CN-122027795-A - Self-adaptive parallel compressed sensing imaging method and system based on low-resolution pre-imaging

CN122027795ACN 122027795 ACN122027795 ACN 122027795ACN-122027795-A

Abstract

The invention relates to the technical field of computational imaging and compressed sensing, and particularly discloses a self-adaptive parallel compressed sensing imaging method and system based on low-resolution pre-imaging. The method first acquires a low resolution pre-imaged image of a scene by all-pass illumination and locates a plurality of target areas therefrom. After the target area is mapped to the plane of the spatial light modulator, a coded modulation pattern matched with the size of the target area is generated for each area, and all patterns are spatially rearranged to form a combined modulation pattern which is not overlapped with each other. After the combined pattern is projected, compressed sensing sampling signals are acquired by using a low-resolution image sensor, and a high-resolution image is independently reconstructed for each target area. The system comprises a spatial light modulator, a low resolution image sensor, an objective lens and a computer for implementing the method.

Inventors

DENG HUAXIA
XIAO ZAIHONG
GONG XINGLONG

Assignees

中国科学技术大学

Dates

Publication Date: 20260512
Application Date: 20260212

Claims (10)

1. An adaptive parallel compressed sensing imaging method based on low-resolution pre-imaging is characterized by comprising the following steps: Projecting an all-pass illumination pattern to the whole scene to be imaged by using a spatial light modulator, and acquiring a low-resolution pre-imaging image reflecting the whole spatial distribution of the scene by using a low-resolution image sensor; positioning at least two target areas in a scene in an image processing mode based on the low-resolution pre-imaging image, and generating an area mask of each target area; Mapping each target area from a low-resolution imaging plane to a projection plane of the spatial light modulator according to the geometric calibration relation of the imaging system, and determining a corresponding modulation area of each target area on the spatial light modulator; For the size information of each target area, respectively generating a code modulation pattern matched with the size of the corresponding target area for each target area; Spatially rearranging the coded modulation patterns corresponding to all the target areas on a projection plane of the spatial light modulator to form combined modulation patterns which are not overlapped and are in one-to-one correspondence with the target areas; Controlling a spatial light modulator to project the combined modulation pattern to a scene to be imaged, and collecting corresponding compressed sensing sampling signals through a low-resolution image sensor; and for each target area, independently reconstructing a high-resolution image of the target area by adopting a compressed sensing reconstruction algorithm based on the compressed sensing sampling signals corresponding to the target area.
2. The method of claim 1, wherein generating a coded modulation pattern for each target region that matches the size of the corresponding target region comprises adjusting the coding density or coding type of the generated coded modulation pattern based on the gray scale distribution characteristics or structural complexity of the target region in the low resolution pre-image.
3. The adaptive parallel compressed sensing imaging method based on low resolution pre-imaging according to claim 2, wherein the type of the coded modulation pattern comprises a random coding pattern, a hadamard coding pattern or a fourier coding pattern.
4. The adaptive parallel compressed sensing imaging method based on low resolution pre-imaging according to claim 1, wherein the step of spatially rearranging the coded modulation patterns corresponding to all the target areas causes the overall measurement matrix to be represented as a block diagonal structure, thereby realizing parallel and mutually independent compressed sensing sampling of a plurality of target areas in the same modulation period.
5. The adaptive parallel compressed sensing imaging method based on low resolution pre-imaging of claim 1, wherein locating target areas based on the low resolution pre-imaging image comprises thresholding and area connectivity analysis of the low resolution pre-imaging image to identify and separate different target areas.
6. The adaptive parallel compressed sensing imaging method based on low resolution pre-imaging of claim 1, wherein the compressed sensing reconstruction algorithm comprises an orthogonal matching pursuit algorithm, a basis pursuit algorithm, an inverse transformation reconstruction algorithm, or a depth learning based reconstruction algorithm.
7. A low resolution pre-imaging based adaptive parallel compressed sensing imaging system for implementing the method of any of claims 1 to 6, comprising: the spatial light modulator is arranged in the illumination light path and is used for projecting an all-pass illumination pattern or a coded modulation pattern to a scene to be imaged; The low-resolution image sensor is arranged in the imaging light path and is used for acquiring a low-resolution pre-imaging image and compressed sensing sampling signals; The first objective lens is arranged in the illumination light path and used for projecting illumination light modulated by the spatial light modulator to an imaging scene; a second objective lens arranged in the imaging light path and used for imaging a light field of an imaging scene to the low-resolution image sensor; the computer is electrically connected with the spatial light modulator and the low-resolution image sensor; The computer is configured to perform the following operations: receiving a low resolution pre-imaging image acquired by a low resolution image sensor and processing the image to locate at least two target areas; Calculating the mapping position of each target area on the projection plane of the spatial light modulator; Generating a coded modulation pattern for each target area, and spatially rearranging all the coded modulation patterns to generate a combined modulation pattern, and sending the combined modulation pattern to a spatial light modulator; Receiving compressed sensing sampling signals acquired by a low-resolution image sensor under the illumination of a combined modulation pattern; And extracting corresponding data from the compressed sensing sampling signals for each target area, and independently reconstructing a high-resolution image of each target area.
8. The adaptive parallel compression-aware imaging system based on low resolution pre-imaging of claim 7, wherein when the computer generates a coded modulation pattern for each target region, the coding density or coding type of the coded modulation pattern is adaptively adjusted according to the gray scale distribution characteristics or structural complexity of the corresponding target region.
9. The adaptive parallel compression-aware imaging system of claim 7, wherein the computer spatially reorders the coded modulation patterns by ensuring that the patterns do not overlap each other on the projection plane of the spatial light modulator and are aligned with the mapping locations of the target region, respectively.
10. The adaptive parallel compressed sensing imaging system of claim 7, wherein the computer processes the parallel acquired compressed sensing sample signals using a block diagonal measurement matrix model and performs an independent image reconstruction operation for each target region.

Description

Self-adaptive parallel compressed sensing imaging method and system based on low-resolution pre-imaging Technical Field The invention relates to the technical field of computational imaging and compressed sensing, and particularly discloses a self-adaptive parallel compressed sensing imaging method and system based on low-resolution pre-imaging. Background In conventional compressed sensing imaging techniques, the entire imaging field of view is typically subjected to uniform code modulation and sampling. When a plurality of discrete target areas exist in a scene, the method can measure the target areas and non-target areas (such as the background) at the same time, so that a large amount of sampling resources are wasted on the areas with sparse information or useless information, and the overall imaging efficiency is reduced. Particularly in the application requiring rapid and high-resolution imaging of a plurality of targets, how to avoid ineffective sampling of non-attention areas and realize adaptive distribution and parallel acquisition of sampling resources is a technical problem to be solved in the prior art. Disclosure of Invention The invention aims to solve the problems of resource waste and low efficiency caused by unified sampling of the whole view field in the existing compressed sensing imaging technology, and provides a method and a system which can adaptively position a plurality of target areas and perform parallel compressed sensing sampling only on the areas so as to improve the imaging efficiency under a multi-target scene. The invention provides a self-adaptive parallel compressed sensing imaging method based on low-resolution pre-imaging, which comprises the following steps: Projecting an all-pass illumination pattern to the whole scene to be imaged by using a spatial light modulator, and acquiring a low-resolution pre-imaging image reflecting the whole spatial distribution of the scene by using a low-resolution image sensor; positioning at least two target areas in a scene in an image processing mode based on the low-resolution pre-imaging image, and generating an area mask of each target area; Mapping each target area from a low-resolution imaging plane to a projection plane of the spatial light modulator according to the geometric calibration relation of the imaging system, and determining a corresponding modulation area of each target area on the spatial light modulator; For the size information of each target area, respectively generating a code modulation pattern matched with the size of the corresponding target area for each target area; Spatially rearranging the coded modulation patterns corresponding to all the target areas on a projection plane of the spatial light modulator to form combined modulation patterns which are not overlapped and are in one-to-one correspondence with the target areas; Controlling a spatial light modulator to project the combined modulation pattern to a scene to be imaged, and collecting corresponding compressed sensing sampling signals through a low-resolution image sensor; and for each target area, independently reconstructing a high-resolution image of the target area by adopting a compressed sensing reconstruction algorithm based on the compressed sensing sampling signals corresponding to the target area. The invention provides a self-adaptive parallel compressed sensing imaging system based on low-resolution pre-imaging, which comprises the following components: the spatial light modulator is arranged in the illumination light path and is used for projecting an all-pass illumination pattern or a coded modulation pattern to a scene to be imaged; The low-resolution image sensor is arranged in the imaging light path and is used for acquiring a low-resolution pre-imaging image and compressed sensing sampling signals; The first objective lens is arranged in the illumination light path and used for projecting illumination light modulated by the spatial light modulator to an imaging scene; a second objective lens arranged in the imaging light path and used for imaging a light field of an imaging scene to the low-resolution image sensor; the computer is electrically connected with the spatial light modulator and the low-resolution image sensor; The computer is configured to perform the following operations: receiving a low resolution pre-imaging image acquired by a low resolution image sensor and processing the image to locate at least two target areas; Calculating the mapping position of each target area on the projection plane of the spatial light modulator; Generating a coded modulation pattern for each target area, and spatially rearranging all the coded modulation patterns to generate a combined modulation pattern, and sending the combined modulation pattern to a spatial light modulator; Receiving compressed sensing sampling signals acquired by a low-resolution image sensor under the illumination of a combined modulation pattern; And extracting cor