Appearance detection method and system for semiconductor chip, and medium

Abstract

The invention discloses an appearance detection method and system for a semiconductor chip and a medium, and the detection method comprises the steps: creating a detection parameter set corresponding to the semiconductor chip, and the detection parameter set comprises standard appearance data and optical system data of the semiconductor chip; adjusting the optical device according to the optical system data; respectively acquiring a front image and a back image of the semiconductor chip; respectively inputting the front image and the back image into a front detection algorithm set and a back detection algorithm set, wherein the front detection algorithm set and the back detection algorithm set respectively carry out at least one defect type detection on the front image or the back image and output a detection result; and dividing qualified products and unqualified products of the semiconductor chips according to the detection result. The method can more accurately and efficiently detect the appearances of different types of semiconductor chips.

Description

Technical Field

[0001] The present invention relates to the field of visual inspection technology, and in particular to a method, system and medium for inspecting the appearance of a semiconductor chip. Background Art

[0002] In the field of semiconductor packaging and testing, the existence of defects in a chip is directly related to the performance and life of the final product, so chip visual inspection is indispensable. However, several common appearance inspection methods currently have some disadvantages. For example, the turret chip appearance inspection requires all the cut chips to be poured into the vibration plate to adjust the direction before feeding the inspection, which usually causes scratches on the chip surface coating, etc., and the inspection is completely separate from the cutting, which increases the manual loading and unloading actions and reduces production efficiency. Secondly, there are mainstream integrated cutting inspection machines, which are mainly chip cutting equipment developed in South Korea and Japan. The teaching process of visual inspection of new products for this type of equipment is cumbersome, the inspection efficiency is low, and the new function development cycle is too long. The software quotation is extremely high and the versatility is not high. Summary of the invention

[0003] In order to overcome the above disadvantages, an object of the present invention is to provide a method, system and medium for detecting the appearance of semiconductor chips, which can detect the appearance of different types of semiconductor chips more accurately and efficiently.

[0004] In order to achieve the above purpose, the technical solution adopted by the present invention is: a method for detecting the appearance of a semiconductor chip, comprising: Creating a detection parameter collection corresponding to the semiconductor chip, wherein the detection parameter collection includes standard appearance data and optical system data of the semiconductor chip; adjusting an optical device according to the optical system data; Respectively acquiring a front side image and a back side image of the semiconductor chip; Inputting the front image and the back image into a front detection algorithm set and a back detection algorithm set respectively, wherein the front detection algorithm set and the back detection algorithm set respectively perform at least one defect type detection on the front image or the back image and output the detection result; The semiconductor chips are sorted into qualified products and unqualified products according to the detection results. The beneficial effect of the present invention is that different detection parameter sets are constructed according to different semiconductor chips, which is convenient for subsequent rapid call. The front detection algorithm set and the back detection algorithm set perform image detection at the same time, which can improve the detection speed and quickly output the detection results.

[0005] Further, the standard appearance data includes the standard size of the product, the number of products to be inspected in the field of view, printing, and defect area, and the standard appearance data is manually input or automatically written through a teaching training of the semiconductor chip; The optical system parameters include camera exposure time and power brightness, and the optical system parameters are input manually. Furthermore, the front detection algorithm set and the back detection algorithm set each include a plurality of detection algorithms, each of which is used for detecting at least one defect type of the semiconductor chip; The detection algorithms in the front detection algorithm set and the back detection algorithm set can detect one front image or one back image at the same time.

[0006] Furthermore, the detection algorithm can be taught and trained by acquiring a front side image or a back side image of the semiconductor chip.

[0007] After enabling the automatic teaching and automatic parameter writing functions, the teaching time can be greatly shortened.

[0008] Furthermore, when the number of defects generated by a detection algorithm for the same semiconductor chip exceeds a set threshold, an alarm is issued to remind the user to retrain the detection algorithm. Specifically, the front image and the back image are input into the front detection algorithm set and the back detection algorithm set respectively, and at the same time, the front image and the back image of another semiconductor chip are acquired. Image acquisition and algorithm detection are performed simultaneously, which improves the detection speed.

[0009] Further, dividing the semiconductor chips into qualified products and unqualified products according to the detection results specifically includes: When the semiconductor chip is a non-conforming product and there are a plurality of defect types, the semiconductor chip is placed in a corresponding waste tray according to the priority of the defect types.

[0010] The present invention also discloses an appearance inspection system for semiconductor chips, which adopts the appearance inspection method described in the claims. The appearance inspection system includes an optical device and a controller. The optical device is used to collect a front image and a back image of the semiconductor chip. The controller includes: A storage module, wherein a detection parameter set corresponding to the semiconductor chip is stored in the storage module, the detection parameter set includes standard appearance data and optical system data of the semiconductor chip, and the optical device is adjusted according to the optical system data; A cache module, the cache module is used to store the front image and the back image; A detection module, wherein a front detection algorithm set and a back detection algorithm set are stored in the detection module, wherein the front detection algorithm set and the back detection algorithm set respectively detect at least one defect type on the front image or the back image and output a detection result; A shunting module is used to shun the semiconductor chips into qualified products and unqualified products according to the shunting instructions below the detection results.

[0011] Further, the optical device comprises a front optical collection mechanism and a back optical collection mechanism, and the front optical collection mechanism and the back optical collection mechanism are used to collect the front image and the back image respectively; The front optical collection mechanism comprises a coaxial first camera, a first annular light source and a second annular light source, wherein the first annular light source and the second annular light source are located on a side close to the lens of the first camera, and the first annular light source and the second annular light source are shadowless light and polarized light respectively; The back optical collection mechanism includes a second camera, a third annular light source and a fourth coaxial light source. The fourth coaxial light source is closely attached to the lens of the second camera, and the third annular light is coaxially arranged with the second camera.

[0012] The present invention further discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the appearance detection method described above are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a flow chart of Embodiment 1 of the present invention; Figure 2 is a system block diagram of Embodiment 2 of the present invention; Figure 3 is a system block diagram of a controller in Embodiment 2 of the present invention; Figure 4 This is a schematic diagram of the structure of the front optical collection mechanism in the second embodiment of the present invention; Figure 5 It is a structural schematic diagram of the back optical collection mechanism in the second embodiment of the present invention; Figure 6 Schematic diagram of the interface display of the detection results in the first embodiment of the present invention. DETAILED DESCRIPTION

[0014] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0015] Embodiment 1 The appearance inspection method of the semiconductor chip of the present invention is used for the appearance inspection of the semiconductor chip, see the attached Figure 1 As shown, the appearance inspection method includes: S100: Create a detection parameter collection corresponding to a semiconductor chip.

[0016] The detection parameter collection includes standard appearance data and optical system data of the semiconductor chip, the standard appearance data includes standard product size, the number of products to be detected in the field of view, printing, and defect area, and the standard appearance data is manually input or automatically written through teaching and training of a semiconductor chip; the optical system parameters include camera exposure time and power brightness, and the optical system parameters are manually input.

[0017] In this embodiment, the detection parameter set corresponds to the type of semiconductor chip. When the same semiconductor chip is tested later, the detection parameter set of this semiconductor chip can be directly called. The types of semiconductor chips include QFN (quadratic flat no-lead package), BGA (ball grid array package) and LGA (grid grid array package), etc. The size of each type of product can be compatible with 2mm*2mm to 25mm*25mm. Different models of the same semiconductor chip only differ in the characters on the front of the chip, and the back of the chip is basically the same.

[0018] The optical system data of different types of semiconductor chips are basically unchanged, so when creating a collection of detection parameters, the existing optical system data can be called. When individual data changes, the parameters can be manually modified.

[0019] In one embodiment, the detection parameter collection is saved in the form of an .XML file and may be named after the semiconductor chip for easy subsequent search.

[0020] S200, adjusting the optical device according to the optical system data.

[0021] Adjust the optical device to the optimal position to ensure the clarity of the photos taken by the optical device and facilitate subsequent image processing.

[0022] S300 , respectively acquiring a front side image and a back side image of a semiconductor chip.

[0023] In this embodiment, the semiconductor chip includes a material strip and a chip body arranged in an array on the material strip. The semiconductor chip is continuously transported, and the front image and back image of the semiconductor chip can be collected at different stations. Because of its large size, a semiconductor chip usually has multiple front images and multiple back images. Taking the detection of QFN4*4 semiconductor chips as an example, a piece of material strip has a total of 855 chip bodies. Each time a photo is taken, 25 chips in 5 rows and 5 columns can be detected, and 36 photos are required. Therefore, there will be 36 front images and 36 back images of this semiconductor chip.

[0024] S400, inputting the front image and the back image into the front detection algorithm set and the back detection algorithm set respectively, the front detection algorithm set and the back detection algorithm set respectively perform at least one defect type detection on the front image or the back image and output the detection result.

[0025] The front detection algorithm set and the back detection algorithm set each include multiple detection algorithms, each of which is used to detect at least one defect type of the semiconductor chip. The detection algorithms in the front detection algorithm set and the back detection algorithm set can detect one front image or one back image at the same time. That is, a front image can be simultaneously detected for different defect types in multiple detection algorithms in the front detection algorithm set, and a back image can be simultaneously detected for defect types in multiple detection algorithms in the back detection algorithm set, so that multiple defect detections can be performed simultaneously to improve detection efficiency.

[0026] In one embodiment, the front image and the back image are input into the front detection algorithm set and the back detection algorithm set respectively, and the front image and the back image of another semiconductor chip are obtained. The front image and the back image are placed in a queue cache, and multiple detection algorithms detect simultaneously to reduce the visual detection time. Each product can be photographed continuously without waiting for the detection results. All detection algorithms detect all chip bodies on one image at the same time, thereby improving the detection efficiency.

[0027] S500: Separate the semiconductor chips into qualified products and unqualified products according to the detection results.

[0028] A qualified product tray and a waste tray are provided. Semiconductor chips that are qualified according to the test results are placed in the qualified product tray, and semiconductor chips that are not qualified according to the test results are placed in the waste tray.

[0029] The waste tray corresponds to the defect type one by one. When the semiconductor chip is a non-conforming product and there are multiple defect types, the semiconductor chip is placed in the corresponding waste tray according to the priority of the defect type. The priority is set in advance. For example, if a semiconductor chip has both scratches and contamination, and the priority of scratches is greater than that of contamination, the semiconductor chip is placed in the waste box for collecting scratched products. Only when all test items are qualified can the semiconductor chip be qualified.

[0030] In one embodiment, the detection results also include defect identification and defect type. The defect identification will display the defect location of the semiconductor chip, marked with an identifier; the defect type will display all defect types that appear here, so as to facilitate viewing the detection details and locating the defect in the entire material strip.

[0031] In one embodiment, the detection algorithm can be trained by acquiring a front image or a back image of the semiconductor chip. In this case, there is no need to manually input standard appearance data, and manual autonomous learning can be performed. Training is generally performed for the front image, because the back of the semiconductor chip is generally unchanged, but the front changes greatly.

[0032] During the production process, when the product batch is changed, a chip body in the training frame area of ​​the current front image or back image is automatically selected as the template feature for training during the first photo inspection, and the trained template center is used as the positioning coordinate system for positioning. The subsequent detection frame positions of other detection algorithms follow the trained template center coordinate system. Users do not need to manually add detection templates or adjust detection parameters, which greatly reduces the teaching and debugging time of new materials.

[0033] When the number of defects detected by a detection algorithm for the same semiconductor chip exceeds the set threshold, an alarm is issued to remind the user to retrain the detection algorithm. At this time, there may be a situation where the first training fails or the chip body selected for training is a defective product. Therefore, when the number of abnormalities detected by the same detection algorithm of a strip exceeds the set threshold, the detection is stopped and an alarm pops up to remind manual confirmation.

[0034] Due to the large variety of products and frequent product changes, the detection method in this embodiment can shorten the teaching time to less than 5 minutes after enabling the automatic teaching and parameter automatic writing functions. At the same time, because the photo and detection are carried out at the same time, multiple defect types are detected at the same time, which greatly shortens the image processing time and quickly displays the detection results.

[0035] In one embodiment, the appearance defect types on the front side of the semiconductor chip include cutting size (PKG Size), character offset (Mark Offset), printing (Mark / Character), empty seal mark (X-Mark), and surface defect (MarkSurface); the appearance defect types on the back side include cutting size (PKG Size), cutting offset (Cut Offset), empty seal mark (X-Mark), heat sink size (PAD Size), heat sink defect (PAD Defect), pin size (Lead Size), pin defect (Lead Defect), ball detection (Ball), and surface defect (Lead Surface).

[0036] The detection algorithms include cutting size algorithm, seal mark algorithm, printing detection algorithm, surface defect algorithm, heat sink size algorithm, pin size algorithm, heat sink defect algorithm, pin defect algorithm and ball detection algorithm. Of course, if there are other types of appearance defects in the semiconductor chip, the corresponding detection algorithm can be added for detection. The actual detection algorithm can be deployed according to the actual detection needs of the semiconductor chip.

[0037] The cutting size algorithm is used for cutting size and cutting offset detection. The detection process of the cutting size algorithm specifically includes: Construct four detection frames on the front image or the back image, and use the Canny detection algorithm to fit the edge curve of each detection frame; The least square method is used to fit the four edges of the chip body: top, bottom, left, and right; Calculate whether the length and width of the chip body are within the qualified range, and determine whether there is a cutting offset by calculating the distance between the center of the product with four sides and the heat sink or pin.

[0038] The empty seal marking algorithm is used for empty seal marking detection. The detection process of the empty seal marking algorithm specifically includes: The front image or the back image is binarized using a threshold segmentation algorithm; Extract the pixel area of ​​the chip in the image and convert it into actual size based on the calibrated pixel equivalent; Determine whether the actual size exceeds the set limit value; If it exceeds, it is judged as an abnormally marked product; otherwise, it is a normal product.

[0039] The printing detection algorithm is used for printing detection, which includes detecting character breakage, misprinting, reprinting, offset, etc. The detection process of the printing detection algorithm specifically includes: Use image binarization to segment each character on the front image or the back image; Extract the outline and area of ​​each character to train a standard template; A template matching algorithm based on product outline is used to detect the similarity of characters at the same position of each product, and the character area is controlled at the same time; Determine whether the similarity and character area are within the set range; If yes, then it is qualified; otherwise, it is unqualified.

[0040] The surface defect algorithm is used for surface defect detection. The detection process of the surface defect algorithm specifically includes: Use the area subtraction algorithm to exclude characters, heat sinks, pins and other areas on the front image or back image; Use common threshold and dynamic threshold algorithms to segment defects such as scratches, contamination, and foreign matter in products; Determine whether the total area, single area, width, and length of the proposed defect area exceed the set threshold; If it exceeds, the product is judged to have surface defects, otherwise, it is a qualified product.

[0041] The heat sink size algorithm is used for heat sink size and cutting offset detection. The detection process of the heat sink size algorithm includes: The four edges of the heat sink on the back image are fitted by edge extraction algorithm and least square method; Calculate the length and width of the heat sink, and determine whether the distance between the center of the heat sink and the center of the product is within the set range; If yes, the product has no cutting offset, if no, there is cutting offset.

[0042] The pin size algorithm is used to detect the pin size. The detection process of the pin size algorithm includes: The pin area around the chip body in the back image is extracted by using a threshold segmentation algorithm; For each of the pin regions, a minimum circumscribed rotated rectangle algorithm is used to calculate the length and width of the pins and the spacing between the pins; Determine whether these dimensions are within the set range; If yes, the pin size is qualified, if no, the pin size is unqualified.

[0043] The heat sink defect algorithm is used to detect heat sink defects. The detection process of the heat sink defect algorithm specifically includes: The template matching algorithm is used to train the template for extracting the contour of the heat sink; Use the trained template to calculate similarity; Determine whether the similarity is qualified, and use the threshold segmentation algorithm to extract the heat sink area. The ordinary threshold and dynamic threshold algorithms are used to segment the area to find defects such as scratches and contamination in the product.

[0044] The pin defect algorithm is used to detect pin defects. The detection process of the pin defect algorithm specifically includes: The template matching algorithm is used to train the template for pin extraction contour; Use the trained template to calculate similarity; To judge whether the similarity is qualified, the threshold segmentation algorithm is used to extract the four rectangular areas of the pins. The ordinary threshold and dynamic threshold algorithms are used to segment the areas to find defects such as scratches and contamination in the product.

[0045] The ball detection algorithm is used to detect solder ball defects. The detection process of the ball detection algorithm specifically includes: The solder ball area on the back image is extracted by threshold segmentation algorithm; Eliminate interference from outliers through minimum area screening; The number of solder balls is calculated by calculating the number of screened areas, and the area and centroid of solder balls are extracted; Calculate the distance between each solder ball through the centroid; The radius of the solder ball is calculated using the minimum circumscribed circle algorithm; Determine whether the distance, diameter and area are within the set threshold; If yes, then it is qualified; otherwise, it is unqualified.

[0046] The various detection algorithms above are only an example of implementing different defect detections, and other algorithms may also be used to perform different types of defect detections.

[0047] Exemplarily, the semiconductor chip is a 3mm*3mm QFN, and the optical system data is set. The chip is 3mm*3mm, and the number of rows and columns of the optical device field of view is set to 7, that is, the optical device field of view can capture 7 rows and 7 columns, a total of 49 products, and the product row and column spacing is 3.3mm. The number of rows and columns in the mapping table is set to 74 and 19 respectively (the number of rows and columns in the mapping table is the number of rows and columns of the chip body of a strip), and the number of blocks is set to 1 (the number of blocks is the number of areas in which the chip body of a strip is arranged). Set standard appearance data, which can be automatically written through teaching and training. The specific parameters of the detection items are set in the detection tool of the teaching interface. Debug the detection algorithms one by one. Each detection algorithm is divided into four major steps: ROI adjustment, training parameter adjustment, advanced parameter confirmation, and testing training results. After each detection algorithm is taught, it can be automatically produced for detection. See the attached for the detection results. Figure 6 As shown, qualified and unqualified products can be displayed on the interface and distinguished by different colors.

[0048] Embodiment 2 The present invention also discloses an appearance inspection system for semiconductor chips, which adopts the appearance inspection method in the first embodiment.

[0049] See attached Figure 2 As shown, the appearance inspection system includes an optical device and a controller, wherein the optical device is used to collect the front image and the back image of the semiconductor chip, and the controller is used to process the front image and the back image.

[0050] See attached Figure 3As shown, the controller includes a storage module, a cache module, a detection module and a shunt module. The storage module stores a set of detection parameters corresponding to the semiconductor chip, and the detection parameter set includes standard appearance data and optical system data of the semiconductor chip, and the optical device is adjusted according to the optical system data. The cache module is used to store the front image and the back image. The detection module stores a front detection algorithm set and a back detection algorithm set, and the front detection algorithm set and the back detection algorithm set respectively perform at least one defect type detection on the front image or the back image and output the detection result. The shunt module shunts the semiconductor chip into qualified products and unqualified products according to the shunt instruction below the detection result.

[0051] The optical device comprises a front optical collection mechanism and a back optical collection mechanism, and the front optical collection mechanism and the back optical collection mechanism are used for collecting a front image and a back image respectively.

[0052] See attached Figure 4 As shown, the front optical collection mechanism includes a coaxial first camera 11, a first annular light source 12 and a second annular light source 13. The first annular light source and the second annular light source are located on one side close to the lens of the first camera. The first annular light source and the second annular light source are shadowless light and polarized light respectively. Figure 5 As shown, the back optical collection mechanism includes a second camera 21, a third annular light source 22 and a fourth coaxial light source 23. The fourth coaxial light source is close to the lens of the second camera, and the third annular light is coaxially arranged with the second camera.

[0053] The first annular light source and the second annular light source use white shadowless light + high-brightness red polarized light, which can clearly present characters or patterns, and the defects of the plastic package are clearly visible, and the grayscale difference between the characters, patterns and defects and the plastic powder is greater than 30. The third annular light and the fourth coaxial light use low-angle white annular light + high-definition red coaxial light, which can clearly present features such as heat sinks, pins and solder balls, and the defects of the plastic package are clearly visible, and the grayscale difference between the heat sinks, pins, solder balls and defects and the plastic package is greater than 30. The optical device in this embodiment can clearly distinguish the various characteristic details of the characters, pins and the surface of the plastic package.

[0054] See attached Figure 2 As shown, the appearance inspection system also includes a display device, which is connected to the controller and is used to display the inspection results.

[0055] It should be understood that the disclosed system can be implemented in other ways. For example, the division of the modules described above is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The above modules may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the embodiment of the present invention.

[0056] Embodiment 3 The present invention further discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the appearance detection method in the first embodiment are implemented.

[0057] In this embodiment, if the detection method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0058] The above implementation modes are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with this technology to understand the content of the present invention and implement it, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting the appearance of a semiconductor chip, characterized in that: include: Creating a detection parameter collection corresponding to the semiconductor chip, wherein the detection parameter collection includes standard appearance data and optical system data of the semiconductor chip; adjusting an optical device according to the optical system data; Respectively acquiring a front side image and a back side image of the semiconductor chip; Inputting the front image and the back image into a front detection algorithm set and a back detection algorithm set respectively, wherein the front detection algorithm set and the back detection algorithm set respectively perform at least one defect type detection on the front image or the back image and output the detection result; The semiconductor chips are sorted into qualified products and unqualified products according to the detection results.

2. The appearance inspection method of a semiconductor chip according to claim 1, characterized in that: The standard appearance data includes the standard size of the product, the number of products to be inspected in the field of view, printing, and defect area, and the standard appearance data is manually input or automatically written through a teaching training of the semiconductor chip; The optical system parameters include camera exposure time and power brightness, and the optical system parameters are input manually.

3. The appearance inspection method of a semiconductor chip according to claim 1, characterized in that: The front detection algorithm set and the back detection algorithm set each include a plurality of detection algorithms, each of the detection algorithms is used for detecting at least one defect type of the semiconductor chip; The detection algorithms in the front detection algorithm set and the back detection algorithm set can detect one front image or one back image at the same time.

4. The appearance inspection method of a semiconductor chip according to claim 3, characterized in that: The detection algorithm can be trained by acquiring a front side image or a back side image of the semiconductor chip.

5. The appearance inspection method of a semiconductor chip according to claim 4, characterized in that: When the number of defects generated by a detection algorithm on the same semiconductor chip exceeds a set threshold, an alarm is issued to remind the user to retrain the detection algorithm.

6. The appearance inspection method of a semiconductor chip according to claim 1, characterized in that: While inputting the front side image and the back side image into the front side detection algorithm set and the back side detection algorithm set respectively, a front side image and a back side image of another semiconductor chip are acquired.

7. The appearance inspection method of a semiconductor chip according to claim 1, characterized in that: The step of separating the semiconductor chips into qualified products and unqualified products according to the detection results specifically includes: When the semiconductor chip is a non-conforming product and there are a plurality of defect types, the semiconductor chip is placed in a corresponding waste tray according to the priority of the defect types.

8. Appearance inspection system for semiconductor chips, characterized by: The appearance inspection method according to any one of claims 1 to 7 is adopted, wherein the appearance inspection system comprises an optical device and a controller, wherein the optical device is used to collect a front image and a back image of the semiconductor chip, and the controller comprises: A storage module, wherein a detection parameter set corresponding to the semiconductor chip is stored in the storage module, the detection parameter set includes standard appearance data and optical system data of the semiconductor chip, and the optical device is adjusted according to the optical system data; A cache module, the cache module is used to store the front image and the back image; A detection module, wherein a front detection algorithm set and a back detection algorithm set are stored in the detection module, wherein the front detection algorithm set and the back detection algorithm set respectively detect at least one defect type on the front image or the back image and output a detection result; A shunting module is used to shun the semiconductor chips into qualified products and unqualified products according to the shunting instructions below the detection results.

9. The appearance inspection system for semiconductor chips according to claim 8, characterized in that: The optical device comprises a front optical collection mechanism and a back optical collection mechanism, wherein the front optical collection mechanism and the back optical collection mechanism are used to collect a front image and a back image respectively; The front optical collection mechanism comprises a coaxial first camera, a first annular light source and a second annular light source, wherein the first annular light source and the second annular light source are located on a side close to the lens of the first camera, and the first annular light source and the second annular light source are shadowless light and polarized light respectively; The back optical collection mechanism includes a second camera, a third annular light source and a fourth coaxial light source. The fourth coaxial light source is closely attached to the lens of the second camera, and the third annular light is coaxially arranged with the second camera.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the appearance detection method according to any one of claims 1 to 7 are implemented.

Images