CN102042878A - Infared nonuniformity correction method for removing temperature shift - Google Patents

Abstract

The invention discloses an infared nonuniformity correction method for removing temperature shift. In the method, the correction gain and offset of a two-point correction algorithm are solved first, and then correction offset of a one-point correction algorithm is solved; and image data during the solving of offset parameters of the one-point correction algorithm is from the data solved after two-point correction. The method has the advantages of low cost, wide correction dynamic range, good effect, and simple principle and the characteristics of high efficiency, accuracy and the like, can be easily implemented in the aspect of hardware, and is suitable to be used in a focal plane imaging system. The technology is technical improvement on uniformity of an infared focal plane and contributes to promoting the good performance of an infared focal plane system.

Description

A kind ofly remove the infrared asymmetric correction method that temperature is floated

Technical field

The present invention relates to non-refrigeration infrared detection technique field, be specifically related to a kind of infrared asymmetric correction method that temperature is floated of removing.

Background technology

Infrared focal plane array is beginning of the eighties late 1970s, grows up under the promotion of national defense applications and the application of other strategy and tactics.It is the important photoelectric device that obtains scenery infrared emanation information.Except that being applied to traditional military imaging, also be widely used in fields such as industrial automatic control, medical diagnosis, chemical process monitoring, infrared astronomy.

Infrared eye is an infrared detection technique development most active fields, is the important component part of infra-red thermal imaging system.The photon type infrared eye of comparative maturity has been applied to fields such as communication, medical science, military affairs and industry widely at present.

Infrared focal plane array (IRFPA) belongs to second generation infrared imaging device, is the core of modern infrared imaging system, with respect to previous generation's infrared imaging system, and advantage such as simple in structure, working stability, noise equivalent temperature difference are little, sensitivity height that it has.But owing to be subjected to the restriction of manufacturing materials and manufacture craft, the response characteristic of each pixel of infrared focal plane array can't be accomplished in full accord, has heterogeneity, has had a strong impact on the detection sensitivity and the spatial resolution of infrared imaging system.Therefore, the infrared focal plane array that uses in the engineering all needs heterogeneity is made correction.

Infrared system in the ideal case, infrared focal plane array is subjected to homogeneous radiation, output amplitude should be just the same.But in fact, because influences such as the semiconductor material inhomogeneous (unevenness of impurity concentration, crystal defect, inner structure etc.) of making device, mask error, defective, process conditions down, its output amplitude is also inequality, the heterogeneity of infrared focal plane array response that Here it is.

The reason that infrared heterogeneity causes has a variety of, and wherein chief component is the heterogeneity that thermal imagery is surveyed unit self, and the extraneous input of infrared focal plane array also can cause influence heterogeneous in addition.As the bias voltage of detector, the difference of bias current, also will cause the unevenness of output, mainly show as fixedly additive noise.

At present, infrared Nonuniformity Correction technology commonly used has a variety of, as a point calibration, the non-homogeneous algorithm of two point calibrations based on calibration, also has based on the time domain high-pass filtering method of scene, adaptive artificial neural network method and mean filter algorithm etc.Though it is a variety of that the Nonuniformity Correction algorithm of thermal imaging has, also do not find the stronger algorithm of a kind of adaptability at present, various nonuniformity correction algorithms all have its deficiency.What extensively be applied to putting into practice at present is a bit to calibrate correcting algorithm and two-point calibration correcting algorithm.But the correction accuracy of some scaling methods is lower, and the choosing of second scaling point of two-point calibration method has difficulties.

One point calibration method is a Nonuniformity Correction algorithm the earliest, the inhomogeneous and inhomogeneous two kinds of situations of biasing at gain coefficient, and a point calibration method also can be divided into two kinds, is divided into the uneven correction of gain coefficient and the uneven correction of setovering.But once can only satisfy a bit.

The inhomogeneous correction of employing to setovering, some correction principles are to suppose that the radiation flux of the output signal of the photosensitive unit of detector and target is linear, that is:

V _ij(φ _ij)＝R _ijφ _ij+θ _ij

In the formula, φ _Ij-----incide (i, j) radiation flux of individual photosensitive unit

V _Ij(φ _Ij)-----Di (i, j) output voltage of individual photosensitive unit

R _Ij------(i, j) responsiveness factor of individual photosensitive unit

θ _Ij-----Di (i, j) the intercept factor of individual photosensitive unit

So-called some correcting algorithms are exactly under even optical radiation, the output signal of each pixel is proofreaied and correct to consistent, and promptly under a certain photoirradiation, it is a certain signal that the output signal of different pixels is proofreaied and correct, this signal can be the signal averaging of this moment, also the maximal value under the condition etc. for this reason.Generally average.Suppose that infrared focal plane array is of a size of S=M * N (M, N are positive integer, respectively the line number and the columns of the photosensitive unit of presentation surface battle array), chooses the irradiance under certain temperature

As scaling point, the output signal V of all photosensitive units of opposite battle array _Ij(φ), average:

In the formula,

For infrared focal plane array at scaling point

The roomage response average at place, S is a detector unit number in the face array,

With

Be respectively the responsiveness factor and section the average of each photosensitive unit in the focal plane arrays (FPA) according to the factor.

The output of any photosensitive unit with the difference of mean value is

$O_{\mathrm{ij}}\left(\mathrm{\φ}\right)=V_{\mathrm{ij}}\left(\mathrm{\φ}\right)-\stackrel{\‾}{V}\left(\mathrm{\φ}\right),(i=\mathrm{0,1},......,N-1;j=\mathrm{0,1},......,M-1.)$

The output valve of proofreading and correct any photosensitive unit in back is

$V_{\mathrm{ij}}^{\′}\left(\mathrm{\φ}\right)=V_{\mathrm{ij}}\left(\mathrm{\φ}\right)-O_{\mathrm{ij}}\left(\mathrm{\φ}\right),(i=\mathrm{0,1},......,N-1;j=\mathrm{0,1},......,M-1.)$

The essence of one point calibration has just been done compensation to the dark current of device, correction is not made in gain, the compensation that we only also just utilize a point calibration that output signal is setovered.

The gain inequality of detector, the factors such as unevenness, thermonoise, black level of setovering have been considered in peg method.Generally, when the incident infrared radiation is zero, the black level that the response of detector output is non-vanishing, just usually said.Peg method is the same with a point calibration method also to be a kind of linearity correction algorithm

V _ij(φ _ij)＝R _ijφ _ij+θ _ij

Choose two kinds of irradiance

With

Calibration just can be determined correction factor R _Ij, θ _IjCan utilize correction factor to carry out the heterogeneity compensation then, then at arbitrary irradiance Down, the output V of each pixel _Ij(φ) can proofread and correct and be V _Ij' (φ):

$\left\{\begin{array}{c}\stackrel{\‾}{V}\left({\mathrm{\φ}}_1\right)=R_{\mathrm{ij}}\left({\mathrm{\φ}}_1\right)+{\mathrm{\θ}}_{\mathrm{ij}}\\ \stackrel{\‾}{V}\left({\mathrm{\φ}}_2\right)=R_{\mathrm{ij}}\left({\mathrm{\φ}}_2\right)+{\mathrm{\θ}}_{\mathrm{ij}}\end{array}\right.$ $\left\{\begin{array}{c}{R}_{\mathrm{ij}}=\frac{\stackrel{\‾}{V}\left({\mathrm{\φ}}_2\right)-\stackrel{\‾}{V}\left({\mathrm{\φ}}_1\right)}{V_i\left({\mathrm{\φ}}_2\right)-V_i\left({\mathrm{\φ}}_1\right)}\\ {\mathrm{\θ}}_{\mathrm{ij}}=\stackrel{\‾}{V}\left({\mathrm{\φ}}_1\right)-R_{\mathrm{ij}}{V}_i\left({\mathrm{\φ}}_1\right)\end{array}\right.$

V _ij′(φ)＝R _ijV _ij(φ)+θ _ij

Peg method be exactly with the resonse characteristic of all probe units by the rotation translation, be transformed to same bar response family curve L.After calibrated, under uniform radiation input condition, the output electric signal of each probe unit is identical, thereby has eliminated the heterogeneity noise of infrared image.Peg method does not have any requirement to object content except that the need Temperature Scaling, comparatively desirable at aspects such as volume, quality, power consumption, costs yet.

Summary of the invention

Problem to be solved by this invention is: how a kind of infrared asymmetric correction method that temperature is floated of removing is provided, this method is to add a point calibration with two point calibrations to be used in combination, it is than only more accurate with a point calibration, and behind two point calibrations, when the appearance temperature is floated, can come the drift of compensation temperature with the method for a point calibration.

Technical matters proposed by the invention is to solve like this: a kind of infrared asymmetric correction method that temperature is floated of removing is provided, it is characterized in that, may further comprise the steps:

Step 101, beginning;

Step 102 reads the low temperature image, promptly is that the temperature when black matrix is T1

Temperature T 1 time, array average response voltage is:

$\stackrel{\‾}{U_1}=\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{U}_1(i,j)}{M\×N}$

Reading the high temperature image, promptly is that the temperature when black matrix is T2

Temperature T 2 times, array average response voltage is:

$\stackrel{\‾}{U_2}=\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{U}_2(i,j)}{M\×N}$

M, N are positive integer, the line number and the columns of the photosensitive unit of difference presentation surface battle array, (i, j) expression (i, j) photosensitive unit;

Step 103 is calculated each pixel correction coefficient

For each pixel, two correction parameters are arranged, be respectively G _IjBe correcting gain and O _IjBe corrects bias:

$G_{\mathrm{ij}}=\frac{\stackrel{\‾}{U_2}-\stackrel{\‾}{U_1}}{U_2(i,j)-U_1(i,j)}$

$O_{\mathrm{ij}}=\stackrel{\‾}{U_1}-G_{\mathrm{ij}}\×U_1(i,j);$

Step 104 generates the correction parameter table, two parameters is placed on respectively among two flash, and deposits according to increasing progressively successively of pixel number, calls over gain and displacement factor when correction program moves;

Whether step 105 has temperature to float, and does not have temperature to float then and carry out step 107, has temperature to winnow and selects step 108;

Step 106 reads image to be corrected;

Step 107 utilizes the parameter of step 104 that original image is carried out peg method;

Step 108 is chosen irradiance φ _MAs scaling point;

Step 109 is done two point calibrations to the scaling point of choosing, and utilizes the parameter and the formula V ' of step 104 _Ij(φ _M)=G _Ij* V _Ij(φ _M)+O _Ij, drawing and proofread and correct back output, battle array all photosensitive units in opposite average:

According to a point calibration method, must proofread and correct the skew be,

Then obtain at 2 and add some corrects bias O _Ij'=O _Ij-O _Ij(φ _M);

Step 110 is to the image of step 106, according to formula V ' _Ij=G _Ij* V _Ij+ O _Ij' calculation correction result generates and proofreaies and correct the back image;

Step 111 is finished real-time correction.

The present invention be directed to low second problem that scaling point is difficult to determine that reaches two-point calibration of correction accuracy of some calibration, two-point calibration and some calibration combination are used for Nonuniformity Correction, it is than only more accurate with a point calibration, and behind two point calibrations, when the appearance temperature is floated, can come the drift of compensation temperature with the method for a point calibration.Beneficial effect of the present invention:

1, use add a point calibration method at 2, not only can utilize two point calibration dynamic ranges big, effective, principle is simple, the characteristics that are easy to realize, and improved the homogeneity of infrared focus plane significantly.Be easy to realize at hardware aspect simultaneously.

2, the warm infrared nonuniformity correction technology of floating of this removal only need realize on FPGA, and just the effect that can obtain wanting is by once calibration, simple.And do not need to revise hardware, and can better eliminate residual heterogeneity to a certain extent, and can not influence the real-time of system, can reach the real-time correction of system.

Description of drawings

Fig. 1 is that two point calibrations add a point calibration method workflow synoptic diagram.

Described sequence number 101 beginnings;

Low temperature and high temperature image are read in described sequence number 102 representatives;

Each pixel correction coefficient is calculated in described sequence number 103 expressions;

Described sequence number 104 is to generate correction parameter table G and O;

Whether described sequence number 105 has temperature to float;

Described sequence number 106 is to read in original image, image promptly to be corrected;

Described sequence number 107 is that image is carried out 2 nonuniformity corrections;

Described sequence number 108 is chosen irradiance φ _MAs scaling point;

Described sequence number 109 is to utilize the parameter of sequence number 104 to obtain at 2 to add some corrects bias O ';

Described sequence number 110 is to utilize parameter G and O ' to proofread and correct to sequence number 106 original images;

Described sequence number 111 is finished real-time correcting imaging.

Embodiment

Below in conjunction with accompanying drawing the present invention is further described:

Detailed technology scheme of the present invention is:

A kind ofly remove the infrared asymmetric correction method that temperature is floated, 2 algorithms are that compensation has all been done in biasing and gain, and focal plane arrays (FPA) can cause temperature drift after operation a period of time sometimes, at this moment just can compensate the coefficient of setovering with a point calibration.

The correction of infrared focal plane asymmetric can realize that independent FPGA correcting circuit has simply, is easy to advantages such as control with independent FPGA.The hardware components of its corrective system mainly contains five parts and forms: detector, FPGA, analog to digital conversion, external memory storage, terminal show.

Because nonuniformity correction mainly needs to add (subtracting) musical instruments used in a Buddhist or Taoist mass and multiplier, and the FPGA that selects for use contains a large amount of multiplication units that adds, so very help Nonuniformity Correction.The hard-wired process of whole infrared nonuniformity correction is as follows: infrared target is through the identification of detector, relevant information is gathered and is exported, A/D converter spare is strengthened delivering to then to signal in the output back, import FPGA after finishing analog to digital conversion, operation and infrared image through correction program strengthen back output, data are through D/A converter spare, and simulating signal at this moment is input to terminal presentation facility then through amplifier again, and infrared target finally appears on the display.

Remove the warm infrared asymmetric correction method that floats and be divided into for two steps, ask the correcting gain and the corrects bias of two point calibration algorithms earlier, ask the corrects bias of some correcting algorithms again, view data when asking for the offset parameter of some correcting algorithms that is to say that from the data after two point calibrations of trying to achieve previously last correction result is the correction again behind two point calibrations.Because some correcting algorithms have just been done correction to dark current, to not influence of gain, so correction parameter of this method, its correcting gain is the correcting gain of two point calibration algorithms, and corrects bias is both poor of the corrects bias of the corrects bias of two point calibration algorithms and some correcting algorithms.

Two temperature spot T of height are chosen in same peg method earlier _HAnd T _L, obtain two irradiance φ _HAnd φ _LAs scaling point, try to achieve correcting gain and proofread and correct displacement factor, afterwards, choose irradiance φ _MAs scaling point, (φ _MAny value between two scaling points of desirable two-point method is generally got intermediate value), two point calibrations are made in the photosensitive unit output of this scaling point earlier, must proofread and correct back output:

V′ _ij(φ _M)＝G _ij×V _ij(φ _M)+O _ij

Battle array all photosensitive units in opposite average,

$\stackrel{\‾}{V}\left({\mathrm{\φ}}_M\right)=\mathrm{\Σ}{V}_{\mathrm{ij}}^{\′}\left({\mathrm{\φ}}_M\right)/N$

Proofreading and correct skew is:

$O_{\mathrm{ij}}\left({\mathrm{\φ}}_M\right)=V_{\mathrm{ij}}^{\′}\left({\mathrm{\φ}}_M\right)-\stackrel{\‾}{V}\left({\mathrm{\φ}}_M\right)$

Updating formula for any photosensitive unit of face array is:

V′ _ij＝G _ij×V _ij+O _ij-O _ij(φ _M)

Can make O _Ij'=O _Ij-O _Ij(φ _M) last updating formula becomes

V′ _ij＝G _ij×V _ij+O _ij′

Some correction parameter that adds 2 correcting algorithm has two, is respectively correcting gain and corrects bias, and whole corrective system mainly is to use to add takes advantage of module, is fit to very much proofread and correct in real time.

Any adds two point calibration algorithms and realizes mainly in two steps with FPGA:

The first step: this step is the same with two point calibrations, is specially and reads low temperature and high temperature image earlier, calculates each pixel correction coefficient, generates the correction parameter table.To demarcate earlier among good gain factor and displacement factor be loaded into correspondence separately respectively through host computer the flash and and deposit successively, and when correction program move, call over and gain and displacement factor according to increasing progressively of pixel number.

Second step: choosing irradiance is φ _MAs calibration point, the data that A/D gathers are carried out two point calibrations earlier, then the result are directly outputed in the host computer, through calculating out the pairing skew O of each point _Ij(φ _M), and by host computer it is loaded among the pairing flash and according to increasing progressively successively of pixel number and deposits.

Fig. 1 is that embodiment of the present invention a bit adds 2 process flow diagram.This flow process starts from

Step 101, beginning.

Step 102 reads the low temperature image, promptly is that the temperature when black matrix is T1.

Temperature T 1 time, array average response voltage is:

$\stackrel{\‾}{U_1}=\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{U}_1(i,j)}{M\×N}$

Reading the high temperature image, promptly is that the temperature when black matrix is T2

Temperature T 2 times, array average response voltage is:

$\stackrel{\‾}{U_2}=\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{U}_2(i,j)}{M\×N}$

Step 103 is calculated each pixel correction coefficient.

For each pixel, two correction parameters are arranged, be respectively G _IjBe correcting gain and O _IjBe corrects bias:

$G_{\mathrm{ij}}=\frac{\stackrel{\‾}{U_2}-\stackrel{\‾}{U_1}}{U_2(i,j)-U_1(i,j)}$

$O_{\mathrm{ij}}=\stackrel{\‾}{U_1}-G_{\mathrm{ij}}\×U_1(i,j)$

Step 104 generates the correction parameter table, two parameters is placed on respectively among two flash, and deposits according to increasing progressively successively of pixel number, calls over gain and displacement factor when correction program moves.

Whether step 105 has temperature to float, and does not have temperature to float then and carry out step 107, has temperature to winnow and selects step 108

Step 106 reads image to be corrected.

Step 107 utilizes the parameter of step 104 that original image is carried out peg method.

Step 108 is chosen irradiance φ _MAs scaling point.

Step 109 is done two point calibrations to this scaling point, utilizes the parameter and the formula V ' of step 104 _Ij(φ _M)=G _Ij* V _Ij(φ _M)+O _Ij, drawing and proofread and correct back output, battle array all photosensitive units in opposite average. According to a point calibration method, must proofread and correct the skew be, Then obtain at 2 and add some corrects bias O _Ij'=O _Ij-O _Ij(φ _M).

Step 110 is to the image of step 106, according to formula V ' _Ij=G _Ij* V _Ij+ O _Ij' calculation correction result generates and proofreaies and correct the back image.

Step 111 is finished real-time correction

A kind ofly remove the infrared Nonuniformity Correction technology that temperature is floated, can be applied in by on 160 * 120 un-cooled infrared focal plane arrays, each non-refrigerating infrared focal plane unit size is 35um * 35um size, among program input FPGA and hardware, can draw the image of enhancing to original image through blind element compensation and nonuniformity correction by infrared Nonuniformity Correction technology.The image that can also be proofreaied and correct in real time according to the software effect.

According to the structure of non-refrigerating infrared focal plane unit, the material of each ingredient and size different can be combined into embodiment like the many types, describe in detail no longer one by one at this.

Claims (1)

1. remove the infrared asymmetric correction method that temperature is floated for one kind, it is characterized in that, may further comprise the steps:

Step 101, beginning;

Step 102 reads the low temperature image, promptly is that the temperature when black matrix is T1

Temperature T 1 time, array average response voltage is:

$\stackrel{\‾}{U_1}=\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{U}_1(i,j)}{M\×N}$

Reading the high temperature image, promptly is that the temperature when black matrix is T2

Temperature T 2 times, array average response voltage is:

$\stackrel{\‾}{U_2}=\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{U}_2(i,j)}{M\×N}$

M, N are positive integer, the line number and the columns of the photosensitive unit of difference presentation surface battle array, (i, j) expression (i, j) photosensitive unit;

Step 103 is calculated each pixel correction coefficient

For each pixel, two correction parameters are arranged, be respectively G _IjBe correcting gain and O _IjBe corrects bias:

$G_{\mathrm{ij}}=\frac{\stackrel{\‾}{U_2}-\stackrel{\‾}{U_1}}{U_2(i,j)-U_1(i,j)}$

$O_{\mathrm{ij}}={\stackrel{\‾}{U_1}-G}_{\mathrm{ij}}\×U_1(i,j);$

Step 104 generates the correction parameter table, two parameters is placed on respectively among two flash, and deposits according to increasing progressively successively of pixel number, calls over gain and displacement factor when correction program moves;

Whether step 105 has temperature to float, and does not have temperature to float then and carry out step 107, has temperature to winnow and selects step 108;

Step 106 reads image to be corrected;

Step 107 utilizes the parameter of step 104 that original image is carried out peg method;

Step 108 is chosen irradiance φ _MAs scaling point;

Step 109 is done two point calibrations to the scaling point of choosing, and utilizes the parameter and the formula V ' of step 104 _Ij(φ _M)=G _Ij* V _Ij(φ _M)+O _Ij, drawing and proofread and correct back output, battle array all photosensitive units in opposite average:

According to a point calibration method, must proofread and correct the skew be,

Then obtain at 2 and add some corrects bias O _Ij'=O _Ij-O _Ij(φ _M);

Step 110 is to the image of step 106, according to formula V ' _Ij=G _Ij* V _Ij+ O _Ij' calculation correction result generates and proofreaies and correct the back image;

Step 111 is finished real-time correction.

Images