JPH0569375B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method of inspecting linear defects such as cracks and scratches existing on the surface of electronic parts or precision processed products by image processing.

[Conventional technology]

Conventionally, in order to inspect linear defects such as cracks and scratches that exist on the surface of electronic components or precision-processed products, there has been a method of placing the object to be inspected within the field of view of a microscope or magnifying glass and visually inspecting it. This was the most common method. However, in the above method, it is difficult to handle particularly small objects to be inspected, even with tweezers, etc., and in the case of mass production, workers are fatigued, making it difficult to continue working for long periods of time. Moreover, there are problems such as inefficiency. In order to solve the above problems, inspection methods using image processing means have been applied in recent years, and many of them are in practical use.

[Problem that the invention seeks to solve]

When the above-mentioned image processing means is applied to the above-mentioned linear defect inspection, there are the following problems. In other words, when inputting an image of the object to be inspected,
The problem is that unevenness in the background of the object to be inspected is also input as an image and becomes noise, making it impossible to distinguish it from linear defects. Furthermore, when the image signal obtained by capturing an image of the inspected object through an imaging means such as a TV camera is binarized using a certain threshold value, the originally continuous image becomes discontinuous. The problem is that it can sometimes happen. A method for displaying such discontinuous images in a continuous manner is disclosed in, for example, Japanese Patent Application Laid-Open No. 124873/1983.
However, with this method, there is a risk that not only the original linear defect but also the noise existing in the vicinity of the linear defect will be continuously displayed, and the non-defective part may be displayed as a defect. This poses a problem in that the reproducibility or reliability is low and it cannot be applied to actual work.

The present invention solves the problems existing in the prior art as described above, eliminates noise in input images, and provides highly reliable linear detection that completely inspects linear defects that appear discontinuously. The purpose is to provide a method for inspecting defects.

[Means for solving problems]

In order to solve the above problems, in the present invention, A. An imaging signal of the object to be inspected obtained through an imaging means is
It is stored in the image memory as binary pixel data in the XY axis directions.

B. A plurality of cells each consisting of a plurality of pixels are set in a matrix in the X and Y axis directions for the binarized pixel data through the central control unit.

C. Calculate the number of defective pixels in each cell.

D. Extract cells with a predetermined number or more of defective pixels as defective cells.

E Let mutually adjacent defective cells among these defective cells be connected components.

F Calculate the number of connected components in the entire image memory as connectivity.

G If the degree of connectivity exceeds a predetermined value, it is determined to be defective.

This technical method was adopted.

Next, the principle structure of the present invention will be explained with reference to FIGS. 1 to 4.

FIG. 1 is a diagram showing cells in the image memory, respectively in the X-axis (left and right) direction and the Y-axis (up and down).
Multiple lines are set in the direction. Each cell 101 shown as a square in the image memory 100 in FIG. 1 is composed of a plurality of pixels that have been converted into binarized pixel data via a central controller, and the cells with hunting indicate defects within the cell. Indicates that the cell is a defective cell whose number of pixels is greater than or equal to a predetermined number. Therefore, FIG. 1 shows that among the plurality of cells 101, A to D are defective cells.

In the present invention, defective cells A, B, and C adjacent to each other are defined as connected components, and the number of connected components is defined as the degree of connectivity. Therefore, in the case shown in FIG. 1, the degree of connectivity is calculated to be 3.

Next, a method for calculating connectivity will be explained.

First, the image memory 100 shown in FIG. 1 is scanned in the X-axis direction (from left to right), and the first defective cell is detected by scanning in which the X-axis direction is sequentially moved in the Y-axis direction (from top to bottom). do. That is, the defective cell A in the top row and leftmost column is detected.

After detecting the defective cell A, 3×3=9 cell masks are set with the defective cell A as the center, as shown in FIG. Then, examine these cell masks clockwise in the order of ~ to see if they are defective cells or not.
The first detected defective cell (defective cell B in FIG. 1) is extracted.

Next, as shown in FIG. 3, 3×3=9 cell masks are set with the defective cell B as the center, as in FIG. 2, and the presence or absence of defective cells is checked in the same manner as described above. However, in this case, since defective cell A has already been detected, the cells after defective cell A are examined clockwise in the order of ~, and the first detected defective cell (defective cell C in Figure 1) is checked. ).

FIG. 4 shows a mode in which the presence or absence of other defective cells is investigated centering on defective cell C, but in this case...
Since no defective cells are detected in any of the cells, the scanning is terminated.

The result of scanning as above is defective cell A.
-D are detected, and in the case shown in FIG. 1, the degree of connectivity of defective cells A to C, which are connected components, is calculated to be "3". Regarding the defective cell D in FIG. 1, even if the cell masks set in the vicinity of the defective cell D are examined in the order of ~, no other defective cells are detected, so the defective cell D is excluded as an upright cell.

FIG. 5 is a flowchart generalizing the above processing procedure. That is, by starting the process (201), scanning is continued to detect a defective cell in the top row and leftmost column (202), and after detecting the defective cell, a cell mask is set around this defective cell. The presence or absence of a defective cell is checked clockwise (203), a defective cell is extracted, and the defective cell is designated as C ₁ and the extracted defective cell is designated as C ₂ (204). The coordinate positions of C ₁ and C ₂ are stored (205).

Next, the above C ₁ and C ₂ are converted to Co _-1 and _Co , respectively (206), and the cell mask is checked clockwise from Co _-1 with _Co as the center (207), and the presence or absence of a defective cell is detected first. Extract the defective cell C _o-1 , and calculate C _o-1 , C _o , Cn
The +1 coordinate position is stored (208).

Convert Co _and Co _{+1 to} Co _and Co respectively (209), and repeat the process from (207) onward.

As a result of examining neighboring cells centered on _Co , if no defective cell is detected, n is calculated as the degree of connectivity (210), and the above processing is terminated. (211).

[Effect]

With the above configuration, most of the noise in the input image of the object to be inspected is removed at the stage of calculating the number of defective pixels in the cell because the number of pixels in the cell is usually small. It has the effect of eliminating the mistake of recognizing something as a defect.
After eliminating noise as described above, the degree of connectivity between defective cells corresponding only to the linear defect is calculated to inspect the linear defect.

〔Example〕

FIG. 6 is a block diagram showing an example of an apparatus according to an embodiment of the present invention. In FIG. 6, an imaging device 3 such as a TV camera is provided opposite the object 1 to be inspected via a lens 2. Reference numeral 4 denotes an AD conversion device, which inputs the imaging signal from the imaging device 3, converts it into binary pixel data, and connects it to the image memory 5 so that it can be input. 6 is a processing circuit connected to process the contents of the memory 5, and 7 is an image memory connected to store the processing results of the processing circuit 6.
The AD conversion device 4, image memories 5 and 7, and processing circuit 6 are electrically and controllably connected to a central control device 9 via a bus 8, respectively.

Next, FIGS. 7a to 7e are schematic diagrams showing main parts of the processing procedure of the present invention, respectively. First, FIG. 7a shows the binarized pixel data stored in the image memory 5 in FIG. 1. In the figure, numeral 10 indicates a crack, which is a type of linear defect existing on the surface of the object 1 to be inspected. Although the cracks 10 are originally continuous, a portion of the cracks 10 is discontinuous as shown by 10a as binarized pixel data. Reference numeral 11 denotes a nozzle, which is stored as binary pixel data due to irregularities in the surface of the object 1 to be inspected.

Next, FIG. 7b shows the state in which the cells are set. That is, the cell 12 is composed of, for example, 8×8=64 pixels, and a plurality of cells are set in rows and columns in the XY axis directions via the central controller 9. FIG. 7b shows an example in which 14 pieces are set on each of the X-axis and the Y-axis, for a total of 196 pieces.

FIG. 7c shows a situation in which the number of defective pixels within the cell 12 is calculated. That is, within each cell 12, the number m of defective pixels occupied by cracks 10 and nozzles 11 is calculated. Note that if the background density is low, the defective pixel
and the nozzle 11, and if the background has a high density, it will be shown with a low density.

FIG. 7d shows a state in which the defective cell 12a is extracted. In other words, cells in which the number m of defective pixels calculated in FIG.
2a, and cells 12 less than the above number are removed. Noise 11 that appears as an input image in FIGS. 7a-c is usually cell 12.
The number of pixels occupied within the area is extremely small. Therefore, if the number m of defective pixels is set in relation to the size of the crack 10 to be detected, most of the noise can be removed at the stage d in FIG. Therefore, only the defective cells 12a that are present can be extracted.
Therefore, even if the defect is in a discontinuous state as shown by 10a in FIG. 7a, it can be extracted as a defective cell 12a as shown in FIG. 7d.

FIG. 7e shows that the degree of connectivity of the defective cell 12a is calculated by the processing procedure described above, and the cracks 10,
10a is shown. As a result, anything above a predetermined value is determined to be defective, and anything below the above value is accepted and determined to be good.

In this embodiment, an example was described in which the linear defect of the object to be inspected is a crack, but even if the linear defect is a scratch, a dent, etc., it is possible to display shading on the pixels in the cell. detectable as long as possible. Also, the cell size is not limited to 8 x 8 = 64,
It should be selected appropriately depending on the type of linear defect to be detected and the noise to be eliminated.

〔Effect of the invention〕

Since the present invention has the structure and operation as described above, the following effects can be expected.

(1) Since the conventional manual visual inspection can be eliminated, not only can the working environment be dramatically improved, but also productivity can be greatly improved.

(2) Even if the input image signal contains noise other than linear defects, it is possible to calculate the number of defective pixels in each cell and extract only cells with a predetermined number or more. The error is extremely small and the reliability is extremely high.

(3) In the input image signal, even if the defect is discontinuous, linear defects that are originally in a continuous state are recognized and processed as a continuous state by calculating the degree of connectivity, so reliability and reproducibility are improved. is extremely high.

(4) Not only can the inspection work be automated through the central control device, but also the processing results can be immediately fed back to the production process, which can improve production efficiency and contribute to quality improvement.

[Brief explanation of the drawing]

1 to 4 are diagrams showing cells in an image memory in an embodiment of the present invention, FIG. 5 is a flow chart showing a generalized processing procedure in an embodiment of the present invention, and FIG. 6 is a diagram showing a generalized processing procedure in an embodiment of the present invention. A block diagram showing an example of the apparatus in the embodiment of FIG. 7a-e.
2A and 2B are schematic diagrams showing main parts of the processing procedure in the embodiments of the present invention, respectively. 1: inspection object, 5, 7: image memory, 6: processing circuit, 10, 10a: crack, 11: noise,
12: Cell.

Claims (1)

[Claims] 1 The imaging signal of the object to be inspected obtained through the imaging means is
The binarized pixel data is stored in an image memory as binary pixel data in the XY axis directions, and cells consisting of a plurality of pixels are
A plurality of defective pixels are set in a matrix in the axial direction, the number of defective pixels in each cell is calculated, cells with a predetermined number or more of defective pixels are extracted as defective cells, and among these defective cells, mutually Inspection of linear defects characterized by using adjacent defective cells as connected components, calculating the number of connected components in the entire image memory as a degree of connectivity, and determining as a defect if the degree of connectivity exceeds a predetermined value. Method.