Micro-scanning super-resolution method for infrared point target detection

Abstract

The present invention belongs to the field of optical system image processing, and specifically discloses a micro-scanning super-resolution method for infrared point target detection, comprising the steps of: a host computer sends a super-resolution instruction, and sends a set external trigger signal to a miniaturized high-frequency controller; the miniaturized high-frequency controller processes and amplifies the signal according to the trigger signal, and outputs it to a high-precision micro-scanning platform to drive the platform to perform micro-displacement movement; after the high-precision micro-scanning platform reaches a stable position, the detector starts integration to obtain low-resolution original image data; the original image data is transmitted to an image algorithm processor, and the image algorithm processor performs super-resolution image processing and point target extraction on the input low-resolution original image data according to the host computer instruction; and the processed result is output to a back-end device. The present invention uses a down-sampled low-resolution sequence obtained by reconstructing a super-resolution image, which greatly reduces the clutter introduced by the phase difference and improves the detection distance.

Description

Technical Field

[0001] The invention relates to the field of mobile optical system imaging processing, in particular to a micro-scanning super-resolution method for infrared point target detection. Background Art

[0002] Due to the field of view of the infrared imaging system and the aperture of the detector pixel, the resolution of infrared targets is generally low, and they often appear in the form of point targets, generally occupying only a few pixels. This characteristic of infrared targets makes target detection technology have many difficulties. First, the area occupied by long-distance infrared targets is small, and the targets have almost no characteristic information such as size, shape, texture, etc., and the amount of useful information that can be provided to the detection algorithm is very small. Secondly, the signal-to-noise ratio of infrared targets is very low, and when propagating in the atmosphere, they will be affected by atmospheric attenuation, rain, snow and other environments, resulting in very low target signal intensity received by the infrared sensor. In addition, the background information is complex. Atmospheric clouds, sea surface, trees, etc. in nature will cause great interference to infrared point source targets and form randomly distributed background noise points, which are close to the characteristics of infrared targets, easily causing false alarms, bringing considerable difficulties to target detection.

[0003] Infrared point target detection is actually the process of automatically detecting targets in cluttered and noisy environments using image processing algorithms. Existing algorithms cannot effectively suppress background clutter for complex backgrounds, resulting in a high false alarm rate in target detection results. Therefore, a micro-scanning super-resolution method for infrared point target detection is needed. Summary of the invention

[0004] The object of the present invention is to provide a micro-scanning super-resolution method for infrared point target detection to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a micro-scanning super-resolution method for infrared point target detection, comprising the following steps:

[0006] S1: The host computer sends the super-resolution command and sends the set external trigger signal to the miniaturized high-frequency controller;

[0007] S2: The miniaturized high-frequency controller processes and amplifies the trigger signal and outputs it to the high-precision micro-scanning platform to drive the platform to perform micro-displacement motion;

[0008] S3: After the high-precision micro-scanning platform reaches a stable position, the detector starts integrating to obtain a low-resolution image sequence;

[0009] S4: The original image data is transmitted to the image algorithm processor, which performs super-resolution image processing and point target extraction on the input low-resolution image sequence according to the instructions of the host computer;

[0010] S5: Output the processed results to the backend device.

[0011] Preferably, the super-resolution image processing and point target extraction process in S4 includes the following steps:

[0012] S4a: Read multiple frames of low-resolution image sequence Y with sub-pixel displacement k ;

[0013] S4b: Read the imaging system calibration function H k ;

[0014] S4c: Calculate the image noise level V k ;

[0015] S4d: Use the optical flow motion estimation algorithm to calculate the displacement between image sequences and obtain the motion information matrix F k ;

[0016] S4e: Constructing the initial super-resolution graph

[0017] S4f: construct the system matrix W using the parameters obtained in steps S4b, S4c, and S4d;

[0018] W=D k H k F k Among them, D k is the downsampling operator;

[0019] S4g: Using the parameters obtained in steps S4e and S4f, reversely solve the super-resolution image

[0020] Y k =D k H k F k X+V k ,k=1,2,…K

[0021] S4h: Super-resolution image obtained in step S4g After forward degradation with the system matrix W, a set of estimated values ​​Y of low-resolution image sequences are obtained k ′;

[0022] S4i: The estimated value Y of the low-resolution image sequence k ′ and the input low-resolution image sequence Y k Perform differential operation to obtain the differential image DIF k ;

[0023] S4j: Calculate the maximum value Max of the low-resolution image sequence and determine the segmentation threshold;

[0024] S4k: The difference image DIF obtained in step 4i k Image segmentation is performed according to the segmentation threshold T_thres, that is, the image grayscale value higher than the segmentation threshold is considered as a point target;

[0025] S41: According to the multi-frame point target position information obtained in step 4k, the targets are screened for a second time to remove unreasonable point targets;

[0026] S4m: point target position mark;

[0027] S4n: Repeat steps S4a-S4m until all images are processed.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. The present invention introduces image micro-scanning technology in the process of infrared point target detection, and uses a controllable displacement module to generate accurate sub-pixel displacement information. Micro-scanning greatly increases the probability of point targets falling into the center area of ​​the pixel, and also enhances the target signal. At the same time, the micro-scanning super-resolution algorithm for infrared point target detection of the present invention uses the optical flow method to estimate the motion between images, has very high scene robustness, and can be applied to moving scenes, such as airborne platforms.

[0030] 2. The micro-scanning super-resolution algorithm used in the present invention for infrared point target detection adopts multi-frame image information, which can effectively suppress image noise. The micro-scanning super-resolution algorithm used for infrared point target detection introduces imaging system calibration to obtain the physical properties of the imaging system, thereby improving algorithm performance and increasing algorithm reliability.

[0031] 3. The present invention constructs a physical model relationship between the original high-resolution image and the low-resolution image, and uses the sequence information of multiple frames of low-resolution images with sub-pixel information to reversely solve the high-resolution image, thereby improving the resolution and resolution of the imaging system.

[0032] 4. The algorithm and method of the present invention utilize the downsampled low-resolution sequence obtained by reconstructing the super-resolution image, and perform differential operation with the input low-resolution image sequence, which greatly reduces the clutter introduced by the phase difference, and can extract point targets and well preserve the energy of point targets, thereby improving the detection distance. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a logic block diagram of the present invention;

[0034] Figure 2 Schematic diagram of the execution flow chart of the super-resolution image processing process in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] See also Figure 1-2 The present invention provides a technical solution: a micro-scanning super-resolution method for infrared point target detection, comprising the following steps:

[0037] S1: The host computer sends the super-resolution command and sends the set external trigger signal to the miniaturized high-frequency controller;

[0038] S2: The miniaturized high-frequency controller processes and amplifies the trigger signal and outputs it to the high-precision micro-scanning platform to drive the platform to perform micro-displacement motion;

[0039] S3: After the high-precision micro-scanning platform reaches a stable position, the detector starts integrating to obtain a low-resolution image sequence;

[0040] S4: The original image data is transmitted to the image algorithm processor, which performs super-resolution image processing and point target extraction on the input low-resolution image sequence according to the instructions of the host computer;

[0041] S5: Output the processed results to the backend device.

[0042] In this embodiment, the super-resolution image processing and point target extraction process in S4 includes the following steps:

[0043] S4a: Read multiple frames of low-resolution image sequence Y with sub-pixel displacement k ;

[0044] S4b: Read the imaging system calibration function H k ;

[0045] S4c: Calculate the image noise level V k ;

[0046] S4d: Use the optical flow motion estimation algorithm to calculate the displacement between image sequences and obtain the motion information matrix F k ;

[0047] S4e: Constructing the initial super-resolution graph

[0048] S4f: construct the system matrix W using the parameters obtained in steps S4b, S4c, and S4d;

[0049] W=Dk H k F k Among them, D k is the downsampling operator;

[0050] S4g: Using the parameters obtained in steps S4e and S4f, reversely solve the super-resolution image

[0051] Y k =D k H k F k X+V k ,k=1,2,…K

[0052] S4h: Super-resolution image obtained in step S4g After forward degradation with the system matrix W, a set of estimated values ​​Y of low-resolution image sequences are obtained k ′;

[0053] S4i: The estimated value Y of the low-resolution image sequence k ′ and the input low-resolution image sequence Y k Perform differential operation to obtain the differential image DIF k ;

[0054] S4j: Calculate the maximum value Max of the low-resolution image sequence and determine the segmentation threshold;

[0055] S4k: The difference image DIF obtained in step 4i k Image segmentation is performed according to the segmentation threshold T_thres, that is, the image grayscale value higher than the segmentation threshold is considered as a point target;

[0056] S41: performing secondary screening of the targets according to the multi-frame point target position information obtained in step S4k to screen out unreasonable point targets;

[0057] S4m: point target position mark;

[0058] S4n: Repeat steps S4a-S4m until all images are processed.

[0059] The present invention actually moves the lens (or lens group) in the optical system to make the image move sub-pixel on the focal plane, collects multiple original images with sub-pixel displacement between each other, and detects infrared point targets after being processed by micro-scanning super-resolution algorithm.

[0060] In this embodiment, by introducing image micro-scanning technology in the process of detecting infrared point targets, using a controllable displacement module to generate accurate sub-pixel displacement information, micro-scanning greatly increases the probability of point targets falling into the center area of ​​the pixel, and also enhances the target signal; at the same time, the micro-scanning super-resolution algorithm for infrared point target detection of the present invention uses the optical flow method to estimate the motion between images, has very high scene robustness, and can be applied to moving scenes, such as airborne platforms.

[0061] In this embodiment, by constructing a physical model relationship between the original high-resolution image and the low-resolution image, the high-resolution image is reversely solved using the information of a multi-frame low-resolution image sequence with sub-pixel information, thereby improving the resolution and resolving power of the imaging system; and the algorithm and method utilize the downsampled low-resolution sequence obtained by the reconstructed super-resolution image to perform a differential operation with the input low-resolution image sequence, thereby greatly reducing the clutter introduced by the phase difference, being able to extract the point target, and well retaining the point target energy, thereby improving the detection distance.

[0062] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A micro-scanning super-resolution method for infrared point target detection, characterized in that: The steps include: S1: The host computer sends the super-resolution command and sends the set external trigger signal to the miniaturized high-frequency controller; S2: The miniaturized high-frequency controller processes and amplifies the trigger signal and outputs it to the high-precision micro-scanning platform to drive the platform to perform micro-displacement motion; S3: After the high-precision micro-scanning platform reaches a stable position, the detector starts integrating to obtain a low-resolution image sequence; S4: The original image data is transmitted to the image algorithm processor, which performs super-resolution image processing and point target extraction on the input low-resolution image sequence according to the instructions of the host computer; S5: Output the processed results to the backend device; The super-resolution image processing and point target extraction process in S4 includes the following steps: S4a: Read multiple frames of low-resolution image sequence Y with sub-pixel displacement k ; S4b: Read the imaging system calibration function H k ; S4c: Calculate the image noise level V k ; S4d: Use the optical flow motion estimation algorithm to calculate the displacement between image sequences and obtain the motion information matrix F k ; S4e: Constructing the initial super-resolution graph S4f: construct the system matrix W using the parameters obtained in steps S4b, S4c, and S4d; W=D k H k F k Among them, D k is the downsampling operator; S4g: Using the parameters obtained in steps S4e and S4f, reversely solve the super-resolution image Y k =D k H k F k X+V k ,k=1,2,…K S4h: Super-resolution image obtained in step S4g After forward degradation with the system matrix W, a set of estimated values ​​Y of low-resolution image sequences are obtained k ′; S4i: The estimated value Y of the low-resolution image sequence k ′ and the input low-resolution image sequence Y k Perform differential operation to obtain the differential image DIF k ; S4j: Calculate the maximum value Max of the low-resolution image sequence and determine the segmentation threshold; The S4 also includes the segmentation processing of the differential image and the screening and marking of the target position information, which specifically includes the steps of: S4k: The difference image DIF obtained in step 4i k Image segmentation is performed according to the segmentation threshold T_thres, that is, the image grayscale value higher than the segmentation threshold is considered as a point target; S41: Perform secondary screening on the targets according to the multi-frame point target position information obtained in step S4k to screen out unreasonable point targets; S4m: point target position mark; S4n: Repeat steps S4a-S4m until all images are processed.

Images