A method for realizing wide-band hyperspectral detection based on coded spectrum device

Abstract

The present invention discloses a method for realizing wide-band hyperspectral detection based on a coded spectrum device, comprising: a coded spectrum array module, a light field-like imaging module and a segmented compressed sensing processing module. The coded spectrum array module is used as a front module to perform structural arrangement of the filter array and spectrum coding design; the light field-like imaging module is used as an optical system to perform compressed sampling in conjunction with the coded spectrum array module; the segmented compressed sensing processing module uses anti-noise clustering and greedy iterative algorithms to reconstruct the sampled image, and finally obtains hyperspectral data. This method does not require complex optical path design, simplifies the system, and reduces costs; and its coding solution is simpler and more accurate, and the detection spectrum can be greatly broadened through structural design.

Description

Technical Field

[0001] The present invention relates to the field of computational optical imaging technology, and in particular to a method for realizing wide-band hyperspectral detection based on a coded spectrum device, which mainly comprises encoding and modulating target spectral information in the wavelength dimension by designing a coded filter with a specific distribution, and then restoring the original hyperspectral information from the modulated information by a compressed sensing algorithm. Background Art

[0002] With the development of imaging technology, spectral imaging has begun to rise and has been widely used in many fields such as medical testing, geological survey, target identification and biological monitoring. Hyperspectral imaging technology combines spectral information with image information through dual innovations in concepts and technologies to obtain image information data with spectral characteristics of ground objects. This technology of simultaneously obtaining spectral information and image information provides a means for people to study the characteristics of ground objects in depth.

[0003] Although traditional spectral imaging instruments have high accuracy, they have moving parts inside, are expensive and are not suitable for dynamic scenes. Currently, there are two commonly used spectral imaging technologies: snapshot and computational. Among them, the snapshot method can obtain a three-dimensional data cube with a single exposure to achieve dynamic scene detection, but its number of spectral channels is limited by the number of bandpass filters and cannot meet the requirements of high spectral resolution. The more common methods in the computational method are aperture coding and deep learning. Among them, the aperture coding method requires the incident light to be encoded in space and then wavelength encoded through a dispersion prism. The detector perceives two-dimensional aliasing information, which is difficult to reconstruct. Although the deep learning method has a faster reconstruction speed, it has a fatal flaw that it cannot be used at all when samples are difficult to collect.

[0004] In view of the above problems, the present invention proposes a method for realizing wide-band hyperspectral detection based on a coded spectrum device. Firstly, a mathematical model of the filter array is established and the optimal parameters are obtained through optimization. Then, the mathematical model is applied to the designed optical system to realize low-cost wide-band hyperspectral detection. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a method for realizing high-spectral detection in a wide spectrum band based on a coded spectrum device, constructs an optimized spectrum based on a multi-cavity Fabry-Perot interference structure, and realizes high-resolution spectral detection in a wide spectrum band based on the array as an observation matrix.

[0006] The technical solution adopted by the present invention is: a method for realizing wide-band hyperspectral detection based on a coded spectrum device, the system used in the method includes: a coded spectrum array module, a light field-like imaging module and a segmented compressed sensing processing module, and the method includes the following steps:

[0007] Step 1: Acquire a characteristic image of the target scene using a combination system of a coded spectrum array module and a light field-like imaging module;

[0008] Step 2: Use the segmented compressed sensing processing module to extract feature images. First, use the clustering algorithm plus the filtering algorithm to classify and denoise the spatial objects; then use segmented compressed sensing to reconstruct the hyperspectral information of the wide spectrum.

[0009] Furthermore, the establishment of the coded spectrum array module includes: according to the thin film transmission theory, a mathematical model of the transmittance of the interference coded filter is established, and the coded spectrum array model can change according to the change of the number of material layers and thickness, so as to construct a filter structure with different characteristic coded spectrum distributions by alternately plating high and low refractive index materials; according to the variation of the RIP theory, the PSO optimization algorithm is used to find the optimal spectrum coding structure with the minimum correlation as the objective function. Construct a filter array with the best imaging effect.

[0010] Furthermore, establishing a light field-like imaging module includes: designing a system optical path, placing a microlens array near the focal plane of the main lens, and placing a CCD detector at the focal plane of the microlens array; wherein the microlens array plays a role in image segmentation and spatial information replication.

[0011] Furthermore, a segmented compressed sensing processing module is established, including: reading feature images, combining clustering algorithm with filtering algorithm to remove sensor noise; extracting denoised feature points of ground objects, using segmented coding to improve the calculation stability of each segment, reconstructing them using greedy iteration method respectively, and finally fusing each sub-spectral segment to obtain a wide-band hyperspectral image of the target scene.

[0012] The principle of the present invention is: a method for realizing wide-band hyperspectral detection based on a coded spectrum device. From the analysis of computational spectral imaging technology, this method is completed by first encoding and sampling the original redundant data, and then decoding and reconstructing it. Since the various bands of material spectral information have extremely high correlation, we can use this redundant information to project it into a sparse high-dimensional space, and design a filter with a specific distribution to encode this information in a single dimension of wavelength, and then use a related decoding algorithm to reconstruct the original information. In this way, the prism or grating used for dispersion in the system can be removed, the system can be simplified, and the reliability of the system can be improved.

[0013] Compared with the existing method, the present invention has the following advantages:

[0014] (1) Compared with the traditional scanning imaging spectrometer, the present invention simplifies the system, reduces the cost, and is more suitable for dynamic scenes.

[0015] (2) Compared with the snapshot spectral imaging system, the present invention has more spectral channels and can achieve higher spectral resolution.

[0016] (3) Compared with the aperture coded CASSI system, the present invention reduces the number of system components and the computational complexity.

[0017] (4) Compared with the spectral imaging system based on deep learning, the present invention is more suitable for hyperspectral detection in complex scenes.

[0018] (5) The present invention can realize high-spectral detection of a wide spectrum at low cost, requiring only 16 filters at least. The operation is simple and easy to implement. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of a method for realizing wide-band hyperspectral detection based on a coded spectrum device of the present invention;

[0020] Figure 2 It is a flow chart of a method for realizing wide-band hyperspectral detection based on a coded spectrum device of the present invention. DETAILED DESCRIPTION

[0021] The specific implementation of the present invention is described in detail below with reference to the accompanying drawings:

[0022] like Figure 1 As shown, a method for realizing wide-band hyperspectral detection based on a coded spectrum device uses a system including: a coded spectrum array module 1, a light field-like imaging module 2, and a segmented compressed sensing processing module 3.

[0023] First, a mathematical model of the multi-cavity Fabry-Perot cavity interference structure is established based on the thin film theory. The incident light will interfere when passing through the dielectric film, and the light of a specific wavelength will be selected to pass through. The Maxwell boundary relationship is used on both sides of each interface to connect the electromagnetic fields on both sides, and then the corresponding relationship between the equivalent admittance of the equivalent medium and the real film system parameters is established. The characteristic matrix of a single-layer dielectric film is:

[0024]

[0025] in, is the equivalent admittance of a single-layer dielectric film, is the phase thickness of the film, is the refractive index of the medium, is the dielectric layer thickness, is the angle of incidence, is the wavelength of incident light, is the membrane admittance, is the admittance of the medium outside the lower interface, and are the incident and outgoing electric field strengths of the dielectric film.

[0026] In the case of multilayer films, similar to single-layer films, the equivalent interface method is used continuously. If there are k layers of films:

[0027]

[0028] in, is the equivalent admittance of the multilayer dielectric film, is the phase thickness of each film layer, It is The refractive index of the layer medium, It is The dielectric layer thickness, It is Layer incident angle, is the wavelength of incident light, It is The admittance of the thin film, is the admittance of the medium outside the last film. and are the incident and outgoing electric field strengths of the multilayer dielectric film.

[0029] According to the above formula, a mathematical model of the multilayer film is established, and then a coded spectrum array model is constructed; then, the minimum matrix correlation is used as the objective function, and the PSO algorithm is used to find the optimal distribution of the coded filter structure. Figure 1 The coded spectrum array module 1 is then built. The experimental optical path is then constructed. The filter array is placed in front of the objective lens to achieve coded modulation of the incident light. The microlens array is placed near the focal plane of the objective lens to achieve spatial information replication and image segmentation. The image to be processed is obtained on the detector. Figure 1 The quasi-light field imaging module 2 in the image processing module is used. Finally, the image to be processed is input into the computer, and the values ​​of multiple encoding channels corresponding to each point in the space are extracted for restoration, and finally the reconstructed hyperspectral data is obtained. Figure 1 The segmented compressed sensing processing module 3 in.

[0030] like Figure 2 As shown in the figure, the specific process of the method for realizing wide-band hyperspectral detection based on the coded spectrum device is to establish a mathematical model of the multi-cavity filter array, and iterate and optimize the coded filter array with the optimal distribution through the optimization algorithm. Then, the target scene is sampled by the combined optical system to obtain the image to be processed; then, the related algorithm is used to extract features and denoise the sub-image, and finally the greedy iterative algorithm is used to reconstruct the hyperspectral data.

[0031] The present invention discloses a method for realizing wide-band hyperspectral detection based on a coded spectrum device. There are many methods and approaches to realize the technical solution. The above is only a preferred implementation mode of the present invention. All components not specified in this embodiment can be realized by existing technologies.

Claims

1. A method for realizing wide-band hyperspectral detection based on a coded spectrum device, characterized in that: The system used in the method comprises: a coded spectrum array module (1), a light field-like imaging module (2) and a segmented compressed sensing processing module (3), and the method comprises the following steps: Step 1: using a combined system of a coded spectrum array module (1) and a light field-like imaging module (2) to obtain a characteristic image of a target scene; establishing the coded spectrum array module (1) includes: according to thin film transmission theory, establishing a mathematical model of a multi-cavity Fabry-Perot cavity interference structure, wherein the coded spectrum array model can be changed according to changes in the number of material layers and thickness, thereby constructing a filter structure with different characteristic coded spectrum distributions by alternately plating materials with high and low refractive indices; according to a variation of the RIP theory, taking the minimum correlation as the objective function, using a PSO optimization algorithm to find the optimal spectrum coding structure, and constructing a filter array with the optimal imaging effect; Step 2: extracting feature images using a segmented compressed sensing processing module (3), firstly using a clustering algorithm plus a filtering algorithm to classify and denoise spatial objects; then using segmented compressed sensing to reconstruct wide-band hyperspectral information; wherein establishing the segmented compressed sensing processing module (3) includes: reading feature images, combining a clustering algorithm with a filtering algorithm to remove sensor noise; extracting denoised feature points of objects, using segmented coding to improve the computational stability of each segment, reconstructing using a greedy iterative method, and finally fusing each sub-spectral segment to obtain a wide-band hyperspectral image of the target scene.

2. The method for realizing wide-band hyperspectral detection based on a coded spectrum device according to claim 1, characterized in that: Establishing a light field-like imaging module (2) includes: designing a system optical path, placing a microlens array near the focal plane of the main lens, and placing a CCD detector at the focal plane of the microlens array; wherein the microlens array plays a role in image segmentation and spatial information replication.

Images