CN202485671U - Compression perception theory based imaging device of aerospace camera - Google Patents

Abstract

The utility model relates to a compression perception theory based imaging device of an aerospace camera. The device comprises a first lens, a second lens, a DMD (Digital Micromirror Device) array, a single pixel detector, an A/D (Analog/Digital) converter and an image reconstruction module, wherein the first lens is used for converging scene incident light to the DMD array; the second lens is used for converging reflection light on the DMD array to the single pixel detector; the DMD array is used for reflecting the light converged from the first lens to the second lens through a micromirror of a measuring matrix; the single pixel detector is used for receiving the light reflected from the DMD array and converting an optical signal into an electric signal; and the image reconstruction module is used for reconstructing a digital signal received from the A/D converter into an image, particularly including the steps of selecting the measuring matrix of the DMD, carrying out sparse representation to the digital signal, calculating a transit matrix in the sparse representation, and calculating the constrained minimum to the 1-norm so as to obtain the reconstructed image. The device is applicable for deep space exploration in a severe environment; the camera is lighter and smaller; and the volume and the power consumption cost are small.

Description

Based on the theoretical camera space imaging device of compressed sensing

Technical field

The utility model belongs to the optical remote sensing technology field, particularly relates to based on the theoretical camera space imaging device of compressed sensing.

Background technology

Shannon's sampling theorem is satisfied in existing camera space sampling strictness, and the sample information amount receives the restriction of sampling thheorem, and picture element is difficult to improve.Detector area is excessive, is not easy to refrigeration.The tradition remote sensing camera can be collected lot of data when earth observation for a long time, on star, be difficult to carry out the high capacity storage.

The utility model content

The purpose of the utility model is to overcome the above-mentioned deficiency of prior art; Provide based on the theoretical camera space imaging device of compressed sensing; This device adopts the detector of unit picture element size to replace the extensive planar array detector of tradition; Be applicable to the survey of deep space that environment is abominable, make camera gentlier little, volume and power consumption cost are little.

The above-mentioned purpose of the utility model is achieved through following technical scheme:

Based on the theoretical camera space imaging device of compressed sensing, comprise first lens, second lens, dmd array, single pixel detector, A/D and image reconstruction module, wherein:

First lens: the scene incident light is converged on the dmd array;

Second lens: the reflection ray on the dmd array is converged on single pixel detector;

Dmd array: comprise addressing integrated chip and the micromirror of measuring matrix, will reflex to light on second lens through the micromirror of measuring matrix from the light that first lens converge;

Single pixel detector: receive the light of dmd array reflection, convert light signal into electric signal, export to A/D;

A/D: receive the electric signal of single pixel detector output, carry out mould/number conversion, digital signal is exported to image reconstruction module;

Image reconstruction module: the digital signal that receives A/D output; Digital signal to receiving is carried out image reconstruction; Concrete grammar is: select the measurement matrix of DMD, the digital signal that receives is carried out rarefaction representation, and obtain the transition matrix in the rarefaction representation; Ask constraint minimum to 1 norm, thus the image that obtains rebuilding.

In the above-mentioned camera space imaging device theoretical based on compressed sensing; Image reconstruction module comprises that measuring matrix selects module, rarefaction representation module, transition matrix module and norm least commitment module; Wherein measuring matrix selects module to be used to select the measurement matrix of DMD; The rarefaction representation module is used for the digital signal that receives is carried out rarefaction representation, and the transition matrix module is used for obtaining the transition matrix of rarefaction representation, and norm least commitment module is used for asking constraint minimum to 1 norm.

The utility model compared with prior art has following advantage:

(1) the utility model has adopted the compressed sensing principle of signal rarefaction representation in image reconstruction module; Before sampling, carry out signal compression; Broken through the traditional sampling theorem and got constraint condition, the quantity of information that obtains is bigger, is applicable to multispectral, Hyper spectral Imaging that amount of image information is bigger;

(2) the utility model adopts the detector of unit picture element size to replace the extensive planar array detector of tradition, has solved the gordian technique difficult problem of detector (especially infrared) dimension-limited, can make the device refrigeration be easy to realize simultaneously;

(3) the utility model imaging device volume and power consumption cost are little, make camera gentlier little, are applicable to survey of deep space or the extremely abominable interstellar space detection of other environment that environment is abominable; Be fit to microsensor (space resources is most deficient);

(4) the utility model imaging device is particularly useful for the low light level and the spectral coverage outside visible spectrum carries out work, such as infrared, the microwave that is used in the sensor dimension-limited, low light level (snowslide pipe) etc., has stronger practicality.

Description of drawings

Fig. 1 is the utility model camera space imaging device structural representation;

Fig. 2 is the structural representation of image reconstruction module in the utility model imaging device.

Embodiment

Below in conjunction with accompanying drawing and specific embodiment the utility model is described in further detail:

Be illustrated in figure 1 as the utility model camera space imaging device structural representation; Can know that by figure this imaging device comprises first lens, second lens, dmd array, single pixel detector, A/D and image reconstruction module; Wherein dmd array adopts the HD3+DMD chip of American TI Company, and resolution is 1280 * 720, and micromirror is a square; Its length of side is 13-16um, is in respectively+12 ° and-12 ° of two duties.System relies on static random-access reservoir (SRAM) unit that each micro mirror is carried out addressing, and the electrostatic force that uses cmos circuit to provide drives the rotation of micro mirror stationary shaft.DMD (digital micromirror device) device is a kind of fast digital photoswitch reflective array that is integrated on the addressing integrated chip, uses it can be used for producing the random measurement matrix.Single pixel detector generally can be processed with the bigger photodiode of single dynamic range.Single detector before the camera lens be single pixel camera be with traditional C CD/CMOS camera between structurally maximum different.Certainly, also can make the small-scale array of forming by small number of pixels (its pixel cell is less than traditional C CD/CMOS detector) to detector.According to the measured value of gained repeatedly, reconstituting initial image information.Because the detector scale is less, be easy to the infrared eye refrigeration.

What single pixel camera was different with traditional C CD/CMOS camera is that single pixel camera has been put into ground high performance parallel Distributed Calculation processing to the work of treatment of big data quantity, thereby accomplishes the ground integrated system of star.Like this, required power consumption reduces greatly on the star, reduces the camera size.Because camera detector sampled value is that the measured value dimension is quite low through the matrix of original signal after the conversion of two kinds of orthogonal basiss, negligible amounts makes things convenient for data-storing and transmission on the star.Compressed sensing can be converted into the cost that forms images on traditional star the ground recovery work of off-line.

The concrete course of work of this imaging device is following:

First lens converge to the scene incident light on the dmd array; Second lens converge to the reflection ray on the dmd array on single pixel detector; Dmd array will reflex to light on second lens through the micromirror of measuring matrix from the light that first lens converge; Single pixel detector receives the light of dmd array reflection, converts light signal into electric signal, exports to A/D; A/D receives the electric signal of single pixel detector output, carries out mould/number conversion, and digital signal is exported to image reconstruction module; Image reconstruction module receives the digital signal of A/D output; Digital signal to receiving is carried out image reconstruction; Concrete grammar is: select the measurement matrix of DMD, the digital signal that receives is carried out rarefaction representation, and obtain the transition matrix in the rarefaction representation; Ask constraint minimum to 1 norm, thus the image that obtains rebuilding.

Image reconstruction module comprises that measuring matrix selects module, rarefaction representation module, transition matrix module and norm least commitment module; Wherein measuring matrix selects module to be used to select the measurement matrix of DMD; The rarefaction representation module is used for the digital signal that receives is carried out rarefaction representation; The transition matrix module is used for obtaining the transition matrix of rarefaction representation, and norm least commitment module is used for asking constraint minimum to 1 norm, and the detailed operation principle of image reconstruction module is following:

(1) at first selects suitable measurement matrix.In encoding measurement, select stable projection matrix, in order to ensure the linear projection of signal can holding signal prototype structure, (sign of approximation ratio between measurement result and the original image just), projection matrix must satisfy constraint isometry condition.The linear projection measured value is made up of original signal and the product form of measuring matrix.

Measure matrix Φ and must satisfy constraint isometry condition.For any K sparse signal c and constant, σ _K∈ (0,1) if

$(1-{\mathrm{\σ}}_K){\left|\right|c\left|\right|}_2^2\≤{{\left|\right|\mathrm{\Φ}}_{T}c\left|\right|}_2^2\≤(1+{\mathrm{\δ}}_K){\left|\right|c\left|\right|}_2^2,\∀c\∈R^{\left|T\right|}---\left(1\right)$

Set up, wherein | T|≤K, Φ _TFor the size that constitutes by the indicated related column of index T among the Φ be K * | the submatrix of T|, claim that then matrix Φ satisfies the constraint isometry.Usually, for a k-factor signal x (position of its K nonzero value is unknown), formula (5) can be that matrix Φ is for 3K sparse signal c and constant δ from the adequate condition that y accurately reconstructs x _3K∈ (0,1) has 3K rank constraint isometry, promptly

$(1-{\mathrm{\σ}}_{3K}){\left|\right|c\left|\right|}_2^2\≤{{\left|\right|\mathrm{\Φ}}_{T}c\left|\right|}_2^2\≤(1+{\mathrm{\δ}}_{3K}){\left|\right|c\left|\right|}_2^2,\∀c\∈R^{\left|T\right|}---\left(2\right)$

Set up wherein

| T|≤3K.Can obtain through the Gauss measurement matrix that to select a size be M * N, wherein each value all satisfy N (0, independent normal distribution 1/N).

(2) prerequisite of compressed sensing is that the hypothesis signal has sparse property and compressibility, and promptly signal can have succinct expression formula under suitable base.Rarefaction representation has surmounted traditional Shannon's sampling theorem, and sampled signal can precisely be rebuild original signal on the basis of Ya Nai Qwest sample mode.The rarefaction representation of signal shows that having only the minority element in this signal is non-zero.In most of the cases all right and wrong are sparse for the natural sign in the time domain, but it possibly be sparse on transform domain.Be transformed to sparse signal, can effectively represent whole signal with the coefficient of minority.

Theoretical according to harmonic analysis, the one-dimensional discrete time signal f that length is N can be expressed as the linear combination of one group of orthonormal basis

$f=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{x}_i{\mathrm{\ψ}}_i---\left(3\right)$

Can be reduced to

f＝xψ，ψ＝[ψ ₁|ψ ₂|…ψ _N] (4)

In the formula (3), ψ _iBe column vector.If x has only K element non-zero, claim that then x is the K rarefaction representation of signal f.Briefly, rarefaction representation is exactly that the absolute value of most conversion coefficients was very little when signal projection was arrived the orthogonal transformation base, and resulting conversion vector is sparse or approximate sparse, can it be regarded as a kind of succinct expression of original signal.Rarefaction representation is the theoretical foundation of compressed sensing, and natural sign is inevitable can be similar to rarefaction representation under certain conversion.

(3) the measurement result matrix of y representation signal M; S represents the matrix of reconstruction signal N, has so

y _j=<x, Φ _j>, and y=Φ x=Φ ψ s=Θ s (5)

Wherein, Φ is for measuring matrix, and ψ is the wavelet basis that projects to.

(4) reconstructed image, as follows

The thought of promptly minimizing with 1 norm is come reconstructed image S.

The content of not doing to describe in detail in the utility model instructions belongs to this area professional and technical personnel's known technology.

Claims (1)

1. based on the theoretical camera space imaging device of compressed sensing, it is characterized in that: comprise first lens, second lens, dmd array, single pixel detector, A/D and image reconstruction module, wherein:

First lens: the scene incident light is converged on the dmd array;

Second lens: the reflection ray on the dmd array is converged on single pixel detector;

Dmd array: comprise addressing integrated chip and the micromirror of measuring matrix, will reflex to light on second lens through the micromirror of measuring matrix from the light that first lens converge;

Single pixel detector: receive the light of dmd array reflection, convert light signal into electric signal, export to A/D;

A/D: receive the electric signal of single pixel detector output, carry out mould/number conversion, digital signal is exported to image reconstruction module;

Image reconstruction module: the digital signal that receives A/D output; Digital signal to receiving is carried out image reconstruction; Concrete grammar is: select the measurement matrix of DMD, the digital signal that receives is carried out rarefaction representation, and obtain the transition matrix in the rarefaction representation; Ask constraint minimum to 1 norm, thus the image that obtains rebuilding.

Images