Measure the diffuse spot diameter D of the target in the historical frame image;

[0057] Calculate the number of historical frame images that need to be cached:

[0058] Nm=M*(D+1)+N;

[0059] The two adjacent frames of images are specifically the image at time t-1 and the image at time t-2;

[0060] The method for determining the estimated value of the average noise level is as follows:

[0061] Divide the image at time t-1 and the image at time t-2 into Np x Np sub-blocks;

[0062] Calculate the standard deviation of each sub-fast data separately;

[0063] The median of all standard deviation data was calculated as the average noise level estimate;

[0064] The specific method for obtaining the target slice difference map is:

[0065] With the target centroid as the center point, double is the diameter, intercept the target slice difference map;

[0066] The current frame image is the image at time t. Obtain the target slice difference map between the image at time t and the background estimation value. Based on the target slice difference map, with the known target centroid position as the center, obtain the radiation information caused by the target in the centroid position and its surroundings. The specific steps are as follows:

[0067] Target spots are extracted from the target slice difference image, and the grayscale threshold is set to 3 times the estimated value of the camera average noise level;

[0068] Mark the locations in the target slice difference image where the grayscale value is greater than 3 times the estimated value of the camera's average noise level;

[0069] Perform morphological dilation processing on one pixel in the determined position image, and mark the position after morphological dilation processing;

[0070] Only the positions after morphological expansion processing are retained, the grayscale of the remaining positions in the target slice difference image is set to zero, and the modified target spot slice information is saved;

[0071] The specific method for counting the grayscale accumulation of target spots is:

[0072] For the modified target spot slice, the grayscale of the position marked as 1 whose grayscale is less than the camera average noise ns is set to 0, and then the grayscale of the slice information is summed to obtain the target grayscale cumulative value DNS;

[0073] The specific steps for calculating the target spot radiation shielding information are as follows:

[0074] Extract the grayscale information G of the target spot centroid position in the background estimation image;

[0075] Calculate the ratio R of the projected area of ​​the target vertical camera angle of view to the projected area of ​​the earth projected by one pixel;

[0076] Calculate the target spot radiation shielding information according to the formula P=G*R, that is, the target spot radiation shielding grayscale P;

[0077] The absolute radiation calibration coefficient is obtained by performing absolute radiation calibration on the camera, including the radiation calibration coefficient K and the radiation calibration coefficient C;

[0078] The calculation method of the target radiation inversion result W is:

[0079] W=K*(DNS+P)+C;

[0080] The calculation method of the ratio R of the projected area of ​​the target vertical camera angle of view to the projected area of ​​the earth projected by one pixel is:

[0081] R=MS / GSD 2 .

[0082] The following is further described based on specific embodiments:

[0083] In the current embodiment, a method for extracting radiation characteristics of a moving weak point target in a high frame rate time-series image includes the following steps:

[0084] 1) Using the cached historical 11 frames of images, perform temporal median filtering on each pixel to obtain background estimation;

[0085] 2) Difference the two frames at time t-1 and time t-2 to obtain a difference image. Divide the difference image into 8x8 = 64 partitions, each with 128x128 pixels. Calculate the standard deviation of each partition and calculate the median of the standard deviations of the 64 regions to obtain the average camera noise level estimate ns.

[0086] 3) Based on the known target position and target spot area 9, extract a 6x6 region image from the difference image between the current frame image and the background estimated in step 1) as the effective area of ​​the target spot slice. Mark the position with a grayscale greater than 3 times the average camera noise level as 1, then perform morphological dilation of one pixel. The expanded position is also marked as 1, and the other pixels are 0. The grayscale of all positions marked as 0 in the image is set to 0.

[0087] 4) The grayscale of the target spot slice whose grayscale is less than the camera average noise ns is set to 0, and the total grayscale DNS of the target spot slice is calculated;

[0088] 5) Extract the grayscale information G of the target spot centroid position in the background estimation image in step 1), calculate the target vertical camera angle projection area MS and the earth projection area GSD of 1 pixel projection 2 The ratio R = MS / GSD 2 ,,Use the formula P=G*R to obtain the target spot radiation shielding grayscale;

[0089] 6) Using the absolute radiation calibration coefficients K and C, the target spot grayscale sum DNS obtained in step 4), and the target spot radiation shielding grayscale P obtained in step 5), the target radiation inversion result W is obtained using the formula W=K*(DNS+P)+C.

[0090] Although the present invention has been disclosed above in terms of preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art may make possible changes and modifications to the technical solutions of the present invention by using the methods and technical contents disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention shall fall within the scope of protection of the technical solutions of the present invention.

[0091] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A method for extracting radiation features of moving weak point targets from high frame rate time series images, characterized in that include: Perform background estimation using a time domain filtering method using cached historical frames of images to obtain a background estimation value; Based on the obtained background estimation value, the average noise level estimation value is obtained by performing difference statistics on two adjacent frames of images; Performing differential processing on the current frame image and the obtained background estimation value to obtain a target slice differential map, and processing the target slice differential map to determine target spot slice information; Count the grayscale accumulation of target spots; Calculate the radiation shielding information of the target spot; The target radiation inversion result is calculated using the absolute radiation calibration coefficient, the target spot grayscale accumulation, and the target spot radiation shielding information; The specific method of using the cached historical image frames to perform background estimation using the time domain filtering method is as follows: Taking the current moment as time t, select the t-Nm frame image to the t-Nm+N frame image as the historical frame image, Nm is the historical frame image count to be cached, N is the time domain filter parameter, and use the median filter to collect the image data of each pixel position in each historical frame image and calculate the median value as the background estimation value; The two adjacent frames of images are specifically the image at time t-1 and the image at time t-2; The method for determining the estimated value of the average noise level is as follows: Divide the image at time t-1 and the image at time t-2 into Np*Np sub-blocks; Calculate the standard deviation of each sub-block data separately; The median of all standard deviation data was calculated as the average noise level estimate; The current frame image is the image at time t. Obtain the target slice difference map between the image at time t and the background estimation value. Based on the target slice difference map, with the known target centroid position as the center, obtain the radiation information caused by the target in the centroid position and its surroundings. The specific steps are as follows: Target spots are extracted from the target slice difference image, and the grayscale threshold is set to 3 times the estimated value of the camera average noise level; Mark the locations in the target slice difference image where the grayscale value is greater than 3 times the estimated value of the camera's average noise level; Perform morphological dilation processing on one pixel in the determined position image, and mark the position after morphological dilation processing; Only the positions after morphological expansion processing are retained, the grayscale of the remaining positions in the target slice difference image is set to zero, and the modified target spot slice information is saved; The specific method for counting the grayscale accumulation of target spots is: The grayscale of the position marked as 1 whose grayscale is less than the estimated value ns of the camera average noise level is set to 0, and then the grayscale of the slice information is summed to obtain the target grayscale cumulative value DNS; The specific steps for calculating the target spot radiation shielding information are as follows: Extract the grayscale information G of the target spot centroid position in the background estimation image; Calculate the ratio R of the projected area of ​​the target vertical camera angle of view to the projected area of ​​the earth projected by one pixel; The target spot radiation shielding information, ie, the target spot radiation shielding grayscale P1, is calculated according to the formula P1=G*R.

2. The method for extracting radiation characteristics of moving weak targets from high frame rate time series images according to claim 1, characterized in that: The calculation method for the historical frame image count that needs to be cached is: According to the target's motion speed v and the pixel ground resolution GSD, the frame count M required for the target to cross the pixel is calculated as M = GSD / v; Measure the diffuse spot diameter D of the target in the historical frame image; Calculate the number of historical frame images that need to be cached: Nm=M*(D+1)+N.

3. The method for extracting radiation characteristics of moving weak point targets from high frame rate time series images according to claim 2, characterized in that: The specific method for obtaining the target slice difference map is: With the target centroid as the center point, double is the diameter, intercept the target slice difference map, and P is the number of target spot pixels.

4. The method for extracting radiation characteristics of moving weak point targets from high frame rate time series images according to claim 3, characterized in that: The absolute radiation calibration coefficient is obtained by performing absolute radiation calibration on the camera, and includes a radiation calibration coefficient K and a radiation calibration coefficient C.

5. The method for extracting radiation characteristics of moving weak point targets from high frame rate time series images according to claim 4, characterized in that: The calculation method of the target radiation inversion result W is: W=K*(DNS+P)+C.

6. The method for extracting radiation characteristics of moving weak point targets from high frame rate time series images according to claim 5, characterized in that: The calculation method of the ratio R of the projected area of ​​the target vertical camera angle of view to the projected area of ​​the earth projected by one pixel is: R=MS / GSD 2 ; MS is the projected area of ​​the target perpendicular to the camera’s viewing angle, GSD 2 The projected area of ​​the earth projected by 1 pixel.

Images