An intelligent deformation small target recognition algorithm for aliasing images based on detector multiplexing

Abstract

The present application relates to a kind of based on detector multiplexing alias image intelligent morphing small target identification algorithm, it is related to alias image target identification technical field, it includes the following steps: design detector multiplexing optical system;Design field of view encoding element, the integral system is spliced;Simulation generates the motion trajectory of early warning target under space-based background;Make target motion trajectory pass through detector multiplexing optical system imaging;Aliasing video simulation of moving target imaging;Aliasing image is applied to intelligent morphing small target identification algorithm, identifies moving target;The correlation of interframe trajectory and light spot shape of moving target is used to realize the resolution and field of view positioning of target.The present application is based on detector multiplexing alias image intelligent morphing small target identification algorithm, the light originally imaged in large array detector is folded and imaged on small array detector, to achieve the purpose of small array detector receiving large field of view imaging.

Description

Technical Field

[0001] This invention relates to the field of target recognition technology for aliased images based on detector multiplexing optical systems, and particularly to an intelligent deformable small target recognition algorithm for aliased images based on detector multiplexing. Background Technology

[0002] Detector multiplexing is a novel space optical early warning payload technology. It utilizes the optical path refraction effect of detector multiplexing components, allowing a single detector to acquire images that would otherwise require multiple detectors. Partition coding elements (cylindrical mirrors) modulate and encode the light rays from different field-of-view areas (i.e., those acquired by different detectors), achieving target differentiation in aliased images. Through an intelligent deformable small target recognition algorithm, the correlation between the inter-frame trajectory of moving targets and the shape of the light spot is utilized to ultimately achieve target resolution and field-of-view localization within the detector multiplexing structure. This leverages the advantages of small-target detectors—low power consumption and low cost—to achieve wide-swath detection for space-based early warning. The detector multiplexing principle is as follows: Figure 1 As shown, the optical system structure for detector multiplexing is as follows: Figure 2 As shown.

[0003] like Figure 3 As shown, the large field of view of the infrared early warning payload is a key factor for successful real-time monitoring and early warning. This requires the infrared detector to be an ultra-large-scale array or ultra-long linear array detector assembly. However, due to limitations in detector materials, research and development level, manufacturing cost and other factors, the scale of a single array or linear array detector is far from meeting the application requirements.

[0004] Inner field of view stitching, also known as image-side field of view stitching, involves stitching multiple imaging devices into an equivalent large-size imaging device within the system in a certain way, thereby achieving large field of view imaging. However, the stitching accuracy requirements of each imaging device in the inner field of view stitching system are high, and the technology is complex to implement. In particular, for infrared cooled detectors, it directly increases the difficulty of temperature control and technical costs.

[0005] External field-of-view stitching, also known as object-side field-of-view stitching, involves dividing the object-side field of view into multiple sub-fields for imaging, and then stitching the images together to obtain a complete large field-of-view image. Examples include common multi-camera arrays, oscillating mirror scanning, or whole-satellite scanning imaging techniques. However, this type of stitching increases the size and weight of the system and places higher precision requirements on the motion control mechanism. Summary of the Invention

[0006] This invention aims to solve the technical problem in the prior art that small array detectors cannot be used to receive large field-of-view imaging, and provides an intelligent deformable small target recognition algorithm based on detector multiplexing of aliased images.

[0007] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0008] A smart deformable small target recognition algorithm based on detector multiplexing in aliased images includes the following steps:

[0009] Step 1: Design the detector multiplexing optical system;

[0010] Step 2: Design the field-of-view encoding element and assemble the overall system; this field-of-view encoding element only has focusing capability in the main axis direction;

[0011] Step 3: Simulate and generate the motion trajectory of the early warning target under a space-based background;

[0012] Step 4: Simulation of aliased video imaging of moving targets;

[0013] Step 5: Apply the intelligent deformable small target recognition algorithm to the aliased image to identify moving targets;

[0014] Step 6: Utilize the correlation between the inter-frame trajectory of the moving target and the shape of the light spot to achieve target discrimination and field of view localization.

[0015] In the above technical solution, when the principal axis direction of the field-of-view encoding element in step 2 is the y-direction, the change in the field of view in the x-direction introduces the tilted incident angle θ. x This causes the equivalent optical axis of the field-of-view encoding element to deflect, resulting in a rotation of the focal line in the xy plane. The rotation angle φ satisfies:

[0016]

[0017] in:

[0018] f y The focal length of the field-of-view encoding element;

[0019] z: Distance from the field-coding element to the final imaging surface;

[0020] Δx provides supplementary meaning;

[0021] Δx: Linear variation of the target's field of view in the x-direction;

[0022] As the field of view angle in the y-direction increases, the degree of beam focusing in the y-direction changes, causing a change in the size of the beam spot, satisfying the following:

[0023]

[0024] in:

[0025] w y : The radius of the light spot in the y-direction;

[0026] w 0,y : Initial beam radius of the incident beam in the y-direction;

[0027] NA: Numerical aperture;

[0028] λ: wavelength of light;

[0029] θ y : Field of view in the y direction.

[0030] In the above technical solution, the field-of-view encoding element is a cylindrical lens.

[0031] In the above technical solution, the intelligent deformable small target recognition algorithm in step 7 is: optical flow method or background modeling method.

[0032] The present invention has the following beneficial effects:

[0033] The intelligent deformable small target recognition algorithm based on detector multiplexing of the present invention folds the light rays that were originally imaged on a large array detector onto a small array detector, thereby achieving the purpose of receiving large field-of-view imaging on a small array detector.

[0034] The intelligent deformable small target recognition algorithm based on detector multiplexing of aliased images of the present invention modulates the light in different fields of view (i.e., those collected by different detectors) by using a partition coding element (cylindrical mirror) to deform the light spots in different fields of view, thereby achieving the marking of light in different fields of view.

[0035] The present invention provides an intelligent deformable small target recognition algorithm for aliased images based on detector multiplexing. By using an intelligent small target recognition algorithm, it can identify moving targets in aliased images, determine the field of view localization of deformable small targets by using inter-frame correlation, and finally realize early warning of small targets in aliased images under detector multiplexing.

[0036] The intelligent deformable small target recognition algorithm based on detector reuse of aliased images in this invention makes up for the shortcomings of existing large field-of-view imaging methods such as oscillating scanning and detector stitching, which are large in size and weight and lack real-time performance. It achieves wide-swath detection with the advantages of low power consumption and low cost of a single target detector. Attached Figure Description

[0037] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0038] Figure 1 This is a schematic diagram illustrating the detector multiplexing principle.

[0039] Figure 2 This is a schematic diagram of the detector multiplexing optical system.

[0040] Figure 3 This is a schematic diagram illustrating the technical characteristics of traditional infrared system large field-of-view imaging techniques.

[0041] Figure 4This is a schematic diagram illustrating the principle of the intelligent deformable small target recognition algorithm for aliased images based on detector multiplexing according to the present invention.

[0042] Figure 5 This is a simulation diagram of the intelligent deformable small target recognition algorithm based on detector multiplexing in aliased images according to the present invention. Detailed Implementation

[0043] The present invention will now be described in detail with reference to the accompanying drawings.

[0044] I. Design of the field-of-view coding element, i.e., the cylindrical mirror

[0045] The cylindrical mirror is designed to have focusing capability in the y-direction but not in the x-direction. Therefore:

[0046] When the field of view changes in the x-direction, the beam remains parallel in the x-direction, the position of the focal line of the light spot will shift, but the size of the light spot remains unchanged.

[0047] When the field of view changes in the y-direction, the beam will focus in the y-direction, the focal point will drift in the y-direction, and the size of the beam spot will change.

[0048] A cylindrical mirror focuses light in the y-direction but not in the x-direction. Therefore, changes in the field of view in the x-direction cause the focal line to rotate. This change in the field of view in the x-direction introduces a tilted incident angle θ. x This causes the equivalent optical axis of the cylindrical mirror to deflect, resulting in a rotation of the focal line in the xy plane. The rotation angle φ can be determined by the following formula:

[0049]

[0050] f y The focal length of a cylindrical mirror.

[0051] z: The distance from the cylindrical mirror to the final image plane.

[0052] Δx: Linear change of the target's field of view in the x-direction.

[0053] Changes in the field of view in the y-direction are equivalent to the beam tilting at different angles θ. y When light is incident on a cylindrical mirror, the focal point of the cylindrical mirror has a converging ability in the y-direction. Changes in the field of view cause changes in the converging quality of the focused beam, thus affecting the length and shape of the beam spot. Since the cylindrical mirror has no focusing effect in the x-direction, the beam spot maintains its original size in the x-direction. As the field of view increases in the y-direction, the degree of focusing of the beam in the y-direction changes, causing a change in the size of the beam spot. This change can be described by the Gaussian beam focusing formula:

[0054]

[0055] wy : The radius of the light spot in the y direction.

[0056] w 0,y : Initial beam radius of the incident beam in the y-direction.

[0057] NA: Numerical aperture.

[0058] λ: wavelength of light.

[0059] θ y : Field of view in the y direction.

[0060] II. Intelligent Deformable Small Target Recognition Algorithm

[0061] In a series of images, the motion of each pixel in the scene changes over time. Assuming that each point on the surface of an object undergoes a translation in time between two adjacent frames, the motion (velocity) of these pixels can be calculated to obtain the motion information of the object in the image. Then, by performing denoising, area filtering, and morphological processing on the foreground target, the recognition of small moving targets against a space-based background can be achieved. Based on this, the shape features of the moving object (the principal axis angle of the line spot and its y-direction dimension) are captured, and the inter-frame change trend of the object's shape features is determined. The unique correlation between the change trend of the shape features and the target's motion trajectory is used to achieve the field-of-view localization of deformable small targets.

[0062] Optical flow equation:

[0063] In a simple case, optical flow can be described by the following equation:

[0064] I x u+I y v+I t =0 (4)

[0065] in:

[0066] I x and I y It is the spatial gradient of the image in the x and y directions.

[0067] u and v are the x and y components of optical flow, representing the velocity of a pixel on the image plane.

[0068] I t It represents the change of an image over time, indicating the change in image brightness over time.

[0069] This equation shows that the temporal variation of brightness in an image is closely related to the spatial gradient and pixel motion.

[0070] The intelligent deformable small target recognition algorithm includes the following steps: modulating and encoding light spots in different fields of view using cylindrical mirrors to establish a unique relationship between the target's trajectory and the changing trend of the light spot shape (the size of the light spot in the y-direction and the rotation angle of the focal line); after imaging by the detector multiplexing infrared early warning system, the intelligent deformable small target recognition algorithm is used to identify targets on the aliased image, thereby achieving the resolution of deformable small targets; and then the field of view is located by using the one-to-one correspondence between the trajectory and the changing trend of the light spot shape.

[0071] like Figure 4 In summary, the intelligent deformable small target recognition algorithm for aliased images based on detector reuse of this invention realizes the stitching and simulation of the detector reuse infrared early warning system and the field-of-view coding element (cylindrical mirror), and completes the deformable small target recognition of the aliased image after field-of-view partition coding under detector reuse. The specific steps are as follows:

[0072] Step 1: Design the detector multiplexing optical system;

[0073] To achieve field-of-view modulation coding at the primary image plane, and to minimize the system aperture while maintaining 100% cold-stop efficiency, a transmissive optical path structure for secondary imaging was selected. Based on this, the front and rear groups of the space-based infrared secondary imaging optical system were designed separately. During the design process, key variables were adjusted to ensure that the lens sizes of the front and rear groups were nearly identical. After the design was completed, the front and rear groups were stitched together. Finally, a reflector assembly was added at the system's back intercept to fold the optical path, enabling a small detector to receive images with a large field of view.

[0074] Step 2: Design the field-of-view encoding elements and assemble the overall system;

[0075] The cylindrical mirror, which serves as the field-of-view encoding element, has its principal axis in the y-direction, meaning it has a focusing effect in the y-direction but not in the x-direction. Therefore, changes in the field of view of the light spot in the x-direction will cause the focal line to rotate.

[0076] The change in the field of view in the x-direction introduces the tilted incident angle θ. x This causes the equivalent optical axis of the cylindrical mirror to deflect, resulting in a rotation of the focal line in the xy plane. The rotation angle φ can be determined by the following formula:

[0077]

[0078] f y The focal length of a cylindrical mirror.

[0079] z: The distance from the cylindrical mirror to the final image plane.

[0080] Δx: Linear change of the target's field of view in the x-direction.

[0081] Changes in the field of view in the y-direction are equivalent to the beam tilting at different angles θ. y When light is incident on a cylindrical mirror, the focal point of the cylindrical mirror has a converging ability in the y-direction. Changes in the field of view cause changes in the converging quality of the focused beam, thus affecting the length and shape of the beam spot. Since the cylindrical mirror has no focusing effect in the x-direction, the beam spot maintains its original size in the x-direction. As the field of view increases in the y-direction, the degree of focusing of the beam in the y-direction changes, causing a change in the size of the beam spot. This change can be described by the Gaussian beam focusing formula:

[0082]

[0083] w y : The radius of the light spot in the y direction.

[0084] w 0,y : Initial beam radius of the incident beam in the y-direction.

[0085] NA: Numerical aperture.

[0086] λ: wavelength of light.

[0087] θ y : Field of view in the y direction.

[0088] In summary, the geometric changes in the light spot under aberration not only exhibit stable directionality and measurability, but also have a clear correspondence with the target's two-dimensional field of view position. Based on this, appropriate cylindrical mirror parameters can be selected, and the cylindrical mirror can be inserted into the primary image plane of the system to form the overall system. By establishing the modulation mapping function of the cylindrical mirror, a highly robust target encoding mechanism can be constructed. This enables the spatial field of view reconstruction and identification of aliased targets.

[0089] Step 3: Simulate and generate the motion trajectory of the early warning target under a space-based background;

[0090] Simulation calculations show that the missile exhaust radiation power received by the space-based infrared early warning system after spectral integration, space propagation, and atmospheric attenuation is approximately 2.920 e^(-24) W. Similarly, the background radiation power received by the space-based infrared early warning system from ground objects is 8.349 e^(-26) W, and the radiation power received from low-Earth orbit satellites is 1.093 e^(-26) W.

[0091] An arc-shaped contour line is used to simulate the Earth's boundary, and a portion of the image is segmented as Earth's background noise and converted to grayscale. The brightness of the target, Earth's background noise, and other low-orbit satellite noise is set with reference to the calculated radiation power ratio. Then, the trajectory of the missile contrail is introduced to generate the trajectory of the early warning target against the space-based background.

[0092] Step 4: Simulation of aliased video for imaging moving targets

[0093] The target's motion trajectory is imaged by a secondary imaging optical system and processed by a cylindrical mirror convolution kernel to obtain a trajectory-encoded video after the target's motion trajectory imaging video is processed by a field-of-view encoding element. The trajectory-encoded video is then processed by image splitting and flipping to obtain a superimposed aliased video.

[0094] Step 5: Apply the intelligent deformable small target recognition algorithm to the aliased image to identify moving targets;

[0095] First, noise is removed from the dynamic video of the missile tail trajectory collected by the detector. After denoising, Gaussian modeling (GMM) is used to identify small targets in the moving missile tail trajectory. During identification, the feature spot in each frame image is analyzed. The bbox module is used to determine the position of the spot centroid, the main axis direction of the spot focal line, the pixel size occupied by the spot, and the size in the y direction. The inter-frame motion trend of the target and the change trend of the shape features of the spot are recorded.

[0096] Step 6: Utilize the correlation between the inter-frame trajectory of the moving target and the shape of the light spot to achieve target discrimination and field of view localization;

[0097] The inter-frame motion trend of the target can be determined by the change in the position of the spot centroid, such as the horizontal component of the spot centroid displacement (left or right) and the vertical component of the spot centroid displacement (up or down). The change trend of the spot shape features can be determined by the change in the principal axis direction of the spot focal line, the pixel size occupied by the spot, and the size of the spot in the y-direction. For example, if the spot focal line rotates clockwise or counterclockwise, the size of the spot in the y-direction increases or decreases. This establishes a connection between the target's motion trend and the change trend of the spot's shape features. Based on the change in the spot shape features corresponding to a certain motion trajectory, the field of view of the target can be uniquely determined, realizing object space field of view discrimination based on the correlation between the target's inter-frame motion trajectory and the change trend of the spot shape.

[0098] like Figure 5 As shown, observing the changes between video frames reveals that the horizontal displacement of the trajectory target is to the right and the trend of the light spot change is that the focal line of the light spot rotates clockwise. The vertical displacement is downward and the trend of the light spot change is that the size of the light spot in the y-direction decreases. Therefore, based on the intelligent small target recognition algorithm, it can be determined that the target comes from the lower field of view of the object.

[0099] The intelligent deformable small target recognition algorithm based on detector multiplexing of the present invention folds the light rays that were originally imaged on a large array detector onto a small array detector, thereby achieving the purpose of receiving large field-of-view imaging on a small array detector.

[0100] The intelligent deformable small target recognition algorithm based on detector multiplexing of aliased images of the present invention modulates the light in different fields of view (i.e., those collected by different detectors) by using a partition coding element (cylindrical mirror) to deform the light spots in different fields of view, thereby achieving the marking of light in different fields of view.

[0101] The present invention provides an intelligent deformable small target recognition algorithm for aliased images based on detector multiplexing. By using an intelligent small target recognition algorithm, it can identify moving targets in aliased images, determine the field of view localization of deformable small targets by using inter-frame correlation, and finally realize early warning of small targets in aliased images under detector multiplexing.

[0102] The intelligent deformable small target recognition algorithm based on detector reuse of aliased images in this invention makes up for the shortcomings of existing large field-of-view imaging methods such as oscillating scanning and detector stitching, which are large in size and weight and lack real-time performance. It achieves wide-swath detection with the advantages of low power consumption and low cost of a single target detector.

[0103] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A smart deformable small target recognition algorithm based on detector multiplexing in aliased images, characterized in that, Includes the following steps: Step 1: Design the detector multiplexing optical system; Step 2: Design the field-of-view encoding elements and assemble the overall system; This field-of-view encoding element has focusing capability only in the principal axis direction; The field-of-view encoding element is a cylindrical lens; Step 3: Simulate and generate the motion trajectory of the early warning target under a space-based background; Step 4: Simulation of aliased video for imaging moving targets; Step 5: Apply an intelligent deformable small target recognition algorithm to the aliased image to identify moving targets; Step 6: Utilize the correlation between the inter-frame trajectory of the moving target and the shape of the light spot to achieve target discrimination and field of view localization.

2. The intelligent deformable small target recognition algorithm based on detector multiplexing aliased images according to claim 1, characterized in that, When the principal axis of the field-of-view encoding element in step 2 is in the y-direction, the change in the field of view in the x-direction introduces a tilted incident angle. This causes the equivalent optical axis of the field-of-view encoding element to deflect, resulting in a rotation of the focal line in the xy plane. The rotation angle is... satisfy: in: The focal length of the field-of-view encoding element; : The distance from the field-of-view encoding element to the final imaging surface; Linear changes in the target's field of view in the x-direction; As the field of view angle in the y-direction increases, the degree of beam focusing in the y-direction changes, causing a change in the size of the beam spot, satisfying the following: in: : The radius of the light spot in the y-direction; : Initial beam radius of the incident beam in the y-direction; Numerical aperture; Wavelength of light; : Field of view in the y direction.

3. The intelligent deformable small target recognition algorithm based on detector multiplexing in aliased images according to claim 1, characterized in that, The intelligent deformable small target recognition algorithm in step 5 is either optical flow or background modeling.

Images