CN108491852B

CN108491852B - Blind pixel detection method for area array infrared focal plane

Info

Publication number: CN108491852B
Application number: CN201810094043.9A
Authority: CN
Inventors: 陈忻; 饶鹏; 夏晖; 赵云峰
Original assignee: Shanghai Institute of Technical Physics of CAS
Current assignee: Shanghai Institute of Technical Physics of CAS
Priority date: 2018-01-31
Filing date: 2018-01-31
Publication date: 2021-08-06
Anticipated expiration: 2038-01-31
Also published as: CN108491852A

Abstract

The invention discloses a blind pixel detection method of an area array infrared focal plane. The method firstly carries out scanning imaging on an area array infrared focal plane and acquires image data. Reading data in a scanning direction in a line unit, constructing a data vector set in a pixel unit, carrying out self-organization clustering on the data vector set, voting on a clustering result, and determining a blind pixel position according to the voting result. And sequentially processing all the rows to complete the blind pixel detection of the focal plane and construct a blind pixel table. The invention can accurately and rapidly complete the detection and positioning of the blind pixels without a reference source.

Description

Blind pixel detection method for area array infrared focal plane

The technical field is as follows:

the invention relates to an infrared detector signal processing technology, in particular to a blind pixel detection method applied to an area array infrared focal plane.

Background

In recent years, the infrared focal plane array (IRFPA) technology has been rapidly developed and is widely used. Because of the influence of manufacturing materials, processes and other factors (such as material nonuniformity, mask errors, defects and the like), the infrared focal plane array device has inevitable nonuniformity, and in extreme cases, part of pixels lose the detection capability and become blind pixels (including dead pixels and overheated pixels). The number and distribution of the blind pixels have certain influence on the performance of the IRFPA, so that bright spots or dark spots exist in the infrared image, and the quality of the infrared image is influenced. The infrared image is processed through blind pixel compensation, the influence of the blind pixels is eliminated, and the imaging quality of the infrared focal plane can be improved. The premise and basis of blind pixel compensation are that the blind pixels can be accurately positioned, namely blind pixel detection. Therefore, the appropriate blind pixel detection method has important application value and significance for improving the performance of the infrared focal plane [1 ].

At present, a large number of blind pixel detection methods have been proposed at home and abroad, and can be generally divided into two categories, namely a detection method based on blackbody calibration and a detection method based on scenes. The black body calibration method obtains a uniform radiation image through black body imaging, determines blind pixels according to the response rate of pixels, noise and the like on the basis, is widely applied to various infrared imaging systems at present, and is simple in principle, but needs black body matching, is complex in steps, needs to interrupt the normal working flow of the system, and cannot process new blind pixels which randomly appear [3] [4 ]. The scene-based detection rule does not depend on a blackbody reference source, but has the defects of poor precision, high possibility of being influenced by image nonuniformity and large computation amount.

[1] GB/T17444-2013 infrared focal plane array parameter testing method [ S ],2013

[2] Yellow, Zhangjianqi, Liude Lianglian infrared image blind pixel self-adaptive detection and compensation algorithm [ J ] infrared and laser engineering, 2011,40(2): 370-.

[3] Zhang Qiao boat, consider the country and the money, etc., based on two-point parameters and the real-time blind pixel detection and compensation technology of the adaptive window [ J ] infrared technology, 2013(3) 139-.

[4] Study on blind pixel detection algorithm of infrared focal plane array detector [ J ] infrared technology, 2011,33(4): 233-.

The invention content is as follows:

in order to overcome the defects of the prior art, the invention provides a blind pixel detection method of an area array infrared focal plane, which can accurately and quickly complete the blind pixel detection and positioning of the area array infrared focal plane without a reference source.

The above purpose of the invention is realized by the following technical scheme:

(1) the area array infrared focal plane performs scanning imaging at a certain speed and acquires image data. The focal plane size is M.N, wherein M is the number of rows, N is the number of columns, the scanning direction is vertical to the column direction, P frame data are collected together, and the total data size is M.N.P;

(2) selecting a row of data from M rows to read out, and reading out N.P data in total;

(3) and constructing data vectors by taking picture elements as units, wherein N.P data are formed into N P-dimensional vectors. Preprocessing a data vector, wherein the data vector obeys N (0,1) distribution;

(4) self-organizing clustering is carried out on the data vectors, whether different data vectors of the data belong to one class or not is evaluated through distance measure calculation, and a threshold value for generating a new class is set as T;

(5) voting is carried out on the clustering result, the most voted class is a normal pixel, and the rest voted classes are blind pixels;

(6) after all rows are processed, the distribution condition of blind pixels on the focal plane can be obtained, and accordingly a blind pixel table is constructed, wherein the blind pixel table is a two-dimensional array which corresponds to focal plane pixels one by one, 1 is filled in the blind pixels, and 0 is filled in the normal pixels.

The speed of the scanning imaging in the step (1) is an arbitrary value.

The distance measure in the self-organizing clusters in the step (4) is obtained by the weighted combination calculation of a correlation function, Euclidean distance, cross entropy or several calculation methods.

Compared with the prior art, the invention has the beneficial effects that

1. And a reference source is not needed, and the complexity and the detection cost of blind pixel detection are reduced.

2. And data clustering is performed in a high-dimensional space through the data vectors, so that the blind pixel detection accuracy is improved.

3. The area array infrared focal plane working in the scanning imaging mode can be subjected to blind pixel detection under the condition of not interrupting the working state.

Drawings

FIG. 1 is a process flow diagram of the present invention;

FIG. 2 is a diagram of data of a column of a focal plane acquired by the present invention;

FIG. 3 shows the result of the present invention after preprocessing the data;

FIG. 4 is a table of blind pixel locations obtained by the present invention;

FIG. 5 is the imaging results before compensation;

fig. 6 shows the result of blind pixel supplement using the blind pixel table found by the present invention.

Detailed Description

The invention is described in further detail below with reference to the figures and specific embodiments. Several parameters are involved that need to be adjusted for a particular processing environment to achieve good performance.

And (3) verifying by using 256 x 256 infrared focal planes, wherein the verification system is placed on a one-dimensional turntable for scanning imaging, and the rotating speed of the turntable is 20 degrees/S.

500 frames of data are collected in total, the total data volume is 256 × 500, the data are sequentially read out in 256 rows in a row unit, and 256 × 500 data are read out at a time, so that 256 500-dimensional data vector sets are formed. The preprocessed 256 500-dimensional data vector sets are self-organized and clustered, and the preprocessing results are shown in fig. 2 and fig. 3. The threshold for generating a new class in the cluster is set to 0.8. Voting is carried out according to the clustering result to obtain normal pixels, and the rest are blind pixels. Fig. 4 is a table of blind pixels produced after all rows are processed, wherein the bright points represent the blind pixels, and the dark points represent normal pixels. Fig. 5 is an image before blind pixel compensation. Fig. 6 shows the result of blind pixel supplement by using the blind pixel table found by the present invention, and the adopted blind pixel compensation method is the neighborhood non-blind pixel mean value, so that the blind pixel in the image can be found out well.

Claims

1. A blind pixel detection method of an area array infrared focal plane is characterized by comprising the following steps:

(1) the area array infrared focal plane performs scanning imaging at a certain speed and acquires image data, wherein the scanning imaging of the area array infrared focal plane is realized by rotating a turntable or swinging a scanning mirror;

(2) reading data in a scanning direction in a row unit, and reading the acquired data in the row unit; the number of focal plane pixels is M.N, wherein M is the number of rows, N is the number of columns, the scanning direction is vertical to the column direction, and P frame data are collected together; therefore, the total data size is M.N.P, and one row of data is selected from M rows by reading the data in row units, wherein the number of the data is N.P;

(3) constructing a data vector set by taking the pixel as a unit, and preprocessing the data vector set; the step of constructing the data vector by taking the pixel as a unit refers to the step of organizing N.P data into N P-dimensional vectors according to the sequence of the pixel; the data vector is preprocessed, namely the data vector is subjected to N (0,1) distribution by adjusting the mean value and the standard deviation of the data vector;

(5) voting is carried out on the clustering result, and the blind pixel position in the pixel of the row is determined; the blind pixel method is that the clustering result is voted, the most voted one is a normal pixel, and the rest is blind pixels;

(6) repeating the step (2), and after all rows are processed, obtaining the distribution situation of blind pixels on the focal plane, and accordingly constructing a blind pixel table; the blind pixel table is a two-dimensional array which is in one-to-one correspondence with focal plane pixels, wherein the blind pixels are filled with 1, and the normal pixels are filled with 0.

2. The method for detecting blind pixels of the area array infrared focal plane according to claim 1, characterized in that: the speed in the scanning imaging in the step (1) is an arbitrary value.

3. The method for detecting blind pixels of the area array infrared focal plane according to claim 1, characterized in that: the distance measure in the self-organizing clusters in the step (4) is obtained by the weighted combination calculation of a correlation function, Euclidean distance, cross entropy or several calculation methods.