CN104142506A - Laser radar imaging system based on compressed sensing - Google Patents

Abstract

The invention discloses a laser radar imaging system based on compressed sensing. A unit APD is adopted for the system, so that the bottleneck that large-scale integration of domestic line array APDs cannot be achieved at present is effectively broken through. The system is formed by a laser emission module, a telescope imaging module, a DMD, a control module, the APD, a synchronization module, a data acquisition module and an image reconstruction module. In the system, pulse laser is emitted to a target through the laser emission module, target reflection echoes are modulated through the DMD, sampling of the unit APD on a time series is achieved through a convergence lens, and finally a three-dimensional image of the target is reconstructed on the basis of a compressed sensing theory by means of related algorithms. The laser radar imaging system based on compressed sensing has the advantages that scanning is not needed, the structure is simple, the data size needed for image reconstruction is small, and detection sensitivity is high.

Description

A kind of laser infrared radar imaging system based on compressed sensing

Technical field

The present invention relates to calculate imaging technique and image reconstruction algorithm, signal processing, laser radar.Be particularly related to a kind of laser infrared radar imaging system based on compressed sensing.

Background technology

Laser radar is a kind of initiatively photoelectric imaging technology, compares that to have resolution high with common passive optical remote sensing with microwave radar, good concealment, extremely strong antijamming capability etc.; Can be through cloud and mist, vegetation etc. are detected real ground surface or terrain.Laser radar is by target emission pulse laser signal, then carries out relevant data processing by receiving the signal (echoed signal) that reflects from target to transmitting, thereby just can extract the relevant information of target, such as target range, orientation, attitude, the parameters such as shape.Utilize this laser radar, militarily just can be to enemy's aircraft, guided missiles etc. are followed the tracks of, detection and identify, thereby realize precision strike.It has become a kind of indispensable technological means in China's military field at present.

Traditional laser radar can be divided into sweep type and the linear array push-broom type of point by point scanning by working method.Sweep type laser radar is very ripe technically, and the advantage of its maximum is that principle is very simple.But also there is very large shortcoming in it, such as the target of the high-speed mobile that defies capture; Owing to there is mechanical scanner, be difficult to accomplish miniaturization and lightness; The data transmission to data acquisition system (DAS) and storage and subsequent treatment are brought great pressure by a large amount of cloud datas; The principle of point by point scanning in addition, and the restriction of flying speed and sweep velocity will cause the spatial resolution of range image lower.Linear array push-broom type laser radar adopts the parallel detecting principle of simultaneously launching multiple laser and multiple detectors, covers efficiency and scan efficiency thereby improve, and overcomes some shortcomings of point by point scanning formula laser radar.At present, the research of China's push-broom type laser radar is just in the starting stage.And the avalanche diode APD detector of linear array is difficult to accomplish large-scale integrated, with regard to current technological means, can only accomplish the APD of 25-50 unit, the technique bottleneck problem of linear array APD will hinder the development of push-broom type laser radar to a great extent.

To be a kind of sampling of being proposed in 2006 by people's (referring to document 1,2,3) such as the mathematician Donoho of Stanford Univ USA and Candes synchronize the theory of carrying out with compressing to compressed sensing (Compressive Sensing, CS).This theory is by excavating redundancy and the sparse property of signal message, in sampling process, not the whole pixel samplings that obtain image, but by specific algorithm, select suitable modulation template, that is: observing matrix carries out overall situation sampling to signal at every turn, then samples in conjunction with relevant recovery algorithms restored image by these.Different from traditional " first sampling, rear compression ", CS theory is the mode of " limit sampling, limit compression ", CS is applied to laser infrared radar imaging system and can significantly saves number of sensors, the mode of this " limit sampling, limit compression " makes the technology burden of signal processing transfer to data processing from sensor.Therefore,, based on compressive sensing theory, the laser infrared radar imaging system of Development of Novel naturally becomes the content that the present invention will study.

List of references:

[1]Donoho D L.Compressed sensing[J].IEEE Transactions on Information Theory，2006，52(4)：1289-1306.

[2]Candès E，Romberg J，Tao T.Robust uncertainty principles：exact signal reconstruction from highly incomplete frequency information[J].IEEE Transactions on Information Theory，2006，52(2)：489-509.

[3]Candès E.Compressive sampling[C].International Congress of Mathematics，2006：1433-1452.

Summary of the invention

The object of this invention is to provide a kind of laser infrared radar imaging system based on compressed sensing.Aspect detector, adopt unit avalanche diode APD detector, overcome the shortcoming and the bottleneck problem of avoiding linear array APD technical matters of traditional point by point scanning formula laser radar.Aspect data acquisition, based on compressive sensing theory, adopting a small amount of data is the 3-D view that restructural obtains target, in the process of sampling just with compressed data, collection, transmission, the storage pressure of big data quantity in the radar imagery of alleviation conventional laser.

The solution thinking that the present invention proposes is as follows:

As shown in Figure 1, this new pattern laser radar imaging system comprises: laser emitting module 1; Telescope image-forming module 2; Digital micro-mirror DMD and control module 3; Optics plus lens 4; Avalanche diode APD 5; Data acquisition module 6; Image Reconstruction module 7; Synchronization module 8.It is characterized in that: laser emitting module 1 adopts the pulsed laser of wavelength 1064nm its repetition frequency 100Hz, pulse energy 200mJ; It is 304.8mm that telescope image-forming module 2 adopts focal length, the telescope that bore is 101.6mm; Digital micro-mirror DMD in digital micro-mirror DMD and control module 3 adopts 1024 × 768 pixels, and pixel size is the DMD of 13.69 μ m; The focal length of optics plus lens 4 is 10cm; The pixel dimension 1.5mm of avalanche diode APD 5, dark current 7nA, rise time 5ns; The capture card quantization digit that data acquisition module 6 adopts is 10, sampling rate 5GSPS; Synchronization module 8 adopts fpga chip to produce three tunnel synchronizing signals;

Workflow between the each module of system is as follows:

Synchronization module 8 is launched synchronizing signal to laser emitting module 1 and digital micro-mirror DMD and control module 3, laser emitting module 1 is received after synchronizing signal, start to scene objects emission pulse laser, set scene target has k, and the echoed signal being reflected by target is designated as successively: x ¹, x ²... x ^k;

Digital micro-mirror DMD and control module 3 are also received synchronizing signal simultaneously, then load a modulation template, send to DMD, and the sum of modulation template is set as M.Modulation template when the M time modulation is designated as: θ _m, concrete value is a m × n rank matrix setting in advance, and the value of matrix element is 0 or 1, and all elements are obeyed gaussian random and are distributed.Change the rollover states of DMD micro mirror by modulation template, thereby reach the effect of modulation target echo.In fact θ _mbe exactly the observing matrix in compressive sensing theory, the span of M is s is signal x ¹degree of rarefication;

Echoed signal after DMD modulation is converged on avalanche diode APD 5 by optics plus lens 4.In the modulated process each time of DMD, the target echo signal of different distance arrives the asynchronism(-nization) on APD, and the time is designated as successively: can there are successively multiple peak values in the signal therefore finally detecting on APD detector, as shown in Figure 2, and the corresponding target of each peak value;

The signal that avalanche diode APD 5 detects is after data acquisition module 6 gathers, in time series upper, obtain successively corresponding M group digital signal value:

The signal that Image Reconstruction module 7 collects data acquisition module 6 is processed, and finally obtains the three-dimensional image of each target;

The specific implementation step of Image Reconstruction module 7 is as follows:

1), for first aim, data acquisition module 6 collects signal and is write as (1) formula as follows:

$\left\{\begin{array}{c}{f_1}^1={\mathrm{\θ}}_1\·x^1\\ f_2^1={\mathrm{\θ}}_2\·x^1\\ ...\\ f_M^1={\mathrm{\θ}}_M\·x^1\end{array}\right.---\left(1\right)$

(1) formula is rewritten as to the matrix equation of (2) formula as follows:

F ¹＝Θ·X ¹ (2)

In above formula, F ¹it is signal 1 M × 1 matrix forming; Θ is M × N matrix, and line number M is modulation template number, and columns N=m × n is each modulation template θ _mthe total number of first number, every a line of Θ is by corresponding modulation template θ _mrearrange and form; X ¹for N × 1 matrix; Based on compressive sensing theory, the value of M is far smaller than N.Therefore, (2) formula is actually an ill-condition equation.Direct solution clearly has infinite multiple solution.But compressive sensing theory is pointed out, as long as X ¹be sparse, or under the expression of certain orthogonal transformation, have sparse property, (2) formula that solves so will have special optimization method.The meaning of sparse property refers to and wherein comprises a large amount of data that go to zero, only has a small amount of nonzero value;

For natural scene target, generally, can under the expression of some orthogonal transformations, there is sparse property.For example: Fourier transform, discrete cosine transform etc.For X ¹, under discrete cosine transform, be following (3) formula by its rarefaction representation:

X ¹＝Ψ·α ¹ (3)

In above formula, α ¹for X ¹rarefaction representation, it is N × 1 matrix; Ψ is N × N rank discrete cosine transform matrix;

So (2) formula is written as to (4) formula as follows again:

F ¹＝Θ·X ¹＝Θ·Ψ·α ¹＝T·α ¹ (4)

In above formula, T is the sensing matrix on M × N rank; Wherein only has α ¹for unknown number;

The method of Image Reconstruction solves the sparse factor alpha in (4) formula exactly ¹.Clearly (4) formula is actually an ill-condition equation.Direct solution has infinite multiple solution, is therefore translated into the optimization problem as shown in the formula (5):

${\widehat{\mathrm{\α}}}^1=\arg \min {\left|\right|{\mathrm{\α}}^1\left|\right|}_{L_1},\mathrm{ST}.F=T\·{\mathrm{\α}}^1---\left(5\right)$

In above formula, L ₁represent 1 norm, for α ¹best fit approximation solution;

(5) the Optimization Solution algorithm steps of formula is as follows:

The first step: empty matrix I=[of initialization], residual matrix R=F;

Bis-Walk: each row in residual error R and T are done respectively to inner product, and find those row of inner product maximum, these row are taken out and add in matrix I;

Tri-Walk: upgrade residual error, R=F-I (I ^ti) ^-1i ^tf, wherein I ^tfor the transposed matrix (I of I ^ti) ^-1for (I ^ti) inverse matrix;

The 4th step: constantly sequential loop bis-Walk and the 3rd step, cycle index is C, its span is: C >=2M;

The 5th step: the solution that final (5) formula is tried to achieve is following (6) formula:

${\widehat{\mathrm{\α}}}^1={(I^T\·I)}^{-1}\·I^T\·F---\left(6\right)$

The image information of the first aim of finally trying to achieve is expressed as (7) formula as follows:

$X^1=\mathrm{\Ψ}\·{\widehat{\mathrm{\α}}}^1---\left(7\right)$

By rank, the N × 1 matrix X in (7) formula ¹be rearranged into the two-dimensional image that m × n rank matrix can obtain target;

2) for second target, by step 1) in (1) formula be rewritten as (8) formula as follows:

$\left\{\begin{array}{c}{f_1}^2={\mathrm{\θ}}_1\·x^2\\ f_2^2={\mathrm{\θ}}_2\·x^2\\ ...\\ f_M^2={\mathrm{\θ}}_M\·x^2\end{array}\right.---\left(8\right)$

(8) formula is written as to the matrix equation of (9) formula as follows:

F ²＝Θ·X ² (9)

Solve the same step 1) of method of (9) formula, the image information of therefore finally trying to achieve second target is expressed as (10) formula as follows:

$X^2=\mathrm{\Ψ}\·{\widehat{\mathrm{\α}}}^2---\left(10\right)$

3), for k target, by step 1) in (1) formula be rewritten as (11) formula as follows:

$\left\{\begin{array}{c}{f_1}^k={\mathrm{\θ}}_1\·x^k\\ f_2^k={\mathrm{\θ}}_2\·x^k\\ ...\\ f_M^k={\mathrm{\θ}}_M\·x^k\end{array}\right.---\left(11\right)$

(11) formula is written as to the matrix equation of (12) formula as follows:

F ^k＝Θ·X ^k (12)

Solve the same step 1) of method of (12) formula, the image information of therefore finally trying to achieve k target is expressed as (13) formula as follows:

$X^k=\mathrm{\Ψ}\·{\widehat{\mathrm{\α}}}^k---\left(13\right)$

By rank, the N × 1 matrix X in (13) formula ^kbe rearranged into the two-dimensional image that m × n rank matrix can obtain target;

4) for the range information of target, the time of data acquisition module (6) record is done on average, obtains following (14) formula:

$\left\{\begin{array}{c}{T}^1=\frac{t_1^1+t_2^1+...+t_M^1}{M}\\ T^2=\frac{t_1^2+t_2^2+...+t_M^2}{M}\\ .\\ .\\ .\\ T^k=\frac{t_1^k+t_2^k+...+t_M^k}{M}\end{array}\right.---\left(14\right)$

In above formula, T ¹be the temporal information of the 1st target, the like T ^kit is the range information of k target;

Then the range information that obtains target is following (15) formula:

$\left\{\begin{array}{c}{d}^1=\frac{1}{2}c\·T^1\\ d^2=\frac{1}{2}c\·T^2\\ .\\ .\\ .\\ d^k=\frac{1}{2}c\·T^k\end{array}\right.---\left(15\right)$

In above formula, d ¹be the range information of the 1st target, the like d ^kit is the range information of k target;

So far,, by formula (13), (15) formula, can obtain the three-dimensional image data of all targets.

The invention has the advantages that:

(1) the present invention adopts the echoed signal of DMD modulation target, compared with traditional point by point scanning formula laser radar, cancels mechanical scanner, realizes miniaturization and the lightweight of laser radar system, has very strong shock resistance.The required sampled data of simultaneously 3-D view reconstruct is few.

(2) system of the present invention adopts unit avalanche diode APD as detector, compared with traditional push-broom type laser radar, the bottleneck problem that the extensive avalanche diode APD of domestic linear array cannot be integrated in technique will be overcome, solve push-broom type laser radar signal to noise ratio (S/N ratio) low, the problem such as detection sensitivity is low simultaneously.

Brief description of the drawings

Fig. 1 is a kind of laser infrared radar imaging system and method based on compressed sensing, the 1st, and laser emitting module; The 2nd, telescope image-forming module; The 3rd, digital micro-mirror DMD and control module; The 4th, optics plus lens; The 5th, avalanche diode APD; The 6th, data acquisition module; The 7th, Image Reconstruction module; The 8th, synchronization module;

Fig. 2 is the signal form that target echo receives at every turn on APD, and wherein figure (a) is modulation signal for the first time, and figure (b) is the M time modulation signal.

Embodiment

Provide a better example of the present invention below in conjunction with Fig. 1, be mainly described in further detail, but not be used for limiting scope of the present invention.

The specific embodiment of the present invention is mainly divided into the following steps:

(1) technical parameter of the first definite each module of system component used, specific as follows: laser emitting module 1 adopts the laser instrument of Shanghai Brillouin laser Science and Technology Ltd., technical indicator is: operation wavelength 1064nm, repetition frequency 100Hz, pulse energy 200mJ; Telescope image-forming module 2 adopts the telescope of Ai Mengte optics (Shenzhen) company limited, and selected focal length is 304.8mm, and bore is 101.6mm.The DMD that digital micro-mirror DMD and control module 3 adopt American TI Company to produce, technical indicator is 1024 × 768 pixels, and pixel size is 13.69 μ m, and control panel adopts the TI-Discovery-4100 coordinating with it; Optics plus lens 4 focal lengths are 10cm; The AD1500-10 that avalanche diode APD 5 adopts Pacific Silicon Sensor company of the U.S. to produce, pixel dimension 1.5mm, dark current 7nA, rise time 5ns, responsiveness 36A/w; Data acquisition module 6 adopts female QT1230 capture card of speeding scientific and technological, and technical indicator is: 10 of quantization digits, and sampling rate is up to 5GSPS; Synchronization module 8 adopts the Spartan-6-XC6SLX9 chip of company of match SEL to produce three tunnel synchronizing signals;

(2) after system hardware is determined, be first that synchronization module 8 is launched synchronous signal impulse, after laser emitting module 1 is received synchronous signal impulse, to target emission pulse laser.

(3) simultaneously, after digital micro-mirror DMD and control module 3 receive synchronization pulse, load a modulation template and send to DMD, modulation template is that a size is 256 × 192, and obey 0 of gaussian random distribution, 1 matrix, these modulation template are generated in advance, and this implements to generate 5000 templates.Due to step (1), selected DMD is 1024 × 768 pixels, and in order to allow DMD correctly identify, it is 1024 × 768 that actual loaded is chosen size to the template of DMD, exceeds 256 × 192 element and all uses " 0 " completion.By modulation, make the micro mirror of DMD in certain on off state, wherein " open " use " 1 " and represent, " passs " is with " 0 " expression (specifically make micro mirror overturn+12 ° (opening) and-12 ° (pass)).Deng micro mirror on off state stable after, can realize the modulation to echo target, the light of-12 ° of reflections of upset is dropped, the light of+12 ° of reflections of upset then enters follow-up optical system.Then, data acquisition module 6 carries out data acquisition.(1) (2) (3) step repeats 5000 times, gets final product data acquisition.

(4) last Image Reconstruction module (7) is processed the three-dimensional image that can obtain target to sampled data.For brief description, set scene has 5 targets, and the spatial resolution of each target is: 256 × 192 pixel sizes.Each target is carried out to space two-dimensional signal reconstruct, need to solve 256 × 192 unknown numbers, and sampled data only has 5000, be equivalent to data acquisition module 6 data in the process of sampling compressed 256 × 192-5000=44152.To the range information reconstruct of each target, solve 5 range informations.

Claims (1)

1. the laser infrared radar imaging system based on compressed sensing, it comprises: laser emitting module (1); Telescope image-forming module (2); Digital micro-mirror DMD and control module (3); Optics plus lens (4); Avalanche diode APD (5); Data acquisition module (6); Image Reconstruction module (7); Synchronization module (8); It is characterized in that:

Described laser emitting module (1) adopts the pulsed laser of wavelength 1064nm, its repetition frequency 100Hz, pulse energy 200mJ;

It is 304.8mm that described telescope image-forming module (2) adopts focal length, the telescope that bore is 101.6mm;

Digital micro-mirror DMD in described digital micro-mirror DMD and control module (3) adopts 1024 × 768 pixels, and pixel size is the DMD of 13.69 μ m;

The capture card quantization digit that described data acquisition module (6) adopts is 10, sampling rate 5GSPS;

Described synchronization module (8) adopts fpga chip to produce three tunnel synchronizing signals;

Synchronization module (8) transmitting synchronizing signal is to laser emitting module (1) and digital micro-mirror DMD and control module (3), laser emitting module (1) is received after synchronizing signal, start to scene objects emission pulse laser, set scene target has k, and the echoed signal being reflected by target is designated as successively: x ¹, x ²... x ^k;

Digital micro-mirror DMD and control module (3) are also received synchronizing signal simultaneously, then load a modulation template by control module, send to DMD, and the sum of modulation template is set as M, and modulation template when the M time modulation is designated as: θ _m, concrete value is m × n rank matrix that a gaussian random distributes, the value of matrix element be 0 or the span of 1, M be: wherein N=m × n, S is x ^kdegree of rarefication;

Echoed signal after DMD modulation is converged on avalanche diode APD (5) by optics plus lens (4).In the modulated process each time of DMD, the target echo signal of different distance arrives the asynchronism(-nization) on APD, and the time is designated as successively: can there are multiple peak values in the final signal detecting on APD detector, each peak value corresponding target object successively;

The signal that avalanche diode APD (5) detects is after data acquisition module (6) gathers, in time series upper, obtain successively corresponding M group digital signal value:

Finally, the signal that Image Reconstruction module (7) collects data acquisition module (6) is processed, and finally obtains the three-dimensional image of each target.