A weak target detection method based on multi-hypothesis velocity of gyroscope

Abstract

The application discloses a weak small target detection method based on multi-hypothesis velocity of a gyroscope, which comprises the following steps: firstly, based on three-axis information of the gyroscope, image sequences under the same coordinate axis are subjected to multi-velocity hypothesis accumulation and difference processing, background information is eliminated on the difference image, line target detection is carried out, target tailing line information is acquired, the real velocity of the target is determined in combination with the line information, the target position is inversely deduced, and finally, a line correlation contrast algorithm based on the centroid extraction is used for weak small target detection, so that the background is suppressed and the target contrast is improved in numerous backgrounds and clutters. In the weak small target detection process, the three-axis information of the gyroscope is used for multi-hypothesis velocity accumulation, so that false alarms caused by the movement of small targets and background points in the image due to the camera movement can be avoided, the method can be used in the scene of camera shaking, and under the condition of low signal-to-noise ratio, the signal-to-noise ratio of the moving small target is effectively improved, the background clutter and strong clutter are suppressed, the detection accuracy is improved, and the method is particularly suitable for the detection of moving small targets under the condition of low signal-to-noise ratio.

Description

Technical Field

[0001] This invention belongs to the field of weak target detection, specifically relating to a weak target detection method based on a gyroscope with multiple hypothetical velocities. Background Technology

[0002] Infrared small target detection plays a crucial role in infrared imaging systems and is widely used in fields such as early warning, precision guidance, and remote sensing. Infrared search has been a popular research topic for decades. Sometimes, due to application environments, our imaging systems are far from the target. At long distances, the target occupies very little space and the signal-to-noise ratio is extremely low, which brings many challenges to the detection of small targets. Furthermore, due to target movement, small targets are always immersed in complex backgrounds, which are affected by high-intensity clouds, sharp edges, and other interference. Therefore, accurately detecting the location of small targets is a challenging task.

[0003] Currently, the main solutions to this problem fall into two categories: single-frame image detection and multi-frame image detection. Typical examples of single-frame image detection include methods that suppress background through spatial filtering, the Human Visual System (HVS) method, deep learning-based methods, and methods based on sparsity and low-quality reconstruction.

[0004] In situations with low signal-to-noise ratios, single-frame detection generates a large number of false alarms. Currently, most methods employ multi-frame detection, such as multi-level hypothesis detection, dynamic programming, 3D filtering detection algorithms, pipeline filtering, and energy accumulation methods. These methods process multiple consecutive frames of images based on the continuity and regularity of the target's motion in the spatial domain. Dynamic programming and energy accumulation methods directly lead to energy diffusion, making it difficult to detect targets with significant motion and resulting in numerous false alarms. Multi-level hypothesis methods accumulate energy as effectively as possible through assumptions and decisions, but they are often applicable to simple backgrounds and struggle to effectively detect targets embedded in complex backgrounds.

[0005] In summary, existing technologies struggle to detect small, distant targets due to their small size, low radiation intensity, and long detection range. When the target's signal-to-noise ratio (SNR) is lower than the algorithm's detectable SNR, the target may be missed. Traditional denoising or enhancement algorithms also struggle to overcome the energy dispersion problem caused by target motion. Summary of the Invention

[0006] To address the aforementioned problems, the present invention aims to provide a method for detecting weak targets based on multiple hypothetical velocities using a gyroscope.

[0007] The specific technical solution for achieving the objective of this invention is as follows:

[0008] A method for detecting weak targets based on multiple hypothetical velocities using a gyroscope includes the following steps:

[0009] Step 1: Calculate the optical axis model based on the three-axis information of the gyroscope using multiple velocity assumptions. Map the continuous images on the time axis to a coordinate system. Accumulate multiple velocity assumptions for the image sequence under the same coordinate axis to reduce the target speed and improve the target signal-to-noise ratio.

[0010] Step 2: Perform differential processing on the accumulated image, remove background information on the differential image, detect line targets, and obtain target trailing line information;

[0011] Step 3: Extract the line target, calculate the length and slope of the trailing line, and use the line information to calculate the target's true speed.

[0012] Step 4: Utilize the target's actual motion information to perform reverse superposition, reverse the target's position, and suppress strong clutter;

[0013] Step 5: Use a centroid-based line correlation contrast algorithm to detect small targets, suppressing background and improving target contrast amidst numerous backgrounds and clutter.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0015] (1) The solution of the present invention uses the three-axis information of the gyroscope to accumulate multiple hypothetical speeds during the detection of weak targets, which can avoid false alarms caused by the movement of small targets and background points in the image due to camera movement. It can be used in scenes with camera shaking to improve detection accuracy.

[0016] (2) The solution of the present invention can effectively improve the signal-to-noise ratio of moving small targets and suppress background clutter and strong clutter under low signal-to-noise ratio conditions, and is particularly suitable for the detection of moving small targets under low signal-to-noise ratio conditions.

[0017] The present invention will be further described below with reference to specific embodiments. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the weak target detection method based on gyroscope with multiple hypothetical velocities according to the present invention.

[0019] Figure 2 This is a flowchart of the FPGA hardware implementation of the present invention.

[0020] Figure 3 This is a schematic diagram illustrating the implementation effect of each step in the embodiments of the present invention.

[0021] Figure 4 This is a schematic diagram of the convolution operation implemented in the FPGA hardware of the present invention.

[0022] Figure 5 This is the original image containing noise in an embodiment of the present invention.

[0023] Figure 6 This is a cumulative graph showing the assumed speed accumulation in an embodiment of the present invention.

[0024] Figure 7 This is a diagram illustrating the detection effect of weak targets in an embodiment of the present invention, wherein... Figure 7 (a) is the grayscale 3D response map after the target detection algorithm. Figure 7 (b) is a graph showing the detection results of line correlation contrast based on centroid extraction. Detailed Implementation

[0025] Example

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0028] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following drawings denote similar items; therefore, once an item is defined in one drawing, it need not be further discussed in subsequent drawings.

[0029] Combination Figures 1 to 3 A method for detecting weak targets based on multiple hypothetical velocities using a gyroscope includes the following steps:

[0030] Step 1: Calculate the optical axis model based on multiple velocity assumptions using gyroscope three-axis information. Map consecutive images on the time axis to a coordinate system. Accumulate multiple velocity assumptions for the image sequence under the same coordinate axis to reduce the target's velocity and improve its signal-to-noise ratio.

[0031] By acquiring gyroscope information and using the three-axis information data of the gyroscope, optical axis calculation is performed on the image. The camera shake compensation image motion is accurately calculated, thereby restoring a series of shaking image sequences on the time axis to the same coordinate system, so that background clutter is stationary in the image and only the motion of the target itself is retained.

[0032] It was found that the error did not change much in a very short time. Therefore, we considered using a linear model to correct the error. At this time, we considered that the target was also moving slowly. Since it was imaging a small target at a long distance, we could regard the target motion as a slow, uniform linear model. Therefore, for two vectors of linear uniform motion, we added them together and merged them into a colinear uniform motion vector. In the FPGA implementation, the gyroscope information was received through the serial port and unpacked to obtain the three-axis information of the gyroscope for each frame of image. The three-axis information was used to restore the jittered image sequence to the same coordinate system.

[0033] We can use the three-axis rotation of a gyroscope to establish a three-axis matrix. To represent the three-axis rotation of an object, we can simply multiply the three single-axis rotation matrices together.

[0034]

[0035] Among them γ, θ, The rotation angles in the x, y, and z directions can be directly obtained from the gyroscope.

[0036] In the FPGA, gyroscope information is received via serial port and unpacked to obtain the three-axis gyroscope information for each frame of image. An image coordinate system is established using the first frame. Starting from the second frame, convolution is performed on each image to correct optical axis deviations caused by camera shake. A convolution module is established in the FPGA as follows: Figure 4 The diagram shows a convolution operation performed within an FPGA. For the 3*3 convolution mentioned above, two shift ram IP cores are used to buffer two rows of the input image. When the third row of the image is output, it together with the outputs of the two shift ram IP cores to form three rows of parallel image information. Convolution operations are then performed sequentially to restore the image sequence to the same coordinate system.

[0037] Figure 5 As shown, in the original image of the target, the small target is submerged in the background and noise, and is almost indistinguishable.

[0038] For a moving target in an image sequence within the same coordinate system, let the target velocity be Vt, and assume multiple initial velocities Vg. After accumulation, the target's velocity in the image sequence is the combined effect of the target velocity and the assumed initial velocities. Since both Vg and Vt are represented using linear motion models, the combined effect of these two velocities can be calculated using vector superposition.

[0039] V = Vg + Vt

[0040] To accurately calculate the magnitude and direction of the motion vector, we approximate the velocity using multiple hypothetical velocities. Typically, we set the velocity magnitude to the following nine values, represented in the form (x, y), where x is the velocity in the horizontal direction of the image, y is the velocity in the vertical direction, and the unit of velocity is pixels per frame. For example, (0.1, 0.1) indicates that the hypothetical frame moves 0.1 pixels horizontally and vertically respectively.

[0041] (-0.1，-0.1) (-0.1，0) (-0.1，0.1) (0.1，-0.1) (0，0) (0，0.1) (0.1，-0.1) (0.1，0) (0.1，0.1)

[0042] Finally, the image sequence in the unified coordinate system is accumulated pixel by pixel to obtain multiple sets of accumulated images with different assumed initial velocities;

[0043] For FPGAs, speed, in terms of image, corresponds to the direction and magnitude of displacement per frame. Therefore, an image displacement module can be designed as a multi-speed assumption module. For an image, we can divide image displacement into right-downward displacement and left-upward displacement. Right-downward displacement can be seen as the image being output after a delay, with a rightward speed of 1, which can be seen as the image being output after a one-clock delay. Downward speed of 1 can be seen as the image being output after a one-line delay. Another type is left-upward displacement. We can first delay the entire image backward, generate an advance signal from the undelayed image as the image output enable, and then shift the delayed image to the left-upward direction.

[0044] To achieve simultaneous accumulation of displacement in nine directions, this embodiment of the FPGA designs an 18-row DDR read / write arbitration module. The DDR is controlled by a MIG IP core, and the write and read requests for the 18 data channels are combined into an arbitration signal. When WR_REQ is detected as high, the arbitration module outputs a write enable signal, and the DDR performs a write operation. When RD_REQ is detected as high, the arbitration module outputs a read enable signal, and the DDR performs a read operation. This DDR accumulation adopts a partitioned design. Taking an initial velocity of 1 in the x-axis direction and 1 in the y-axis direction as an example, the DDR is divided into two regions: a 100-frame image region for storing a sequence of 100 images, and a 1-frame image position for storing the accumulation image. To save hardware computation time, after the accumulation of frames 1-100 is completed, when calculating the accumulation of frames 2-101, since the portion of frames 2-100 was already included in the first accumulation, the operation of adding the 101st frame minus the 1st frame to the accumulation of frames 1-100 can be performed. This greatly reduces the computational load of the FPGA and helps meet the real-time requirements of the hardware implementation.

[0045] Figure 6 The cumulative graph after the effective accumulation of the target shows that the target's capabilities have been effectively accumulated, and a slight trailing line can be seen.

[0046] Step 2: Perform difference processing on the accumulated image to remove background information from the difference image, detect line targets, and obtain target trailing line information:

[0047] Step 2-1: The image accumulated in Step 1 is easily affected by background information. Therefore, the accumulated image at intervals of m frames is extracted and image difference processing is performed to obtain the initial trailing line region in the difference image, which effectively reduces the interference of the background.

[0048] Because the target is in motion, and after the difference is accumulated, the target will leave a line of alternating light and dark on the difference map, which is the initial trailing line obtained.

[0049] Step 2-2

[0050] The grayscale values ​​of points within a line target are correlated to a certain extent, while the grayscale values ​​of the line target and its surrounding background area differ significantly. This difference can be represented by the grayscale gradient. The grayscale difference between the line target and the surrounding background is large, resulting in a large gradient on the 3D surface. In contrast, the correlation between pixels in the background area is strong, and the grayscale gradient between individual pixels is small, without abrupt changes.

[0051] For each point in the initial trailing line region, determine the difference between the grayscale value of the connected components centered at that point at different angles (0°, 45°, 90°, and 135°) and the grayscale value of the surrounding region:

[0052]

[0053] Where f(x,y) is the original image, g(θ) are the filter templates for 0°, 45°, 90° and 135° respectively, and g(x,y,θ) is the corresponding filtered image;

[0054] Taking 0° as an example, open the convolution window:

[0055]

[0056] All other direction line templates are obtained by rotating the 0° direction template.

[0057] Steps 2-3: Fuse the filtering results of each point in the four directions (0°, 45°, 90°, 135°), take the maximum value after filtering in the four directions as the gradient information of that pixel, and perform binarization segmentation on the resulting gradient image:

[0058]

[0059] Where D(x,y) represents the image after four-directional filtering, and Th represents the binarization threshold;

[0060] For each pixel after four-way filtering, a threshold comparison is performed. Pixels with values ​​greater than the threshold are set to 1, and those with values ​​less than the threshold are set to 0. Pixels with non-zero values ​​are saved, which means that all target trailing lines and suspected targets are processed together for subsequent connected component processing.

[0061] Step 3: Extract the line target, calculate the length and slope of the trailing line, and use the line information to calculate the target's true velocity.

[0062] Step 3-1: Perform connected component processing on the points with a value of 1 in the binarized segmented image, and store the targets under the same connected component. The length of the connected component is the length of the target's trailing line, and the slope of the two endpoints of the connected component is the target's direction of motion.

[0063] In FPGA, connected component processing is implemented using a growth method. A 15x15 convolution kernel is created and grown outward from the center point. When the detected line information is continuous and meets a certain length, the trailing line is considered a line target within the same connected component. The starting coordinates and length information of the trailing line are recorded and stored. This convolution method enables the implementation of connected component algorithms, significantly reducing FPGA computation time and improving algorithm speed.

[0064] Step 3-2: The signal-to-noise ratio of weak targets is low. The energy of moving small targets is dispersed when they are directly superimposed. When the assumed velocity is close to the actual velocity of the target, its energy can be effectively concentrated.

[0065] By comparing the connected component information under different speed assumptions, the group with the largest connected component energy is the group whose assumed speed is closest to the target speed, and the target's true speed is calculated based on the trailing line information.

[0066] Assume the angle between the trailing line and the positive x-direction is α, the angle between the velocity and x-direction is β, and the horizontal length component of the trailing line is L. x The target's true velocity in the horizontal direction is V. xt The target's assumed velocity in the horizontal direction is V. xh When the cumulative frame count is M, we have:

[0067]

[0068] L x =L·cosα#

[0069] L x =(V xt -V xh )·M#

[0070]

[0071] The vertical length component of the trailing line is L. y The target's true velocity in the vertical direction is V. yt ,but:

[0072] L y =L·sinα

[0073]

[0074] Among them, V xt V represents the target's true velocity in the horizontal direction. yt This represents the target's true velocity in the vertical direction.

[0075] Step 4: Using the target's actual motion information, perform reverse superposition to deduce the target's position and suppress strong clutter.

[0076] Based on the target's true velocity determined in step 3, the target's true position is determined by reverse calculation. Then, the calculated target's true velocity is used to replace the previously assumed initial velocity of the target. Multiple frames of images are accumulated, and the point target region is obtained in the accumulated images.

[0077] For a line target, the target may have two directions of motion. By summing the two directions in reverse, we can obtain the correct direction of motion for the point target.

[0078] Step 5: Use a centroid-based line correlation contrast algorithm to detect small targets, suppressing background and improving target contrast amidst numerous background and clutter elements.

[0079] Step 5-1: All target enhancement and noise reduction algorithms ultimately require target detection. Based on the target velocity calculated above, the accumulated image is used to reconstruct the point target. To find the termination condition for this reconstructing process, this paper employs a patch contrast target detection algorithm to detect point targets after energy accumulation. During the target reconstruction process based on the target's motion trajectory, we have essentially determined the target's range. We then perform point target detection on areas within this range that may contain the target.

[0080] This paper employs a target neighborhood contrast detection algorithm. Since the international standard specifies the size of small targets as 2*2-9*9, we open a 9*9 window and slide it across the potential target area. The 9*9 window is divided into a central region where the target exists, and eight neighborhoods in eight directions around the central region. The neighborhood mean is then calculated.

[0081] This paper focuses on long-distance imaging of small targets, where the target size on the camera's target surface is only a few pixels. Therefore, a certain size region (3x3) within the point target region obtained in step 4 is defined as the target's central region, and its eight surrounding 3x3 neighborhoods are defined as the target's background region. The average grayscale value of the central region is:

[0082]

[0083] Where, m T The average gray value within the target area is represented by F, the number of pixels in the center area is represented by N, the gray value at (i,j) is represented by F, and the target area is represented by T.

[0084] The average grayscale value of the target background area is:

[0085]

[0086] Where, m Bk N represents the average grayscale value of the target background region. B Indicates the number of pixels within the background area;

[0087] Calculate the average grayscale values ​​of the background and target regions separately, and determine the intensity difference between the background and target regions based on these values:

[0088] Z(x,y)=minl i

[0089] l1=d1·d9

[0090] l2=d2·d8

[0091] l3=d3·d7

[0092] l4 = d4·d5

[0093]

[0094] To implement the above operations in an FPGA, a 9*9 convolution window is opened, the average gray values ​​of the background region and the target region are calculated separately, the difference is calculated, and the line contrast of the point target is calculated. The line contrast value calculated in the convolution window is used to replace the value of the center pixel in the centroid detection operation.

[0095] Step 5-2: Determine the target centroid based on the intensity difference between the background and target regions. Considering the real-time performance of the algorithm, this paper adopts a sliding window traversal of the image instead of the 8-connected component centroid detection algorithm. A fixed window is used to traverse the image, and the intensity value of the center pixel is compared with the intensity values ​​of its neighboring pixels to determine whether it is the maximum value of the current window. If it is, the pixel is retained; otherwise, it is discarded.

[0096]

[0097] Where U is the convolution region, and maxZ(i,j) is the maximum contrast intensity between each pixel in the convolution region and its surroundings;

[0098] Step 5-3: Determine weak targets in the image based on the determined target centroid and the binarized segmented image:

[0099] CLPCM(x,y) = D(x,y) × C(x,y)

[0100]

[0101] Where T represents the local region of the target trajectory, and Th represents the set threshold.

[0102] Figure 7 (a) is the grayscale 3D response map of the image after centroid extraction. The target is effectively enhanced and the background is weakened. Figure 7 (b) The target is filtered out after binarization, and the background and noise are suppressed.

[0103] This invention also provides a weak target detection system based on multiple hypothetical velocities using a gyroscope, comprising the following steps:

[0104] Multi-velocity accumulation module: used to calculate the optical axis model based on the three-axis information of the gyroscope by multiple assumptions. It maps continuous images on the time axis to a coordinate system, accumulates multiple velocity assumptions on the image sequence under the same coordinate axis, and decelerates the target to improve the target signal-to-noise ratio.

[0105] Trailing line extraction module: used to perform differential processing on the accumulated image, remove background information on the differential image, detect line targets, and obtain target trailing line information;

[0106] The true velocity calculation module is used to extract line targets, calculate the length and slope of the trailing line, and calculate the true velocity of the target using the line information.

[0107] Target position determination module: Used to perform inverse superposition using the target's real motion information to infer the target position and suppress strong clutter;

[0108] Target detection module: Used to detect weak targets by suppressing background and enhancing target contrast amidst numerous backgrounds and clutter by utilizing centroid-based line correlation contrast.

[0109] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0110] Step 1: Calculate the optical axis model based on the three-axis information of the gyroscope using multiple velocity assumptions. Map the continuous images on the time axis to a coordinate system. Accumulate multiple velocity assumptions for the image sequence under the same coordinate axis to reduce the target speed and improve the target signal-to-noise ratio.

[0111] Step 2: Perform differential processing on the accumulated image, remove background information on the differential image, detect line targets, and obtain target trailing line information;

[0112] Step 3: Extract the line target, calculate the length and slope of the trailing line, and use the line information to calculate the target's true speed.

[0113] Step 4: Utilize the target's actual motion information to perform reverse superposition, reverse the target's position, and suppress strong clutter;

[0114] Step 5: Use a centroid-based line correlation contrast algorithm to detect small targets, suppressing background and improving target contrast amidst numerous backgrounds and clutter.

[0115] The present invention also provides a computer-storable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, performs the following steps:

[0116] Step 1: Calculate the optical axis model based on the three-axis information of the gyroscope using multiple velocity assumptions. Map the continuous images on the time axis to a coordinate system. Accumulate multiple velocity assumptions for the image sequence under the same coordinate axis to reduce the target speed and improve the target signal-to-noise ratio.

[0117] Step 2: Perform differential processing on the accumulated image, remove background information on the differential image, detect line targets, and obtain target trailing line information;

[0118] Step 3: Extract the line target, calculate the length and slope of the trailing line, and use the line information to calculate the target's true speed.

[0119] Step 4: Utilize the target's actual motion information to perform reverse superposition, reverse the target's position, and suppress strong clutter;

[0120] Step 5: Use a centroid-based line correlation contrast algorithm to detect small targets, suppressing background and improving target contrast amidst numerous backgrounds and clutter.

[0121] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for detecting weak targets based on multiple hypothetical velocities using a gyroscope, characterized in that, Includes the following steps: Step 1: Calculate the optical axis model based on multiple velocity assumptions using gyroscope three-axis information. Map consecutive images on the time axis to a coordinate system. Accumulate multiple velocity assumptions for the image sequence under the same coordinate axis to reduce the target's velocity and improve its signal-to-noise ratio. By acquiring gyroscope information and using the three-axis information data of the gyroscope, the optical axis of the image is calculated, thereby restoring a series of jittery image sequences on the time axis to the same coordinate system; For a moving target in an image sequence within the same coordinate system, let the target velocity be... And assume multiple initial velocities for it. After accumulation, the velocity of the target in the image sequence is the combined effect of the target velocity and the assumed initial velocity: ； Finally, the image sequence in the unified coordinate system is accumulated pixel by pixel to obtain multiple sets of accumulated images with different assumed initial velocities; Step 2: Perform differential processing on the accumulated image, remove background information on the differential image, detect line targets, and obtain target trailing line information; Step 3: Extract the line target, calculate the length and slope of the trailing line, and use the line information to calculate the target's true speed. Step 4: Utilize the target's actual motion information to perform reverse superposition, reverse the target's position, and suppress strong clutter; Step 5: Use a centroid-based line correlation contrast algorithm to detect small targets, suppressing background and improving target contrast amidst numerous background and clutter elements. Step 5-1: Take a certain size area from the target area obtained in Step 4 as the center area of ​​the target, and define the eight surrounding neighborhoods of the same size as the target background area. The average gray value of the center area is: ； in, This represents the average gray value within the target area. This indicates the number of pixels in the central region. express grayscale values ​​on Indicates the target area; The average grayscale value of the target background area is: ； in, This represents the average grayscale value of the target background area. Indicates the number of pixels within the background area; Calculate the average grayscale values ​​of the background and target regions separately, and determine the intensity difference between the background and target regions based on these values: ； ； ； ； ； ； Step 5-2: Determine the target centroid based on the intensity difference between the background region and the target region: ； in, For convolution regions, This represents the maximum contrast intensity between each pixel within the convolution region and its surroundings. Step 5-3: Determine weak targets in the image based on the determined target centroid and the binarized segmented image: ； ； Where T represents a local region of the target trajectory. The threshold value is set.

2. The weak target detection method based on gyroscope with multiple hypothetical velocities according to claim 1, characterized in that, The acquisition of target trailing line information in step 2 specifically involves: Step 2-1: For the image accumulated in Step 1, extract the accumulated images at intervals of m frames and perform image difference processing to obtain the initial trailing line region in the difference image. Step 2-2: For each point in the initial trailing line region, determine the center point of the line. , , , The difference between the gray values ​​of connected components in different directions and the gray values ​​of the surrounding areas: ； in, This is the original image. They are respectively , , , The filter template, The filtered image is shown below. Steps 2-3: For each point , , , The filtering results from the four directions are fused, and the maximum value after filtering in the four directions is taken as the gradient information of the pixel. The resulting gradient image is then binarized and segmented. ； in, This represents the image after four-way filtering. The threshold representing binarization; For each pixel after four-way filtering, a threshold comparison is performed. Pixels with values ​​greater than the threshold are set to 1, and those with values ​​less than the threshold are set to 0. Pixels with non-zero values ​​are saved, which means that all target trailing lines and suspected targets are processed together for subsequent connected component processing.

3. The weak target detection method based on gyroscope with multiple hypothetical velocities according to claim 2, characterized in that, The calculation of the target's true velocity using line information in step 3 specifically involves: Step 3-1: Perform connected component processing on the points with a value of 1 in the binarized segmented image, and store the targets under the same connected component. The length of the connected component is the length of the target's trailing line, and the slope of the two endpoints of the connected component is the target's direction of motion. Step 3-2: Compare the connected component information under different speed assumptions. The group with the largest connected component energy is the group whose assumed speed is closest to the target speed. Then, calculate the target's true speed based on the trailing line information. Assuming the trailing line and The included angle in the positive direction is Assuming speed and The included angle is The horizontal length component of the trailing line is The target's true speed in the horizontal direction is The target's assumed velocity in the horizontal direction is When the cumulative frame count is Sometimes: ； ； ； ； The vertical length component of the trailing line is The target's true velocity in the vertical direction is ,but: ； ； in, This represents the target's true velocity in the horizontal direction. This represents the target's true velocity in the vertical direction.

4. The weak target detection method based on gyroscope with multiple hypothetical velocities according to claim 2, characterized in that, Step 4, which involves using the target's actual motion information for reverse superposition to infer the target's position, specifically involves: Based on the target's true velocity determined in step 3, the target's true position is determined by reverse calculation. Then, the calculated target's true velocity is used to replace the previously assumed initial velocity of the target. Multiple frames of images are accumulated, and the point target region is obtained from the accumulated images.

5. A weak target detection system based on a gyroscope with multiple hypothetical velocities, characterized in that, Includes the following steps: Multi-velocity accumulation module: Used for calculating optical axis models based on multiple velocity assumptions using gyroscope three-axis information. It maps continuous images on the time axis to a coordinate system, accumulates multiple velocity assumptions for image sequences under the same coordinate axis, and reduces the target's velocity to improve the target's signal-to-noise ratio. By acquiring gyroscope information and using the three-axis information data of the gyroscope, the optical axis of the image is calculated, thereby restoring a series of jittery image sequences on the time axis to the same coordinate system; For a moving target in an image sequence within the same coordinate system, let the target velocity be... And assume multiple initial velocities for it. After accumulation, the velocity of the target in the image sequence is the combined effect of the target velocity and the assumed initial velocity: ； Finally, the image sequence in the unified coordinate system is accumulated pixel by pixel to obtain multiple sets of accumulated images with different assumed initial velocities; Trailing line extraction module: used to perform differential processing on the accumulated image, remove background information on the differential image, detect line targets, and obtain target trailing line information; The true velocity calculation module is used to extract line targets, calculate the length and slope of the trailing line, and calculate the true velocity of the target using the line information. Target position determination module: Used to perform inverse superposition using the target's real motion information to infer the target position and suppress strong clutter; Target detection module: Used for detection based on centroid-derived line-correlated contrast to suppress background and enhance target contrast amidst numerous backgrounds and clutter, enabling the detection of weak targets. A region of a certain size within the acquired target area is designated as the target's central region, and its eight surrounding neighborhoods of the same size are defined as the target's background region. The average grayscale value of the central region is: ； in, This represents the average gray value within the target area. This indicates the number of pixels in the central region. express grayscale values ​​on Indicates the target area; The average grayscale value of the target background area is: ； in, This represents the average grayscale value of the target background area. Indicates the number of pixels within the background area; Calculate the average grayscale values ​​of the background and target regions separately, and determine the intensity difference between the background and target regions based on these values: ； ； ； ； ； ； Determine the target centroid based on the intensity difference between the background region and the target region: ； in, For convolution regions, This represents the maximum contrast intensity between each pixel within the convolution region and its surroundings. Identify weak targets in the image based on the determined target centroid and the binarized segmented image: ； ； Where T represents a local region of the target trajectory. The threshold value is set.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-4.

7. A computer-storable medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-4.

Images