CN116577349A

CN116577349A - System and method for detecting defects on smooth surface of semiconductor

Info

Publication number: CN116577349A
Application number: CN202310855811.9A
Authority: CN
Inventors: 郑飞; 洪敬柱; 吴绪波; 何海双; 陈志强
Original assignee: Hefei Tuxun Electronic Technology Co ltd
Current assignee: Hefei Tuxun Electronic Technology Co ltd
Priority date: 2023-07-13
Filing date: 2023-07-13
Publication date: 2023-08-11
Anticipated expiration: 2043-07-13
Also published as: CN116577349B

Abstract

The invention discloses a system and a method for detecting defects on a smooth surface of a semiconductor, and belongs to the technical field of semiconductor product detection. The invention comprises a camera component, a light source component and analysis equipment for mutually transmitting information with the camera component, wherein the light source component comprises a plurality of lamp beads which are in a circular ring shape around the camera component as a central circumferential array; the lamp beads are divided into a plurality of groups; each group of lamp beads in the light source assembly is lightened to irradiate the sample to be detected; controlling each group of lamp beads to light and irradiate the sample to be detected, and when each group of lamp beads light, the camera component shoots an image of the sample to be detected and sends the image to the industrial personal computer; and synthesizing an image for detecting the defects of the sample by a multi-image synthesis method according to a plurality of images acquired from the camera assembly and according to the azimuth of each group of lamp beads, and carrying out defect detection judgment according to the gray scale of the defects through the image. The effect of defects on the image is enhanced by synthesizing one image, so that the defects on the smooth surface of the semiconductor are detected, and the omission ratio and the false detection ratio are reduced.

Description

System and method for detecting defects on smooth surface of semiconductor

Technical Field

The invention relates to the technical field of semiconductor product detection, in particular to a system and a method for detecting defects on a smooth surface of a semiconductor.

Background

The semiconductor field often has many smooth surface products, such as CMOS photo-sensitive chips, PCB circuit boards and wafer products, which generally have high requirements for surface flatness.

The CMOS photosensitive chip is required to be used as a camera photosensitive chip, so that the surface appearance requirement on one end of the sample photosensitive element is extremely high, and defects such as dirt, cracks, scratches and the like cannot occur. However, since one end of the photosensitive element is usually made of a relatively brittle glass material, scratches are occasionally generated under the influence of external force factors, and the scratches greatly affect the imaging effect of the photosensitive chip. In order to ensure the yield of the shipment of the CMOS photosensitive chip, defective products with scratches must be detected, and the imaging effect is often not obvious enough when the scratches are shallower, and the existing detection scheme also often cannot well meet the detection requirements of CMOS manufacturers.

Surface nodulation and scratches of the PCB may be a major cause of rejection thereof. Most of the printed circuit boards rejected due to scratches are scratched after the final surface coating is completed, and when scratches occur on the final surface coating, scratches of different depths produce different results. Deep scratches may directly block the internal leads, causing circuit blockage, or may pass through the final surface plating, causing oxidation of copper exposed to air, which in turn causes circuit blockage. Only the slight scratches that wear without completely wiping off the gold plating will not affect the subsequent use. Therefore, detecting and judging the specific depth of the scratch is very important to the yield of the PCB manufacturing company.

The wafer is one of the stages of chip manufacture, and the stage needs to carry out the power on test by pressing a plurality of pins to the fixed position of the wafer, and because of errors of a mechanical system, the pressing position of the pins occasionally shifts, if the pressing area shifts too much, the power on test result of the sample may be inaccurate, even the sample is destroyed or scrapped, if the pressing force of the pins is too large, the sample may be damaged due to too deep dent, and if the pressing force is too small, the power on test result may be inaccurate, and the re-detection is needed. In order to know the real condition of the sample after the power-on test and improve the yield of shipment, the needle mark position and the depth of the needle mark pit after the thimble is pressed need to be detected, and the next procedure of the sample is determined by combining the power-on test result.

The prior art for defect detection of such semiconductor samples (hereinafter referred to as "sample") has the following schemes:

1. and 2D visual detection or morphological detection, judging whether defects exist according to the form of the sample on the image, and judging the types of the defects.

2. AI visual detection or deep learning detection, comparing with the detected object image according to the deep learning model, further judging whether defects exist or not, and classifying the defects.

3. The method can obtain the 3-dimensional depth of the surface of the sample by laser or structured light 3-dimensional imaging, so that the sample with the defects of scratch and dent can be conveniently detected.

The prior art has the following technical defects:

1. the existing 2D detection has higher requirements on the light source effect and the image effect, and the conventional 2D detection only has 1 to 2 light sources, or is a plurality of layers of annular light sources with different colors, and cannot reflect the forms of defects under different angles of illumination, so that the shot defects are often not obvious enough. When the defects on the shot image are not obvious, missed detection is easy to occur; when the image effect is poor, false detection is easy to occur, and the defect types are difficult to classify conveniently by the method.

2. The AI detection can make up the defects of partial 2D detection, such as higher robustness, and has no too high requirement on image effect, and can classify the defect types as long as the training samples are enough, but if a good detection effect is to be achieved, the required sample size is larger, more samples are difficult to obtain, and the light source effect still has a certain requirement.

3. Because the surface reflection effect of the smooth product is close to specular reflection, laser or structured light can hardly project effective patterns on the surface of the smooth product, and a sample model cannot be correctly reconstructed, so that defect detection cannot be performed on the product. Moreover, the accuracy of the depth reconstruction in the mode is difficult to achieve, and the calculation process is complex.

Disclosure of Invention

1. Technical problem to be solved by the invention

Aiming at the defect detection defect situation of the semiconductor surface in the prior art, the invention provides a system and a method for detecting the defect of the smooth surface of the semiconductor, which are based on the multi-angle light source multi-image synthesis technology, utilize the characteristic that when the defect of the protrusion or the depression exists on the smooth surface of the semiconductor, the defect part can reflect illumination with different angles, and synthesize one image through a plurality of images generated by illumination with different angles, thereby enhancing the effect of the defect on the image, further detecting the defect of the smooth surface of the semiconductor, and reducing the omission ratio and the false detection ratio.

2. Technical proposal

In order to achieve the above purpose, the technical scheme provided by the invention is as follows:

the system comprises a camera component, at least one light source component and analysis equipment for mutually transmitting information with the camera component, wherein the light source component comprises a plurality of lamp beads which are in a circular ring shape around the camera component in a central circumferential array, and the camera component is positioned at one side far away from the irradiation of the light source component; the lamp beads are divided into A groups, and in each group, a plurality of lamp beads form a strip-shaped light source; and each group of lamp beads in the light source assembly is lightened to irradiate the sample to be detected.

Further, when the number of the light source assemblies is more than two, the radii of the circular rings formed by the lamp beads in each light source assembly around the camera assembly as a central circumferential array are different; the plurality of light source components are concentrically arranged, and the lamp beads on the different light source components irradiate the sample to be detected without shielding each other.

Further, the lamp beads are arranged on the lamp holder, and the lamp holder is annular.

Further, the camera module comprises a stand column, wherein a chute is formed in the stand column, and the camera module and the light source module are installed on the chute through a bracket.

Further, the sliding groove extends along the irradiation direction of the light source component, and the camera component and the light source component can adjust the relative position between the camera component and the light source component along the sliding groove.

The invention also provides a method for detecting the defect of the smooth surface of the semiconductor, which adopts the system for detecting the defect of the smooth surface of the semiconductor, and comprises the following specific modes:

placing a sample to be detected at a designated position under the irradiation of a light source assembly, and calibrating the azimuth of each group of lamp beads;

controlling each group of lamp beads in the light source assembly to be sequentially lightened to irradiate a sample to be detected, and when each group of lamp beads are lightened, shooting an image of the sample to be detected by the camera assembly and sending the image to analysis equipment;

the analysis device synthesizes an image for detecting the defects of the sample by a multi-image synthesis method according to a plurality of images acquired from the camera component and according to the azimuth of each group of lamp beads, so that the display of the defects on the image is enhanced, and defect detection judgment is carried out according to the gray scale of the defects through the image.

Further, synthesizing an image for detecting a defect of a sample by a multi-image synthesizing method includes the steps of:

processing the images to obtain gray values of corresponding pixel points in the images, wherein the pixel points at the same position in the images correspond to the same point in the sample to be detected;

calculating the normal vector of each pixel point on the picture through the light source direction matrix and the gray value of each pixel point;

obtaining gray mapping of each pixel point according to the normal vector of each pixel point, and synthesizing an image;

further, the gray mapping formula of the composite image is:

in the middle ofFor Fourier transform, ++>Is an inverse Fourier transform>Is imaginary unit, ++>Is gray scale mapping; the number of pixels per gray picture is +.>；/>；/>，/>，/>All 1 row->Normal vector of column pixel points.

Further, the number of pixels and the arrangement mode of the pixels of the plurality of images are the same.

Further, the method also comprises the following steps: the method for acquiring the azimuth data of each group of lamp beads comprises the following steps: and carrying out normalization processing on the position coordinates of each group of lamp beads, and obtaining a light source direction matrix according to the normalized position coordinates of each group of lamp beads.

3. Advantageous effects

Compared with the prior art, the technical scheme provided by the invention has the following remarkable effects:

(1) The invention discloses a system and a method for detecting defects on a smooth surface of a semiconductor, which are used for solving the problem of detecting defects on the surface of a smooth sample in the prior art, and are based on a multi-angle light source multi-image synthesis technology.

(2) According to the system and the method for detecting the defects on the smooth surface of the semiconductor, disclosed by the invention, the defects in different directions can be detected by irradiating the same sample by using light sources with different angles. The synthesized image can enhance the defect effect and can more conveniently detect the defect. There is no need to collect a large number of sample pictures in advance for training as in AI detection, nor is it disturbed by the effect of smooth surface reflection. The depth of the defect does not need to be calculated as 3D modeling, the calculation process is complex, the occupied memory of the computer is large, and the speed is low.

(3) The invention relates to a system and a method for detecting defects on a smooth surface of a semiconductor, wherein a light source assembly comprises a lamp holder and a plurality of lamp beads arranged on the lamp holder, all the lamp beads of each light source assembly are arranged on one lamp holder, and only the light source assembly is adjusted to instantly adjust the positions of all the lamp beads, so that the relative positions of a light source, a camera assembly and a sample are convenient to adjust.

(4) The system and the method for detecting the defects on the smooth surface of the semiconductor are flexible in combination and adjustment of the light source components, more than two light source components can be arranged, and the detection system can be adapted to detecting a plurality of products, such as non-planar products, ball products and the like. The lamp bead arrays on each light source component are in a ring shape, so that the angles of each angle towards the sample are the same, and the illumination of each group of lamp beads is uniform.

Drawings

FIG. 1 is a schematic diagram of a three-dimensional structure of a detection system;

FIG. 2 is a front view of the detection system of FIG. 1;

FIG. 3 is a front view of the detection system with two light source modules;

FIG. 4 is a schematic perspective view of a light source assembly;

FIG. 5 is a flow chart of a method for detecting defects on a smooth surface of a semiconductor;

FIG. 6 is a schematic diagram of grouping light source module beads according to an embodiment;

FIG. 7 is a schematic diagram of the positional relationship among the camera assembly, the light source assembly and the sample under inspection;

FIG. 8a is a schematic diagram showing a sample to be inspected being a CMOS photosensitive chip, and an image of the CMOS photosensitive chip being captured under illumination of a lamp bead set;

FIG. 8b is a schematic illustration showing a sample to be inspected being a CMOS light sensitive chip, and a second light bead set being irradiated to capture an image of the CMOS light sensitive chip;

FIG. 8c is a photograph of a sample to be inspected being a CMOS light sensitive chip, the light bead set being illuminated by three illumination lights;

FIG. 8d is a photograph of a sample to be inspected being a CMOS photosensitive chip, the light bead set being illuminated by four light beads;

FIG. 8e is a photograph of a sample to be inspected being a CMOS light sensitive chip, the light bead set being illuminated by five illumination lights;

FIG. 8f is a photograph of a sample to be inspected being a CMOS light sensitive chip, under six illumination of a lamp bead set;

FIG. 8g is a photograph of a sample to be inspected being a CMOS light sensitive chip, under illumination of a set of beads;

FIG. 8h is a photograph of a sample to be inspected being a CMOS light sensitive chip, with eight illumination of the light bead set;

FIG. 9 is an original view of a sample of a CMOS photo-sensitive chip;

FIG. 10 is a sample image of a synthesized CMOS photo-sensitive chip for defect detection;

FIG. 11 is a schematic diagram of marking defects on an artwork of a CMOS photo-sensitive chip sample;

FIG. 12a is a view of a sample to be inspected being a PCB, and an image of the PCB being captured under illumination by a set of beads;

fig. 12b is a diagram showing a sample to be inspected being a PCB, and an image of the PCB being captured under illumination by a second lamp bead set;

fig. 12c is a view of a sample to be inspected being a PCB, and a shot image of the PCB being illuminated by a third set of beads;

fig. 12d is a photograph of a sample to be inspected being a PCB, and a lamp bead group being irradiated with four lights;

fig. 12e is a diagram of a sample to be inspected being a PCB, and a fifth lamp bead set shooting an image of the PCB under irradiation;

fig. 12f is a view of a sample to be inspected being a PCB, and a shot image of the PCB being illuminated by a bead set six;

fig. 12g is a diagram of a sample to be tested being a PCB, and a seventh lamp bead set shooting an image of the PCB under irradiation;

fig. 12h is a photograph of a sample to be inspected being a PCB, and a lamp bead set being illuminated by eight lights;

fig. 13 is an original view of a PCB circuit board sample;

FIG. 14 is a sample image of a PCB circuit board after synthesis for detecting defects;

FIG. 15a is a view of a wafer as a sample to be inspected, taken under illumination from a set of beads;

FIG. 15b is a view of a wafer as the sample to be inspected, and a second shot of the lamp bead set;

FIG. 15c is a view of a wafer as the sample to be inspected, with a third illumination of the set of beads;

FIG. 15d is a view of a wafer as the sample to be inspected, and a wafer image captured under illumination of the lamp bead set four;

FIG. 15e is a view of a wafer as the sample to be inspected, and a wafer image captured under illumination of a fifth lamp bead set;

FIG. 15f is a view of a wafer as a sample to be inspected, taken under six illumination from a set of beads;

FIG. 15g is a view of a wafer as a sample to be inspected, taken under illumination by a seventh set of beads;

FIG. 15h is a photograph of a wafer with eight illumination of the lamp bead set, with the sample to be inspected being the wafer;

FIG. 16 is an original view of a wafer sample;

fig. 17 is an image of a wafer sample after synthesis for defect detection.

Reference numerals in the schematic drawings illustrate: 10. a column; 101. a chute; 20. a camera assembly; 30. a light source assembly; 301. a lamp holder; 302. a lamp bead; 40. and (5) a sample to be detected.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

It should be noted that, in the embodiments of the present invention, terms such as "up, down, left, and right" are generally used with respect to the directions shown in the drawings; also, for ease of understanding and description, "inner and outer" refers to inner and outer relative to the profile of each component itself. The terms "first," "second," and "third," etc. are used for descriptive purposes only.

Aiming at the defect detection defect situation of the semiconductor surface in the prior art, the embodiment of the invention is based on the multi-angle light source multi-image synthesis technology, utilizes the characteristic that when the defect exists on the smooth surface of the semiconductor, the defect can reflect illumination with different angles, synthesizes one image by a plurality of images generated by illumination with different angles, and enhances the effect of the defect on the image, thereby detecting the defect of the smooth surface of the semiconductor and reducing the omission factor and the false detection rate.

The multi-image synthesizing technology uses light sources in multiple directions to irradiate the convex or concave defects on the surface of the sample, the method can reflect the brightness change of the defects on multiple images according to the reflectivity of the defects under the light sources in different directions, and the multiple images are synthesized into one image by using the multi-image synthesizing method, so that the effect of the defects on the images is enhanced. The synthesized image defects are enhanced to detect the sample.

To achieve the object of defect detection of a sample, as shown in fig. 1, 2 and 4, an embodiment of the present invention firstly provides a semiconductor smooth surface defect detection system, which comprises a camera assembly 20, a light source assembly 30 and an analysis device for transmitting information with the camera assembly 20. In this embodiment, the analysis device is integrated in the industrial personal computer, and meanwhile, the industrial personal computer is also integrated with a control device, and the control device can control the light source assembly 30. For ease of analysis and control, the industrial personal computer is connected to the camera assembly 20 and the light source assembly 30.

The light source assembly 30 comprises a lamp holder 301 and a plurality of lamp beads 302 arranged on the lamp holder 301, the lamp beads 302 are circularly arrayed around the camera assembly 20 in a central circumferential direction, and the camera assembly 20 is positioned on one side far away from the light source assembly 30. And the beads 302 are divided into a groups, and in each group, a plurality of beads form an arc-shaped band-shaped light source. The number of beads 302 in each set may or may not be the same. Each group of the lamp beads 302 in the light source assembly 30 can be sequentially lighted to irradiate the sample 40 to be inspected.

In this embodiment, a light source assembly 30 is provided, the lamp beads 302 provided in the lamp holder 301 are LED lamps, 32 lamp beads are provided in total, and the 32 lamp beads are divided into 8 groups of 4 lamp beads each. Other combinations may be used in some cases, for example, in 32 groups of only 1 bead, or in 4 groups of 8 beads. When the quantity of the lamp beads in each group is inconsistent, one group can be three lamp beads, the other group is five lamp beads, and the two groups are mutually staggered. Since the light beads 302 are circularly arrayed around the camera assembly 20 in a central circumferential direction, the lamp holders 301 are also circularly disposed for convenience of disposing the light beads.

In other embodiments, the beads 302 may be provided in other numbers, such as 24. Further, a plurality of light source modules 30 may be provided. For example, two light source modules having 32 beads, or one light source module 30 having 32 beads and one light source module having 24 beads, or three or more light source modules may be provided. When the light source components are arranged in a plurality, the adjacent lamp beads on the light source components can be divided into a group of lamp bead groups, and two lamp beads on one light source component and two lamp beads on the other light source component which are adjacent to the two lamp beads form a group of lamp bead groups.

As shown in fig. 3, when the light source assemblies 30 are more than two, the light beads 302 in each light source assembly 30 have different radii of the circular ring formed by the circumferential array around the camera assembly 20; the light source assembly 30 with smaller circular radius is positioned at one side of the light source assembly 30 with larger circular radius, a plurality of light source assemblies are concentrically arranged, and the light beads 302 on different light source assemblies irradiate the sample 40 to be detected without shielding each other. For example, the two light source modules 30 are a first light source module and a second light source module, wherein r1 is a radius of a circular ring formed by the array of the light beads 302 in the first light source module, r2 is a radius of a circular ring formed by the array of the light beads 302 in the second light source module, and r1 > r2, and the second light source module is located at a side irradiated by the light beads 302 of the first light source module or at a side irradiated by the light beads 302. The combination of the light source components in different forms can be set according to different detection samples, so that the detection system can be used for detecting defects of more types of samples.

In this embodiment, the camera module 20 and the light source module 30 are mounted on the chute 101 through a bracket, and the chute 101 is provided on the column 10. Specifically, the chute 101 extends along the irradiation direction of the light source assembly 30, and the camera assembly 20 and the light source assembly 30 can adjust the relative position between the two along the chute 101. The lamp beads 302 are all arranged on the lamp holder 301, and the light source assembly 30 is adjusted so that all the lamp beads 302 can be adjusted at the same time without single adjustment. The array of beads 302 is circular in shape, enabling easier calibration of the orientation of each set of beads.

As shown in fig. 5, a flow chart of a method for detecting defects on a smooth surface of a semiconductor according to an embodiment of the present invention specifically includes the following steps:

s1, placing a sample 40 to be detected at a designated position under the irradiation of a light source assembly 30, and calibrating the direction of each group of lamp beads by an industrial personal computer;

s2, the industrial personal computer controls each group of lamp beads 302 in the light source assembly 30 to sequentially light and irradiate the sample 40 to be detected, and when each group of lamp beads 302 light, the camera assembly 20 shoots an image of the sample 40 to be detected and sends the image to the industrial personal computer;

s3, the industrial personal computer synthesizes an image for detecting the defects of the sample through a multi-image synthesis method according to a plurality of images acquired from the camera assembly 20 and according to the directions of each group of lamp beads 302, so that the display of the defects on the image is enhanced, and defect detection judgment is carried out through the image according to the gray scale of the defects.

In step S1, the acquiring, by the industrial personal computer, the azimuth data of each group of lamp beads 302 specifically includes: and (3) carrying out normalization processing on the position coordinates of each group of lamp beads 302, and obtaining a light source direction matrix according to the normalized position coordinates of each group of lamp beads 302.

Further, in step S3, the method of synthesizing an image for detecting a defect of a sample by using the multiple image synthesis method specifically includes: s31, processing the images to obtain gray values of corresponding pixel points in the images, wherein the pixel points at the same position in the images correspond to the same point in the sample to be detected;

s32, calculating the normal vector of each pixel point on the picture through the light source direction matrix and the gray value of each pixel point;

s33, obtaining gray mapping of each pixel point according to the normal vector of each pixel point.

The number of pixels and the arrangement of the pixels of the plurality of images are the same.

As shown in fig. 6 and 7, in this embodiment, the sample to be inspected 40 is a CMOS photosensitive chip.

In fig. 6, the 32 beads are first divided into 8 groups, namely, a first bead group, a second bead group, a third bead group, a fourth bead group, a fifth bead group, a sixth bead group, a seventh bead group and an eighth bead group. The sample 40 to be detected is placed at the position of the sample 40 to be detected, and the industrial personal computer calibrates the positions of eight groups of lamp beads.

Referring to fig. 7, the diameter of the circular light source formed by the array of the lamp beads 302 is 2r, and the angle of the plane of the sample 40 to be detected relative to the line connecting the sample 40 to be detected and the circular ring of the lamp beads 302 is 60 ° (of course, the angle can be set to other values). The distance between the plane of the lamp beads 302 and the sample 40 to be detected isr，

The line connecting the center of the ring of the lamp beads 302 and the center of the lamp bead group I and the center of the ring of the lamp bead 302 is taken as an x axis, the line connecting the center of the ring of the lamp bead 302 and the center of the lamp bead group III and the center of the ring of the lamp bead 302 are taken as a y axis, and the line connecting the center of the ring of the lamp bead 302 and the center of the camera component 20 is taken as a z axis, and r is taken as 1.

In step S1, eight groups of bead position coordinates and directions are obtained as follows:

table 1 eight sets of lamp bead position coordinates

Normalizing the position coordinates (X, Y, Z) of the first lamp bead group to the eighth lamp bead group, wherein the position coordinates are as follows:

where X, Y, Z represents the original coordinate values and X ', Y ', Z ' represent the normalized coordinate values.

Table 2 eight sets of coordinates after lamp bead position normalization

Obtaining a light source direction matrix according to the normalized coordinates of eight groups of lamp beadsThe method is characterized by comprising the following steps:

in step S2, the industrial personal computer controls the light source assembly 30 to sequentially light the first lamp bead group, the second lamp bead group … … and the eighth lamp bead group, and each light a group of lamp beads, and the camera assembly 20 shoots a picture of the sample 40 to be detected and transmits the picture to the industrial personal computer. After the photographing of all eight samples to be tested is completed, as shown in fig. 8a-8 h. The industrial personal computer combines eight pictures with the obtained light source direction matrix to synthesize eight images into one image for defect detection, as shown in fig. 10.

In step S31, a gray value of each corresponding pixel point in the plurality of images is obtained, and a gray value matrix is obtained. Processing eight chip images to be detected to obtain eight pixels with the number ofGray picture of (1), let us assume->First->Line->The pixel gray value of the column is +.>The gray value of each corresponding pixel point on eight images is taken out to obtain gray value matrix of all the pixel points +.>At this time->For 8 rows->A matrix of columns. The method comprises the following steps:

in step S32, a normal vector of each pixel on the picture is calculated by the light source direction matrix and the gray value of each pixel, which is specifically as follows:

after the normal vector of each pixel point on the picture is calculated, a unit normal vector matrix of each position of the chip is obtainedWherein->Is->Is the pseudo-inverse of the transposed matrix of (1), the normal vector matrix obtained>For 3 rows->Matrix of columns, i.e.)>，/>，/>All 1 row->Normal vector of column, wherein each element represents normal vector of current pixel +.>。

In step S33, the gray mapping of each pixel is obtained according to the normal vector of each pixel, and the resultant image with enhanced defects is calculated.

The gray mapping formula of the composite image is as follows:

where I' is the gray scale map of the image,for Fourier transform, ++>Is an inverse Fourier transform>Is imaginary unit, ++>The gray mapping is used for converting the storage type of the result in the computer, so that the calculation speed is improved, and the memory occupation is reduced. Because the numerical value type of the result of the inverse Fourier transform in the computer is double type, the comparison occupies the memory of the computerCalculating time, in view of that the gray scale change of 0 to 255 is detected enough when we detect the defect, we use +.>And mapping the double type data to gray values in a range of 0-255.

As shown in fig. 9 and 11, the CMOS light sensitive chip detects the presence of deep scratches and shallow scratches on a sample, and a composite image is obtained according to a system and method for detecting a smooth surface defect of a semiconductor in this embodiment. According to the original drawing of the sample in fig. 11, the inspector marks the deep scratches and the shallow scratches on the surface of the sample. One of which is a shallow scratch, which is not evident in the image taken in fig. 11. However, as shown in fig. 10, the gray level image of the scratch defect is enhanced in the scratch area after the image synthesis, the shallow scratch is also obviously displayed, the scratch defects with different degrees can be detected through the image, the gray level of the deep scratch is deeper, the gray level of the shallow scratch is shallower, and the severity of the defect is reflected to a certain extent.

The embodiment not only can detect the defects of the CMOS photosensitive chip, but also can detect the surface defects of the PCB and the wafer products.

As shown in fig. 12a-12h, fig. 13 and fig. 14, eight sample images of the PCB in fig. 12a-12h are obtained, the images for detection as shown in fig. 14 are synthesized, and the sample defects are judged according to the gray scale on the images, so that the scratch defects can be detected obviously.

Similarly, as shown in fig. 15a to 15h, fig. 16 and fig. 17, the sample images of eight wafers are acquired at 15a to 15h, the images for detection as shown in fig. 17 are synthesized, and the defects of the pits on the surface of the wafer can be detected remarkably by judging the defects of the sample based on the gray scale on the images.

In the embodiment, only eight sample chip pictures under irradiation of different angles need to be shot, and the detection speed is high. The number of pictures can be increased or decreased according to the requirement.

In addition, the detection system provided by the embodiment of the invention has a simple structure, the detection system and the detection method can be compatible with various samples, and based on the difference of different samples, the embodiment of the invention realizes the following specific scheme of compatibility:

1. for different specific forms of the sample, the embodiment of the invention can synthesize the image of the object in the whole visual field range of the camera, and the combination form of the lamp beads in the light source component is flexible and changeable, so that the effect of the specific form of the sample can be avoided.

2. For different sizes of sample products, if the sample products are larger, the light sources with different sizes can be flexibly replaced, the camera is lifted to adjust the field of view of the camera or the lens with larger field of view is replaced, and as long as the sample products can be completely covered by the light sources and the camera can shoot the whole sample products, the scheme can shoot the sample products, synthesize images and highlight defects.

3. For the condition that the sample product is smaller or the required precision is higher, the embodiment of the invention can flexibly replace cameras with different precision according to the required precision, the precision of the result calculated by the high-precision camera is correspondingly higher, and the sample and scratch which can be detected can be smaller.

4. For samples with different colors or samples with different color light sources affecting the detection result, the embodiment of the invention can flexibly adjust the color of the lamp beads to achieve the best detection effect.

5. For objects with different reflectivities, the embodiment of the invention can flexibly change the number of the lamp beads or adjust the brightness of the lamp beads so as to achieve the best detection effect.

Claims

1. A semiconductor smooth surface defect inspection system comprising a camera assembly (20) and at least one light source assembly (30), and an analysis device for transmitting information to and from the camera assembly (20), characterized in that: the light source assembly (30) comprises a plurality of lamp beads (302), the lamp beads (302) are circularly arrayed around the camera assembly (20) in a central circumferential direction, and the camera assembly (20) is positioned at one side far away from the light source assembly (30) for irradiation; the lamp beads (302) are divided into A groups, and in each group, a plurality of lamp beads form a strip-shaped light source; each group of lamp beads (302) in the light source assembly (30) is lightened to irradiate the sample (40) to be detected.

2. The semiconductor smooth surface defect detection system of claim 1, wherein: when the number of the light source assemblies (30) is more than two, the radius of a circular ring formed by the lamp beads (302) in each light source assembly (30) circumferentially arrayed around the center of the camera assembly (20) is different; the plurality of light source components are concentrically arranged, and the lamp beads (302) on different light source components irradiate the sample (40) to be detected without shielding each other.

3. The semiconductor smooth surface defect detection system of claim 1, wherein: the lamp beads are arranged on the lamp holder, and the lamp holder (301) is in a circular ring shape.

4. A semiconductor smooth surface defect detection system according to any of claims 1-3, wherein: still include stand (10), set up spout (101) on stand (10), camera subassembly (20) and light source subassembly (30) are installed on spout (101) through the support.

5. The semiconductor smooth surface defect detection system of any one of claims 4, wherein: the sliding groove (101) extends along the irradiation direction of the light source assembly (30), and the camera assembly (20) and the light source assembly (30) can adjust the relative position between the camera assembly and the light source assembly along the sliding groove (101).

6. A method for detecting defects on a smooth surface of a semiconductor, which adopts the system for detecting defects on a smooth surface of a semiconductor according to any one of claims 1 to 5, characterized in that:

placing a sample (40) to be detected at a designated position under the irradiation of a light source assembly (30), and calibrating the direction of each group of lamp beads;

controlling each group of lamp beads (302) in the light source assembly (30) to be sequentially lightened to irradiate the sample (40) to be detected, and when each group of lamp beads (302) are lightened, the camera assembly (20) shoots an image of the sample (40) to be detected and sends the image to analysis equipment;

the analysis device synthesizes an image for detecting the defects of the sample by a multi-image synthesis method according to a plurality of images acquired from the camera assembly (20) and according to the azimuth of each group of lamp beads (302), so that the display of the defects on the image is enhanced, and defect detection judgment is carried out according to the gray scale of the defects through the image.

7. The method for detecting a smooth surface defect of a semiconductor according to claim 6, wherein: synthesizing an image for detecting defects of a sample by a multiple image synthesis method comprises the following steps:

and obtaining gray mapping of each pixel point according to the normal vector of each pixel point, and synthesizing an image.

8. The method for detecting a smooth surface defect of a semiconductor according to claim 7, wherein: the gray mapping formula of the composite image is as follows:

in the middle ofFor Fourier transform, ++>Is an inverse Fourier transform>Is imaginary unit, ++>For grey scale mappingShooting; the number of pixels per gray picture is +.>；/>；/>，/>，/>All 1 row->Normal vector of column pixel points.

9. The method for detecting a smooth surface defect of a semiconductor according to claim 7, wherein: the number of the pixels of the plurality of images is the same, and the arrangement modes of the pixels are the same.

10. The method for detecting a smooth surface defect of a semiconductor according to any one of claims 7 to 9, wherein: the method also comprises the following steps: the acquisition of azimuth data of each group of lamp beads (302) is specifically as follows: and carrying out normalization processing on the position coordinates of each group of lamp beads (302), and obtaining a light source direction matrix according to the normalized position coordinates of each group of lamp beads (302).