A method and device for detecting filter defects

Abstract

A method and device for detecting filter defects of the present invention include: parallel light passing through a filter falls on the sensor of a camera; performing flat-field correction on the camera and saving the flat-field correction coefficient; changing the relative pose between the camera and the filter with respect to the parallel light so that the position where the parallel light passes through the filter and projects on the sensor changes relatively; keeping the flat-field correction coefficient of the camera unchanged and collecting pictures with different relative poses between the camera or the filter and the parallel light; preprocessing the pictures, finding the areas where the gray values change in the pictures, and obtaining the defect detection result based on the preprocessed pictures. The present invention detects defects by the change in gray values at the defect locations before and after flat-field correction, can detect defect areas that are difficult to directly see with the human eye, can detect dust that is close to the background, and is applicable to scenarios with relatively high detection accuracy.

Description

Technical Field

[0001] The present invention relates to the field of optical detection, and particularly to a method for detecting filter defects. Background Art

[0002] Currently, coated filters are widely used in fields such as cameras, playing a role in filtering stray light and protecting the sensor. When there are defects on their surfaces, it will have a greater impact on the imaging effect of the camera, and their applications in fields such as precision detection will be greatly affected. Therefore, in order to improve the imaging effect of the camera and the accuracy of detection, it is necessary to accurately detect the defects of the filter.

[0003] A method for locating glass defects is proposed in Chinese Patent CN115018829A, which requires comparing the image obtained by photographing the filter to be detected with the image obtained by photographing a defect-free filter. The detection accuracy depends on the cleanliness of the defect-free filter. A method and system for imaging lens defects are proposed in Chinese Patent CN115096912A, which uses different image patterns for imaging different types of defects. The detection accuracy is related to the image pattern and it is difficult to cover all types of defects.

[0004] In summary, the disadvantages of the prior art are as follows:

[0005] (1) Existing detection methods often only collect images from one angle and compare them with the images collected by a reference defect-free filter, which have high requirements for the light source, reference filter, and collection equipment, and it is difficult to completely avoid problems caused by uneven illumination.

[0006] (2) As the detection progresses, new dust will inevitably fall on the sensor or filter, and existing methods cannot filter out such dust or defects.

[0007] (3) For defects that are close to the background, it is difficult for the prior art to distinguish and detect them well. Summary of the Invention

[0008] A method and device for detecting filter defects proposed by the present invention can solve at least one of the above technical problems.

[0009] To achieve the above object, the present invention proposes the following technical solutions:

[0010] A method for detecting filter defects includes:

[0011] Parallel light passing through the filter is projected onto the sensor of the camera;

[0012] Perform flat-field correction on the camera and save the flat-field correction coefficient;

[0013] Change the relative pose between the camera or the filter and the parallel light so that the position where the parallel light passes through the filter and projects onto the sensor changes relatively;

[0014] Keep the flat-field correction coefficient of the camera unchanged, and capture images with different relative poses between the camera or the filter and the parallel light;

[0015] Preprocess the images, find the areas where the gray values change in the images, and obtain the defect detection result based on the preprocessed images.

[0016] Further, keeping the flat-field correction coefficient of the camera unchanged and capturing images with different relative poses between the camera or the filter and the parallel light includes:

[0017] Adjust the pose of the camera and the filter to be measured relative to the parallel light so that the parallel light vertically passes through the filter and falls on the camera sensor. Keep the flat-field correction coefficient of the camera unchanged and capture Image 1;

[0018] With the incident angle of the parallel light unchanged, rotate the camera and the filter to be measured together to the left and right by a set angle respectively. Keep the flat-field correction coefficient unchanged, and the camera captures Image 2 and Image 3 respectively after the rotation.

[0019] Further, preprocessing the images, finding the areas where the gray values change in the images, and obtaining the defect detection result based on the preprocessed images includes:

[0020] Perform difference operations between Image 1 and Image 2 and Image 3 respectively to obtain two difference images;

[0021] Merge the defect areas on the two difference images. The defect areas are the areas where the gray values change, and obtain the defect detection result.

[0022] Further, it also includes:

[0023] After capturing Image 2 and Image 3, adjust the pose of the camera and the filter to be measured relative to the parallel light again so that the parallel light vertically passes through the filter and falls on the camera sensor;

[0024] Remove the filter to be measured and capture Image 4;

[0025] Compare Image 4 with Image 1, Image 2, and Image 3 to determine whether new defects have occurred on the filter and the camera sensor during the detection process.

[0026] Further, it also includes:

[0027] Before the camera performs flat-field correction, adjust the light intensity of the parallel light and the exposure time of the camera so that the gray values of the images captured by the camera are within a set range.

[0028] Further, it further includes:

[0029] After collecting Picture 2 and Picture 3, rotate the filter to be measured, and the rotation is to rotate the filter 180 degrees to the left or right within its plane;

[0030] Keep the flat field correction coefficient of the camera unchanged, and collect Picture 4, Picture 5 and Picture 6 at the same positions as Picture 1, Picture 2 and Picture 3 respectively.

[0031] Further, preprocess the picture, find the area where the gray value changes in the picture, and obtain the defect detection result according to the preprocessed picture, including:

[0032] Perform difference between Picture 1 and Picture 2, and between Picture 1 and Picture 3 respectively to obtain two difference images, and merge the defect areas on the two difference images to obtain Picture 7;

[0033] Perform difference between Picture 4 and Picture 5, and between Picture 4 and Picture 6 respectively to obtain two difference images, and merge the defect areas on the two difference images to obtain Picture 8;

[0034] Extract the repeated defect areas in Picture 7 and Picture 8, where the defect area is the area where the gray value changes, merge the repeated defect areas in Picture 7 and Picture 8 to obtain Picture 9, and obtain the defects on the sensor according to Picture 9.

[0035] Further, it further includes:

[0036] Perform difference between Picture 9 and Picture 7, and between Picture 9 and Picture 8 respectively to obtain two difference images as Picture 10 and Picture 11;

[0037] Rotate Picture 10, and the rotation is to rotate Picture 10 180 degrees to the left or right within its plane to obtain Picture 101;

[0038] Extract the repeated defect areas in Picture 101 and Picture 11, merge the repeated defect areas in Picture 101 and Picture 11 to obtain Picture 12, and merge the non-repeated defect areas in Picture 101 and Picture 11 to obtain Picture 13;

[0039] Obtain the inherent defects on the filter to be measured according to Picture 12, and obtain the newly added defects on the filter to be measured during the detection process according to Picture 13.

[0040] On the other hand, the present invention proposes a filter defect detection device, including:

[0041] An integrating sphere light source for generating uniform parallel light;

[0042] A camera, with the filter to be measured installed in front of the camera sensor; the filter to be measured is located between the integrating sphere light source and the camera;

[0043] A rotation angle stage for fixing the camera and driving the camera to rotate within a set angle;

[0044] A host computer for calculating the defect detection result of the filter to be tested according to the above filter defect detection method.

[0045] Furthermore, a filter holder is installed in front of the camera sensor, and the filter holder includes:

[0046] A frame fixedly connected to the camera, on which a filter support is slidably connected, and the filter to be tested is detachably installed on the filter support. The filter to be tested is installed at the corresponding position of the camera by sliding the filter support into the frame.

[0047] The beneficial effects of the present invention are as follows:

[0048] (1) The present invention detects defects by the change in the gray value at the defect location before and after flat field correction, can detect defect areas that are difficult to directly see with the human eye, can detect dust that is close to the background, and is applicable to scenarios with high detection accuracy;

[0049] (2) Due to the adoption of flat field correction, the requirements for the light source are reduced, and the cost of the detection equipment can be reduced;

[0050] (3) It can distinguish between the inherent defects of the filter, the defects generated during the detection process, and the inherent defects of the sensor, improving the reliability and accuracy of the detection. Description of the Drawings

[0051] Figure 1 is a flowchart of the filter defect detection method of the present invention;

[0052] Figure 2 is a structural diagram of the filter defect detection device of the present invention;

[0053] Figure 3 is a structural diagram of the installation of the filter in the filter defect detection device of the present invention.

[0054] In the figure: 1 - integrating sphere light source; 2 - rotation angle stage; 3 - large area array camera; 4 - frame; 5 - filter support; 6 - filter; 7 - angle scale; 8 - limit post. Detailed Embodiments

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in 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. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.

[0056] Embodiment 1

[0057] This embodiment proposes a filter defect detection device, the overall structure of which is as shown in Figure 1 , including:

[0058] An integrating sphere light source 1 and a rotation angle stage 2 are installed on the platform base. A camera is installed on the rotation angle stage 2. The camera used in this embodiment is a large area array camera 3, which can better detect the filter 6 to be measured. The filter 6 to be measured is installed in front of the sensor of the large area array camera 3 through a filter holder. Among them, the integrating sphere light source 1, the filter holder, and the large area array camera 3 are located on the same axis, and the filter holder is located between the integrating sphere light source 1 and the large area array camera 3.

[0059] For the filter holder in front of the sensor of the large area array camera 3 to facilitate the replacement of the filter 6 and the change of the direction in which the filter 6 is placed, a frame 4 of the filter holder and a filter support 5 are designed. The frame 4 is fixed in front of the sensor of the large area array camera 3, and there is a rectangular through hole in the middle to accommodate the filter 6 and allow the light source to pass through. The filter support 5 is slidably connected to the frame 4, and in this embodiment, a hinge connection is adopted. The filter 6 is detachably installed on the filter support 5, which is convenient for replacing the filter 6 or changing the direction in which the filter 6 is inserted into the filter support 5. The filter support 5 drives the filter 6 to slide into the frame 4, and the filter 6 is installed in the frame 4 through the positioning structure in the frame 4. The filter holder designed in this embodiment facilitates the installation and removal operations of the filter 6 and can adapt to the detection of a large number of filters.

[0060] The integrating sphere light source 1 is used to emit uniform light to ensure the stability of the illumination as much as possible; and the integrating sphere light source 1 can control the light field intensity, adjust the gray value of the image, and control the light to pass through the lens to be detected for irradiation to ensure that the illumination in the image acquisition area is as uniform as possible. In this embodiment, uniform parallel light needs to be used as the light source to complete the subsequent detection. Because after the parallel light passes through the filter 6, the defect is projected onto the sensor at only one fixed position. When the relative position relationship between the parallel light and the filter 6 changes, this position will also change. However, the light source in this embodiment can be other light source devices that emit parallel light. Therefore, other light sources that can emit parallel light should also be within the scope of the present invention.

[0061] The rotation angle stage 2 is used to carry the large area array camera 3 and drive the large area array camera 3 to rotate by a small angle. The large area array camera 3 is used to collect filter images according to the set acquisition strategy. The filter holder is used to place the filter 6 and avoid the trouble of disassembly and assembly when replacing the filter 6.

[0062] It also includes a host computer, which is used to execute the following detection method and calculate the detection result according to the filter images collected by the large area array camera 3.

[0063] As shown in Figure 2 and Figure 3Shown is the detection tooling of the filter flaw detection imaging device of this embodiment, including the angle scale 7 on the rotation angle stage 2, which is used to fix the large area array camera 3 and drive the large area array camera 3 to rotate; the angle scale 7 is rotatably connected to the rotation angle stage 2, and through holes are provided on the angle scale 7, and the limit posts 8 on the rotation angle stage 2 pass through the through holes. Through the limit posts 8 and the through holes, the rotation angles of the large area array camera 3 to the left and right can be restricted to be the same. The large area array camera 3 and the angle scale 7 are fixedly connected through a fixing device; among them, the large area array camera 3 does not install a lens, a filter holder is installed at the lens mount, the filter holder is adjacent to the sensor of the large area array camera 3, and the filter 6 to be detected is placed on the filter holder, so that the collected image has an approximately uniform background. Reflecting on the image, it is: a 10×10 pixel area is selected on the image, and the average difference in gray value between the flaw part and the image background before processing is within the range of 2-3 pixels.

[0064] The detection principle of this embodiment is as follows:

[0065] The flaw points of the filter 6 include dust falling on the filter 6 and flaws generated during the production, manufacturing, transportation, and handling of the filter 6. These flaws will affect the passage of light. When the incident light is parallel light, the flaws on the filter 6 will weaken the light intensity passing through the flaw area, resulting in a smaller light response in the flaw area projected onto the sensor. If the flaw is small, it is difficult to distinguish this difference in light response by general detection means. Judging the flaw points through the method of flat field correction can amplify this difference in light response. Under the existing sensor acquisition accuracy, the resolution of this difference can be improved. In addition, after flat field correction, the influence of uneven illumination on the detection result can be eliminated, the requirement for the light source is reduced, and the cost of the detection equipment is reduced. And the amplification principle of the flat field correction of the present invention is to amplify this response difference from the sensor level through physical methods, rather than through mathematical amplification assisted by software. Therefore, the detection method of the present invention can identify flaw points that are very close to the background.

[0066] When the large area array camera 3 faces the light source directly, the large area array camera 3, the filter to be measured 6 and the integrating sphere light source 1 are on the same axis. At this time, the light source passes through the filter 6 and projects onto the sensor. The connection line between the defect point on the filter 6 and the corresponding projection area on the sensor is perpendicular to the filter to be measured 6. At this time, flat field correction is performed. Flat field correction is used to eliminate the non-uniformity of the response of each pixel point and is a relatively mature technology. Flat field correction uses flat field correction coefficients to correct the non-uniform response of different pixel points, and the correction coefficients of different pixel points are different. The light passing through the defect area is dimmer, so the correction coefficient of the corresponding pixel point area on the sensor is larger; at this time, the camera is rotated by a certain angle through the rotation angle stage 2. The connection line between the defect point on the filter 6 and the corresponding projection area on the sensor is no longer perpendicular to the filter to be measured 6, and the projection area also deflects relative to the central position. The original projection area of the defect point has no defect projection at this time, but because the correction coefficient remains unchanged, this position will become brighter relative to the background, and the new projection area of the defect point will become darker. In the image, it will be shown as two circles with prominent light and dark. That is, the flat field correction coefficients for the defect-free area and the defect area are different, and the coefficient remains unchanged after correction. During the process of rotating the camera, the position of the defect moves, and the gray value change of the new position and the old position of the defect is relatively prominent due to the different correction coefficients. By capturing this change, the position of the defect can be found and the position of the defect can be judged. However, the rotation at a single angle increases the risk of missed detection, so it is necessary to rotate the same angle to the left and right respectively to prevent some special defect points from not being recognized.

[0067] The detection process of this embodiment is as follows:

[0068] Insert the filter to be measured 6 into the filter holder 5, and rotate the filter holder 5 to install the filter to be measured 6 on the filter rack.

[0069] Adjust the light intensity of the integrating sphere light source 1 and the exposure time of the large area array camera 3 to make the background light illumination uniform, and the gray value of the image collected by the large area array camera 3 is within the set reasonable range, which is convenient for subsequent defect detection.

[0070] The large area array camera 3 faces the integrating sphere light source 1 directly. Here, facing directly means that the parallel light emitted by the integrating sphere light source 1 vertically passes through the filter to be measured 6 and reaches the sensor. At this time, flat field correction is performed on the camera, and the flat field correction coefficient is saved. Later, when the large area array camera 3 takes pictures at different angles, this flat field correction coefficient is used to obtain images.

[0071] The large area array camera 3 respectively captures images when facing the light source directly, rotating left by a preset angle, and rotating right by a preset angle, which are respectively recorded as Picture 1, Picture 2, and Picture 3. Additionally, when the large area array camera 3 faces the camera directly, the filter 6 to be tested is taken out, and a filterless picture N is captured. The setting process of the above-mentioned preset angle is as follows: after rotating left by a certain angle, for the image captured by the large area array camera 3, since the flat field correction coefficient remains unchanged, the original defect area will become brighter, and the new position where the defect area is projected will become darker, forming two bright and dark circles. When these two circles are tangent, the angle by which the large area array camera 3 rotates at this time meets the requirements, and the set angle for rotating right is the same as the set angle for rotating left.

[0072] Perform differential processing on the above-mentioned captured images, and then process the images through connected component analysis to obtain the defects of the filter 6 and the position information of the defects. The differential image reflects the change magnitude of the gray value of each pixel point. The defect area and the background area can be distinguished through a change threshold. The purpose of this embodiment is to find the area where the gray value changes in the image, and after processing, a defect detection result is obtained. Therefore, if other methods are used to analyze the change in the gray value to obtain defect information, it should also be considered within the scope described in this specification.

[0073] Specifically, Picture 1 is respectively subjected to differential processing with Picture 2 and Picture 3 to obtain Differential Picture 1 and Differential Picture 2; the defect areas on the two differential images are merged, and the defect area is the area where the gray value changes; the image is further processed from the perspective of connected components to obtain a binary image. If other methods are used to obtain the binary image, it should also be considered within the scope described in this specification. Analyze and obtain the defect detection result on the filter 6, including information such as the shape, size, and position of the defect; and further determine whether the filter 6 is qualified, and it can detect dust or defect points that are very close to the background.

[0074] Finally, compare Picture 1, Picture 2, and Picture 3 with Picture N respectively to observe whether there are new defects on the sensor and the filter 6 to be tested. The new defects include dust that has fallen or other new defects generated during the detection process.

[0075] Embodiment 2

[0076] In Embodiment 1, only two rotations are performed on the filter 6 to be tested and the camera, which can achieve the detection of the defects of the filter 6. However, during the detection process, new dust will inevitably intervene or new defects will occur on the filter 6 to be tested and the sensor of the large area array camera 3. Embodiment 1 of the present invention solves this problem by taking a filterless picture N. This method only compares pictures and has a poor effect. Therefore, on the basis of Embodiment 1, Embodiment 2 of the present invention uses the same detection device and proposes an improved detection method to solve the above problem.

[0077] The detection process of Embodiment 2 of the present invention is as follows:

[0078] After connecting to the large area array camera 3, collect images in 12-bit mode and observe the images through the lower 8 bits. Adjust the appropriate exposure time and the light field intensity of the integrating sphere light source 1 until the difference between the defect points and the background can be clearly seen. Some defect points close to the background are difficult to see and need further processing. When saving the images, save the lower 8-bit images.

[0079] Collect images, which are divided into:

[0080] (1) Collect the same pictures 1, 2, and 3 in the same method as in Embodiment 1;

[0081] (2) Take out the filter 6 to be measured from the filter holder, rotate it and then put it back into the filter holder. After rotation, the normal vector of the plane where the filter 6 is located rotates by 180 degrees;

[0082] (3) Except for rotating the filter 6, the rest of the device and the flat-field correction coefficient of the camera remain unchanged. Then collect pictures 4, 5, and 6 at the same positions as pictures 1, 2, and 3 above.

[0083] Picture 1 is differentiated from pictures 2 and 3 respectively to obtain two differential images, differential picture 1 and differential picture 2. The differential images reflect the change magnitude of the gray level of each pixel point. The defect area and the background area can be distinguished through the change threshold. The defect areas on the two differential images are merged to obtain the first defect image, which is picture 7. In order to obtain a more obvious boundary area, further downsampling operation is performed on picture 7 here.

[0084] Picture 4 is differentiated from pictures 5 and 6 respectively to obtain two differential images, differential picture 3 and differential picture 4. The defect area and the background area can be distinguished through the change threshold. The defect areas on the two differential images are merged to obtain the second defect image, which is picture 8. Similarly, downsampling operation is performed on picture 8.

[0085] Find the defect areas that exist repeatedly on pictures 7 and 8, and extract them separately to obtain picture 9. During the rotation process, the corresponding relationship between the filter 6 and a single pixel point on the sensor will change, but the defects on the sensor itself will not change due to this. Therefore, the defect area that has not changed is the defect on the sensor, that is, picture 9 is the defect image on the sensor.

[0086] Images 7 and 8 are respectively differentiated from Image 9 to filter out the defects on the sensor, resulting in Images 10 and 11. Image 10 is rotated to obtain Image 101. Since Filter 6 rotated during image acquisition, here the image is rotated back so that the normal vector of the plane where the rotated Image 10 is located rotates 180 degrees. The rotation here is opposite to the rotation of Filter 6 mentioned above. Then, Image 101 is compared with Image 11 to extract the repeatedly occurring defects, denoted as Image 12. The defects represented by Image 12 are the defects on Filter 6, and the remaining non-repeated defects are the defects added during the process of replacing Filter 6, denoted as Image 13.

[0087] Finally, combining Image 12, Image 13, and Image 9, analyze whether the tested Filter 6 is qualified according to the set judgment criteria.

[0088] Image 12, Image 13, and Image 9 obtained by the method of this embodiment can respectively reflect the inherent defects on Filter 6, the defects added to Filter 6 during the detection process, and the defects on the sensor. Compared with the detection method in Embodiment 1, it can more accurately distinguish different types of defects, and the information of the detection result is richer.

[0089] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0090] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting filter defects, characterized in that, Including: Project parallel light passing through a filter onto the sensor of a camera; Perform flat-field correction on the camera and save the flat-field correction coefficient; Change the relative pose between the camera, the filter, and the parallel light so that the position where the parallel light passes through the filter and projects onto the sensor changes relatively; The camera keeps the flat-field correction coefficient unchanged and acquires images with different relative poses between the camera, the filter, and the parallel light; Preprocess the said images, find the areas where the gray values change in the images, and obtain the defect detection result based on the preprocessed images; Among them, the camera keeps the flat-field correction coefficient unchanged and acquires images with different relative poses between the camera, the filter, and the parallel light, including: Adjust the pose of the camera and the filter to be measured relative to the parallel light so that the parallel light perpendicularly passes through the filter and falls on the camera sensor. The flat-field correction coefficient of the camera remains unchanged, and Image 1 is acquired; With the incident angle of the parallel light unchanged, the camera and the filter to be measured rotate left and right by a set angle respectively. The flat-field correction coefficient remains unchanged, and Images 2 and 3 are acquired by the camera after the rotation respectively; The preprocessing of the said images, finding the areas where the gray values change in the images, and obtaining the defect detection result based on the preprocessed images, including: Image 1 is differentiated from Images 2 and 3 respectively to obtain two differential images; the defect areas on the two differential images are merged. The defect areas are the areas where the gray values change, and the defect detection result is obtained; Among them, the process of setting the angle is as follows: after rotating left by a certain angle, for the image acquired by the camera, due to the unchanged flat-field correction coefficient, the original defect area will become brighter, and the new position of the projection of the defect area will become darker, forming two bright and dark circles; the set angle for rotating right is the same as the set angle for rotating left.

2. The filter defect detection method according to claim 1, wherein Also including: After acquiring Images 2 and 3, adjust the pose of the camera and the filter to be measured relative to the parallel light again so that the parallel light perpendicularly passes through the filter and falls on the camera sensor; Remove the filter to be measured and acquire Image 4; Image 4 is compared with Images 1, 2, and 3 to determine whether new defects have occurred on the filter and the camera sensor during the detection process.

3. The filter defect detection method according to claim 1, wherein Also including: Before the camera performs flat-field correction, adjust the light intensity of the parallel light and the exposure time of the camera so that the gray value of the image acquired by the camera is within a set range.

4. The filter defect detection method according to claim 1, wherein Also including: After acquiring Images 2 and 3, rotate the filter to be measured. The rotation is that the filter rotates 180 degrees left or right in its plane; The camera keeps the flat-field correction coefficient unchanged and acquires Images 4, 5, and 6 at the same positions as Images 1, 2, and 3 respectively.

5. The filter defect detection method according to claim 4, wherein The preprocessing of the said images, finding the areas where the gray values change in the images, and obtaining the defect detection result based on the preprocessed images, including: Image 1 is differentiated from Images 2 and 3 respectively to obtain two differential images, and the defect areas on the two differential images are merged to obtain Image 7; Image 4 is differentiated from Images 5 and 6 respectively to obtain two differential images, and the defect areas on the two differential images are merged to obtain Image 8; Extract the overlapping defect regions in Image 7 and Image 8, where the defect regions are areas with changing grayscale values. Merge the overlapping defect regions in Image 7 and Image 8 to obtain Image 9. Determine the defects on the sensor based on Image 9.

6. The filter defect detection method according to claim 5, wherein Further included are: Performing differential operations on Image 9 with Image 7 and Image 8 respectively to obtain two differential images, namely Image 10 and Image 11; Rotating Image 10, where the rotation is a 180-degree left or right rotation of Image 10 within its plane to obtain Image 101; Extract the overlapping defect regions in Image 101 and Image 11, merge the overlapping defect regions in Image 101 and Image 11 to obtain Image 12, and merge the non-overlapping defect regions in Image 101 and Image 11 to obtain Image 13; Determine the inherent defects on the filter under test based on Image 12, and determine the newly added defects on the filter under test during the detection process based on Image 13.

7. A filter defect detection device, characterized in that, Included are: An integrating sphere light source for generating uniform parallel light; A camera with the filter under test installed in front of the camera sensor; the filter under test is located between the integrating sphere light source and the camera; A rotation angle stage for fixing the camera and driving the camera to rotate within a set angle; A host computer for calculating the defect detection result of the filter under test according to the filter defect detection method described in any one of Claims 1-6.

8. The filter defect detection device according to claim 7, wherein A filter holder is installed in front of the camera sensor, and the filter holder includes: A frame fixedly connected to the camera. A filter support is slidably connected to the frame, and the filter under test is detachably installed on the filter support. The filter under test is installed at the corresponding position of the camera by sliding the filter support into the frame.

Images