CN104406699A - Infrared thermal imager based on adaptive infrared image correction algorithm - Google Patents

Abstract

The invention relates to an infrared thermal imager based on an adaptive infrared image correction algorithm. The problem to be resolved is that non-uniformity of an infrared focal plane array exists due to the fact that elements of the infrared focal plane array are inconsistent in response generally resulting from materials, processes and the like. A real-time adaptive infrared image non-uniformity correction method based on traditional two-point calibration correction is adopted and overcomes the shortcomings of repetitive calibration of traditional two-point correction to enable a system to track change of input signals during operation and keep optimal correction effects, so that real-time adaptive correction of infrared images is achieved. The problems of gradual burring and ghosting in a traditional correction method are improved to a great degree by adding correction judgment during coefficient correction.

Description

Based on self-adaptation infrared image correcting algorithm thermal infrared imager

Technical field

The present invention relates to infrared imagery technique field, especially a kind of based on self-adaptation infrared image correcting algorithm thermal infrared imager.

Background technology

Infrared focal plane device is the core component of Military IR Thermal Imaging system, and to be developed country be ensures the high-tech product that its military strength competitively develops, the gordian technique that the western countries such as the Ye Shi U.S. enforce a blockade to China/limit.At civil area, infrared focal plane detector is widely used in space flight, industry, agricultural, medical science and scientific research, is Infrared Imaging Spectrometer, the core component in the instruments such as infrared camera.Along with the raising of focus planardetector technological level, the scale of infrared focus plane has been extended to pixel up to a million.But inconsistent accordingly owing to detecting between unit, result in the fixed pattern noise of detector array, be usually expressed as striated, latticed or mottled sense pattern noise.

In the ideal case, when infrared focal plane array is by homogeneous radiation, output amplitude should be just the same.And in fact, due to the unevenness of the processing technology of detector, material, temperature and bias conditions, cause output amplitude not identical, namely infrared focal plane array responds the inconsistency of output between each pixel when the input of same homogeneous radiation field, the external world, the heterogeneity of Here it is so-called infrared focal plane array.From the angle of noise, infrared focal plane array noise equals the summation of transient noise and spatial noise.Transient noise is photon noise, dark current noise and the coefficient result of sensing circuit noise.And the heterogeneity that spatial noise is infrared focal plane array causes, also referred to as intrinsic spatial noise.Transient noise can by repeatedly asking on average eliminating of measured value, and intrinsic spatial noise just can must be eliminated by correcting.

The typical image signal characteristic of infrared focal plane array equipment is infrared detecting unit corresponding to different elementary area.Compared with how far detector, owing to have employed the signal processing technologies such as multipath transmission, overall imaging performance is obtained significantly improve, but owing to abandoning amplifier and detector mode of operation one to one, higher technical requirement is proposed to the spatially uniform of infrared focal plane array equipment.Present stage, the overall imaging performance of infrared focal plane array mainly limited by detector fixed pattern noise, and not by the restriction of temporal noise.The heterogeneity of infrared imaging limits infrared focal plane array image-forming systematic difference, and non-uniformity reduces the temperature resolution of infrared imaging system, and the visuality of target image is had a strong impact on.Solve the non-uniformity of infrared focal plane array, just must analyse in depth various heteropical source and the form of expression thereof, work out effective alignment technique and later stage algorithm correction recovery is carried out to it.

Summary of the invention

The present invention will solve the shortcoming of above-mentioned prior art, provide a kind of can realize to the self-adaptation real time correction of infrared image based on self-adaptation infrared image correcting algorithm thermal infrared imager.

The present invention solves the technical scheme that its technical matters adopts: this based on self-adaptation infrared image correcting algorithm thermal infrared imager, its self-adaptation infrared image correcting algorithm comprises following steps:

1) correction coefficient of one group of each unit is calculated with two-point potentionmetric;

2) in the makeover process of correction coefficient, first calculate the statistical property of the image after correcting, obtain one group of average statistical data of image, then utilize these group average statistical data and original image correction data to obtain the image error information of each pixel;

3) in coefficient correction, add correction to judge, image border or impulsive noise is distinguished with this, if image border, then retain former correction coefficient, if impulsive noise, then according to control information correction factor, carry out Nonuniformity Correction with revised coefficient, just can obtain high quality graphic all the time.

As preferably, described step 1) specific as follows: the black matrix inserting a homogeneous radiation in the optical path, then goes out gain factor G according to the RESPONSE CALCULATION of each array elements to blackbody radiation even under high temperature and low temperature _ijwith displacement factor O _ij, under supposing high temperature and low temperature, the response of all array elements is respectively V _hand V _l, the response of pixel (i, j) under high temperature and low temperature homogeneous radiation is respectively X _ij(φ _h) and X _ij(φ _l), line number and the columns of array elements are respectively M and N, then

$\left\{\begin{array}{c}{V}_H=G_{\mathrm{ij}}{X}_{\mathrm{ij}}\left({\mathrm{\φ}}_H\right)+O_{\mathrm{ij}}\\ V_L=G_{\mathrm{ij}}{X}_{\mathrm{ij}}\left({\mathrm{\φ}}_L\right)+O_{\mathrm{ij}}\end{array}\right.$

Wherein,

$V_L=\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{X}_{\mathrm{ij}}\left({\mathrm{\φ}}_L\right)}{M\×N}{V}_H=\frac{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{X}_{\mathrm{ij}}\left({\mathrm{\φ}}_H\right)}{M\×N}$

Thus, obtain:

$\left\{\begin{array}{c}{G}_{\mathrm{ij}}=\frac{V_H-V_L}{X_{\mathrm{ij}}\left({\mathrm{\φ}}_H\right)-X_{\mathrm{ij}}\left({\mathrm{\φ}}_L\right)}\\ O_{\mathrm{ij}}=\frac{V_H{X}_{\mathrm{ij}}\left({\mathrm{\φ}}_L\right)-V_L{X}_{\mathrm{ij}}\left({\mathrm{\φ}}_H\right)}{X_{\mathrm{ij}}\left({\mathrm{\φ}}_L\right)-X_{\mathrm{ij}}\left({\mathrm{\φ}}_H\right)}\end{array}\right.$

By the gain coefficient G of each array elements calculated _ijwith deviation ratio O _ijstored in Flash, in the initial procedure of Startup calibration, read the coefficient data in Flash, then multiply-add operation is carried out to it, just complete the heteropical real time correction of initial pictures.

As preferably, described step 2) specific as follows: the statistical property first calculating the image after correcting, obtain one group of average statistical data of image, then these group average statistical data and original image correction data is utilized to obtain the image error information of each pixel, when obtaining control information, scene information must be removed, by the average statistical data of the medium filtering process image of 3 × 3, the difference of the intermediate value tried to achieve and former correction pixels is image error, if for intermediate value, for the original pixel value after correction, e _jfor image error, then have

$e_j={\tilde{U}}_j^k-U_j^k$

The e tried to achieve _jbe exactly the control information of pixel j.

As preferably, described step 3) specific as follows: add in coefficient correction and revise judgement, image border or impulsive noise is distinguished with this, if image border, then retain former correction coefficient, if impulsive noise, then according to control information correction factor, 9 of medium filtering template values are arranged from small to large, namely then maximal value and minimum value is removed, that is:

${\left(K_{\mathrm{up}}\right)}_j=\sqrt{\frac{1}{7}\underset{m=2}{\overset{7}{\mathrm{\Σ}}}{({\tilde{U}}_J^K-U_m^k)}^2}$

The judgment value obtained, depends on comparing of judgment value and threshold value to the correction of coefficient, if judgment value is less than threshold value, then and correction factor, on the contrary then retain former coefficient, i.e. (K _up) _j< (K _s) _j, wherein, K _sit is the threshold value of coefficient correction.

Inventing useful effect is: the present invention is a kind of thermal infrared imager correcting the bearing calibration of infrared image heterogeneity real-time adaptive based on traditional two-point calibration, the method overcome the shortcoming repeating to calibrate of traditional two point correction, system is enable to follow the tracks of the change of input signal at work, moment keeps best calibration result, thus the self-adaptation real time correction realized infrared image, it judges by adding to revise in coefficient makeover process, improves the problem of the fuzzy gradually and ghost in conventional correction methods largely.

Embodiment

Embodiment:

During affecting infrared focal plane array, heteropical factor is a lot, first be the heterogeneity of detector pixel responsiveness or spectral responsivity, next is the coupling factor noise of signal read circuit self and sensing circuit and detector, and the heterogeneity etc. of dark current.Self-adaptation peg method is a kind of infrared image heterogeneity real-time correction method corrected based on traditional two-point calibration.The method overcome the shortcoming repeating to calibrate of traditional two point correction, enable system follow the tracks of the change of input signal at work, the moment keeps best calibration result, thus realizes the self-adaptation real time correction to infrared image.It judges by adding to revise in coefficient makeover process, improves the problem of the fuzzy gradually and ghost in conventional correction methods largely.

First, the correction coefficient of one group of each unit is calculated with traditional two-point potentionmetric.Two-point potentionmetric is the black matrix by inserting a homogeneous radiation in the optical path, then goes out gain factor G according to the RESPONSE CALCULATION of each array elements to blackbody radiation even under high temperature and low temperature _ijwith displacement factor O _ij, thus realize Nonuniformity Correction.Under supposing high temperature and low temperature, the response of all array elements is respectively V _hand V _l, the response of pixel (i, j) under high temperature and low temperature homogeneous radiation is respectively X _ij(φ _h) and X _ij(φ _l), line number and the columns of array elements are respectively M and N, then

$\left\{\begin{array}{c}{V}_H=G_{\mathrm{ij}}{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_H\right)+O_{\mathrm{ij}}\\ V_L=G_{\mathrm{ij}}{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_L\right)+O_{\mathrm{ij}}\end{array}\right.$

Wherein,

$V_L=\frac{\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_L\right)}{M\&times;N}{V}_H=\frac{\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_H\right)}{M\&times;N}$

Thus, obtain:

$\left\{\begin{array}{c}{G}_{\mathrm{ij}}=\frac{V_H-V_L}{X_{\mathrm{ij}}\left({\mathrm{\&phi;}}_H\right)-X_{\mathrm{ij}}\left({\mathrm{\&phi;}}_L\right)}\\ O_{\mathrm{ij}}=\frac{V_H{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_L\right)-V_L{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_H\right)}{X_{\mathrm{ij}}\left({\mathrm{\&phi;}}_L\right)-X_{\mathrm{ij}}\left({\mathrm{\&phi;}}_H\right)}\end{array}\right.$

By the gain coefficient G of each array elements calculated _ijwith deviation ratio O _ijstored in Flash, in the initial procedure of Startup calibration, read the coefficient data in Flash, then multiply-add operation is carried out to it, just complete the heteropical real time correction of initial pictures.

Due to the reason such as temperature variation and noise, the resonse characteristic of each unit of detector has DC shift in various degree.Utilize original correction coefficient cannot obtain high-quality infrared image, so must revise correction coefficient.Under guarantee real-time and constringent prerequisite, the periodicity correction coefficient according to detector each cell response characteristic carries out dynamic adaptive correction.In the makeover process of correction coefficient, first calculate the statistical property of the image after correcting, obtain one group of average statistical data of image, then utilize these group average statistical data and original image correction data to obtain the image error information of each pixel.When obtaining control information, must remove scene information, otherwise will comprise scene information in correction coefficient, by the average statistical data of the medium filtering process image of 3 × 3, the difference of the intermediate value of trying to achieve and former correction pixels is image error.We establish for intermediate value, for the original pixel value after correction, e _jfor image error, then have

$e_j={\tilde{U}}_j^k-U_j^k$

The e tried to achieve _jbe exactly the control information of pixel j.The correction effect of the control information of on average being tried to achieve by the timing statistics of one group of sequential picture to deviation ratio is better.

Optimal makeover process is that image error does not comprise any scene information, but this is inevitable in practice, and especially the marginal information of image often can be added in correction coefficient by force.When scene changes, before image just has, the edge shadow of image, is commonly called as " ghost ".In order to removal of images shade, we add correction and judge in coefficient correction, distinguish image border or impulsive noise with this.If image border, then retain former correction coefficient; If impulsive noise, then according to control information correction factor, 9 of medium filtering template values are arranged from small to large, namely $\{U_1^k,U_2^k,U_3^k,U_4^k,U_5^k,U_6^k,U_7^k,U_8^k,U_9^k\},U_{m-1}^k<U_k^m<U_{m+1}^k.$ Then maximal value and minimum value is removed, that is:

${\left(K_{\mathrm{up}}\right)}_j=\sqrt{\frac{1}{7}\underset{m=2}{\overset{7}{\mathrm{\&Sigma;}}}{({\tilde{U}}_J^K-U_m^k)}^2}$

The judgment value obtained, depends on comparing of judgment value and threshold value to the correction of coefficient.If judgment value is less than threshold value, then correction factor, on the contrary then retain former coefficient, i.e. (K _up) _j< (K _s) _j.Wherein, K _sit is the threshold value of coefficient correction.So simple deterministic process will make calibration result be improved significantly.By once calibrating, according to explorer response characteristic revision correction coefficient, and Nonuniformity Correction need only be carried out with revised coefficient, just can obtain high quality graphic all the time.

In addition to the implementation, the present invention can also have other embodiments.All employings are equal to the technical scheme of replacement or equivalent transformation formation, all drop on the protection domain of application claims.

Claims (4)

1. based on a self-adaptation infrared image correcting algorithm thermal infrared imager, it is characterized in that: its self-adaptation infrared image correcting algorithm comprises following steps:

1) correction coefficient of one group of each unit is calculated with two-point potentionmetric;

2) in the makeover process of correction coefficient, first calculate the statistical property of the image after correcting, obtain one group of average statistical data of image, then utilize these group average statistical data and original image correction data to obtain the image error information of each pixel;

3) in coefficient correction, add correction to judge, image border or impulsive noise is distinguished with this, if image border, then retain former correction coefficient, if impulsive noise, then according to control information correction factor, carry out Nonuniformity Correction with revised coefficient, just can obtain high quality graphic all the time.

2. according to claim 1 based on self-adaptation infrared image correcting algorithm thermal infrared imager, it is characterized in that: described step 1) specific as follows: the black matrix inserting a homogeneous radiation in the optical path, then goes out gain factor G according to the RESPONSE CALCULATION of each array elements to blackbody radiation even under high temperature and low temperature _ijwith displacement factor O _ij, under supposing high temperature and low temperature, the response of all array elements is respectively V _hand V _l, the response of pixel (i, j) under high temperature and low temperature homogeneous radiation is respectively X _ij(φ _h) and X _ij(φ _l), line number and the columns of array elements are respectively M and N, then

$\left\{\begin{array}{c}{V}_H=G_{\mathrm{ij}}{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_H\right)-O_{\mathrm{ij}}\\ V_L=G_{\mathrm{ij}}{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_L\right)+O_{\mathrm{ij}}\end{array}\right.$

Wherein,

$V_L=\frac{\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_L\right)}{M\&times;N}{V}_H=\frac{\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_H\right)}{M\&times;N}$

Thus, obtain:

$\left\{\begin{array}{c}{G}_{\mathrm{ij}}=\frac{V_H-H_L}{X_{\mathrm{ij}}\left({\mathrm{\&phi;}}_H\right)-X_{\mathrm{ij}}\left({\mathrm{\&phi;}}_L\right)}\\ O_{\mathrm{ij}}=\frac{V_H{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_L\right)-V_L{X}_{\mathrm{ij}}\left({\mathrm{\&phi;}}_H\right)}{X_{\mathrm{ij}}\left({\mathrm{\&phi;}}_L\right)-X_{\mathrm{ij}}\left({\mathrm{\&phi;}}_H\right)}\end{array}\right.$

By the gain coefficient G of each array elements calculated _ijwith deviation ratio O _ijstored in Flash, in the initial procedure of Startup calibration, read the coefficient data in Flash, then multiply-add operation is carried out to it, just complete the heteropical real time correction of initial pictures.

3. according to claim 1 based on self-adaptation infrared image correcting algorithm thermal infrared imager, it is characterized in that: described step 2) specific as follows: the statistical property first calculating the image after correcting, obtain one group of average statistical data of image, then these group average statistical data and original image correction data is utilized to obtain the image error information of each pixel, when obtaining control information, scene information must be removed, by the average statistical data of the medium filtering process image of 3 × 3, the difference of the intermediate value tried to achieve and former correction pixels is image error, if for intermediate value, for the original pixel value after correction, e _jfor image error, then have

$e_j={\tilde{U}}_j^k-U_j^k$

The e tried to achieve _jbe exactly the control information of pixel j.

4. according to claim 1 based on self-adaptation infrared image correcting algorithm thermal infrared imager, it is characterized in that: described step 3) specific as follows: add in coefficient correction and revise judgement, image border or impulsive noise is distinguished with this, if image border, then retain former correction coefficient, if impulsive noise, then according to control information correction factor, 9 of medium filtering template values are arranged from small to large, namely $\{U_1^k,U_2^k,U_3^k,U_4^k,U_5^k,U_6^k,U_7^k,U_8^k,U_9^k\},U_{m-1}^k<U_m^k<U_{m+1}^k,$ Then maximal value and minimum value is removed, that is:

${\left(K_{\mathrm{up}}\right)}_j=\sqrt{\frac{1}{7}\underset{m=2}{\overset{7}{\mathrm{\&Sigma;}}}{({\tilde{U}}_J^K-U_m^k)}^2}$

The judgment value obtained, depends on comparing of judgment value and threshold value to the correction of coefficient, if judgment value is less than threshold value, then and correction factor, on the contrary then retain former coefficient, i.e. (K _up) _j< (K _s) _j, wherein, K _sit is the threshold value of coefficient correction.