A TFT-LCD full dynamic threshold binary processing method based on pseudo frame difference

Abstract

In order to overcome the technical problem that the high efficiency and reliability cannot be considered simultaneously in the existing TFT-LCD panel ACF conductive particle detection method, the application provides a TFT-LCD full dynamic threshold binarization processing method based on pseudo frame difference; the method is inspired by the frame difference idea in motion multi-frame image processing, a new type of efficient pseudo frame difference processing method is established by establishing a plurality of plane image augmented matrices and using two plane image augmented matrices for frame difference in the pre-processing part of the ACF conductive particle identification in the TFT-LCD image, and the technical problem that the high efficiency and reliability cannot be considered simultaneously in the TFT-LCD panel ACF particle detection is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial production detection, in particular to a TFT-LCD full dynamic threshold binarization processing method based on pseudo frame difference, and belongs to the electronic information field. BACKGROUND

[0002] Thin film transistor liquid crystal display (TFT-LCD) is widely used in television, notebook computer, monitor, mobile phone and other products. TFT-LCD panel defects have always affected the display quality of TFT-LCD, and the inability to efficiently analyze the problem and feedback has seriously affected the production efficiency and delivery efficiency of the panel; as a crucial part of TFT-LCD production, the quality problem detection of conductive particles in anisotropic conductive film (ACF) has become a key detection technology for existing TFT-LCD production; a fast, efficient and accurate detection system will have great economic value and is of great significance to quality inspection and production in the assembly line.

[0003] At present, the image processing detection methods for this problem at home and abroad mainly fall into two directions, each with three types of processing schemes, which are: edge blur defect segmentation, difference method and filtering method based on image processing for TFT-LCD panel defect detection; and template matching method, gradient method and neural network method based on image recognition for ACF conductive particle detection.

[0004] The above various schemes perform very well in the general ACF conductive particle image processing which is not complex and can meet the requirements of industrial production process; however, due to the complexity of actual industrial production, low-precision and uneven brightness images often occur, and general simple calculation schemes are not practical under such conditions; other schemes propose many algorithms that focus on overcoming these difficulties by using filtering, image enhancement algorithms and full convolution neural networks, but due to the complexity of the algorithm, the timeliness requirement of industrial production cannot be met.

[0005] The problem of general simple processing algorithm is that it cannot cope with the complex and variable actual situation of industrial production, and has low reliability; although the complex processing algorithm has high reliability, most of the existing schemes are background processing schemes, which cannot meet the timeliness requirement of actual industrial production. It can be seen that the TFT-LCD panel ACF particle detection method still has the technical problem that high efficiency and reliability cannot be achieved. SUMMARY

[0006] In order to overcome the technical problem that the existing TFT-LCD panel ACF conductive particle detection method cannot simultaneously consider high efficiency and reliability, the application provides a TFT-LCD full dynamic threshold binarization processing method based on pseudo frame difference; the method is inspired by the frame difference idea in motion multi-frame image processing, and a new efficient pseudo frame difference processing method is established by establishing multiple plane image augmented matrices and using two plane image augmented matrices for frame difference in the processing part before recognizing the ACF conductive particles in the TFT-LCD image, so that the technical problem that the TFT-LCD panel ACF particle detection cannot simultaneously consider high efficiency and reliability is solved.

[0007] The technical solution adopted by the application to solve the technical problem is: a TFT-LCD full dynamic threshold binarization processing method based on pseudo frame difference, characterized by the following steps:

[0008] (1) the image collected by the acquisition system enters the image processing and recognition system;

[0009] (2) the original image collected is subjected to a to-be-processed image block area selection, a plurality of single TFT panels to be detected are screened out, and are stored as corresponding plane image matrices , the subscripts and are the row number and column number of the plane image matrix respectively, and the following description of the subscript meaning of the plane image matrix is the same;

[0010] (3) the following three plane image augmented matrices are established

[0011]

[0012] wherein is the plane image augmented matrix of the original plane image matrix, is the plane image augmented matrix established after the original plane image matrix is moved right columns in the horizontal direction, is the plane image augmented matrix established after the image is moved down rows in the vertical direction, , is an arbitrary non-zero positive integer selected according to the actual image; the translation before the pseudo frame difference processing of the image simulates the moving image targeted by the general frame difference method,

[0013] (4) the pseudo frame difference processing of the image, that is

[0014]

[0015] wherein , respectively represent the pseudo-frame difference planar image matrix in horizontal and vertical directions respectively;

[0016] (5) Selecting the pseudo-frame difference images in horizontal and vertical directions, the sub-matrix composed of the first row and the first column of the sub-matrix composed of the first row and the first column of

[0017] (6) Taking absolute value of all elements of and respectively in the following way, and defining:

[0018] , ,

[0019] ,

[0020] wherein, is the element in the i-th row and the j-th column; is the element in the i-th row and the j-th column;

[0021] (7) Further binarization fusion is performed in the following weighted sum way: defining:

[0022]

[0023] wherein ; ; , and all are positive numbers;

[0024] , ; is the outline point of the ACF particle after binarization fusion;

[0025] (8) The number and orientation information of the ACF conductive particles are extracted according to the results of (6) or (7).

[0026] ​​​​​​​​​​​​​​The beneficial results of the present application are: by pseudo frame difference operation on single image, a combined pseudo frame difference image after horizontal and vertical direction pseudo frame difference is obtained, and the new pseudo frame difference image is binarized based on the combined pseudo frame difference image. The uneven brightness problem in the image is efficiently solved, the binarization of the full dynamic threshold of any pixel point of the original image is realized, a new efficient pseudo frame difference processing method is established, and the technical problem that the high efficiency and reliability of TFT-LCD panel ACF particle detection cannot be achieved is solved.

[0027] The specific content aspects of the present application will be described in detail below in combination with specific embodiments. EMBODIMENT

[0028] (1) The image collected by the collection system enters the image processing and recognition system;

[0029] (2) The original image collected is subjected to a to-be-processed image block region selection, a plurality of single TFT panels to be detected are screened out, and are stored as corresponding plane image matrices , the subscripts and are the row number and column number of the plane image matrix respectively, and the following description of the subscript meaning of the plane image matrix is the same;

[0030] (3) The following three plane image augmented matrices are established

[0031]

[0032] , wherein is the plane image augmented matrix of the original plane image matrix, is the plane image augmented matrix established after the original plane image matrix is moved right by columns in the horizontal direction, is the plane image augmented matrix established after the image is moved down by rows in the vertical direction; the translation of the image before pseudo frame difference processing simulates the moving image targeted by the general frame difference method,

[0033] (4) The image is subjected to pseudo frame difference processing, that is

[0034]

[0035] , wherein , are respectively the pseudo frame difference plane image matrices of the original image in the horizontal and vertical directions;

[0036] (5) The pseudo frame difference images in the horizontal and vertical directions are selected, is the first row and the first column of to sub-matrix composed of the first for the first to sub-matrix composed of the first column;

[0037] (6) The thresholded binarization processing is performed on the and respectively in the following manner, and the definitions are:

[0038] ,

[0039] ,

[0040] wherein, is the element in the first the first column; is the element in the first the first column;

[0041] (7) Further binarization fusion is performed in the following manner of weighted sum: the definitions are:

[0042]

[0043] wherein ; ; , and are all positive numbers;

[0044] , ; is the outline point of the highlighted ACF particle after binarization fusion;

[0045] (8) The subsequent ACF conductive particle number and orientation information extraction is performed according to the results of (6) or (7).​​

Claims

1. A method for full dynamic threshold binarization processing of TFT-LCD based on pseudo-frame difference, characterized by the following steps: (1) Images acquired by the acquisition system are entered into the image processing and recognition system; (2) Select the image block region to be processed from the original image to filter out multiple individual TFT panels to be detected, and store them as corresponding planar image matrices. subscript and These represent the row and column numbers of the planar image matrix, respectively. The subscripts for the planar image matrix have the same meaning in the following descriptions. (3) Establish the following three planar image augmentation matrices in This is the planar image augmentation matrix of the original planar image matrix. Shift the original planar image matrix to the right horizontally The augmented matrix of the planar image established after column motion, Move the image vertically downwards The augmented matrix of the planar image established after the motion. , Let be an arbitrary non-zero positive integer selected based on the actual image; the translation of the image before pseudo-frame difference processing simulates the motion image targeted by the general frame difference method. (4) Perform pseudo-frame difference processing on the image, i.e. in , These are the pseudo-frame difference plane image matrices representing the original image in the horizontal and vertical directions, respectively; (5) Select pseudo frame difference images in the horizontal and vertical directions. for The former line, number to A submatrix composed of columns; for The to row, before A submatrix composed of columns; (6) To and Take the absolute value of all elements in the following ways, and define them as follows: ， , , in, for No. Line 1 Column elements; for No. Line 1 Column elements; (7) Further binarization fusion is performed using the following weighted sum method: Definition: in ; ; , And all are positive numbers; , ; The outline points that highlight the ACF particles are obtained after binarization and fusion. (8) Extract the number and orientation information of ACF conductive particles based on the results of (6) or (7).

Images