Crystal ingot defect detection system and method

Abstract

The application provides a crystal ingot defect detection system and method, and relates to defect detection technology; the system comprises: a multi-directional illumination module configured to provide illumination light from at least two different directions to a surface of a crystal ingot to be detected, so that the flat area and the inclined area of the crystal ingot can be effectively illuminated during detection; an image acquisition module having an acquisition light path facing the surface to be detected, configured to acquire optical signals modulated by the surface of the crystal ingot and entering the acquisition light path under the illumination of the multi-directional illumination module, and convert the optical signals into image data; and a processing module configured to receive and process the image data to identify and output defect information of the surface of the crystal ingot. Through the application, the accuracy of defect detection of the surface of the crystal ingot can be effectively improved.

Description

Technical Field

[0001] This application relates to defect detection technology, and more particularly to a crystal ingot defect detection system and method. Background Technology

[0002] Silicon carbide, as a new generation of semiconductor material, is widely used in power electronics, 5G communications and other fields due to its excellent properties. Its ingot surface defects directly affect wafer yield and device performance, making ingot defect detection a core quality control link in the industry chain.

[0003] Currently, defect detection in silicon carbide ingots mainly relies on optical imaging and surface morphology analysis techniques. Mainstream methods include phase-deflection detection, spectral confocal detection, and bright-field reflection imaging. Phase-deflection detection projects coded stripe light onto the ingot surface, collects the phase information of the reflected light, and reconstructs the three-dimensional morphology using algorithms. Spectral confocal detection, based on the principle of dispersive optics, utilizes the wavelength-height correspondence after focusing a broadband light source to achieve high-precision imaging. Bright-field reflection imaging illuminates the ingot surface with light at a fixed angle and identifies defects by imaging the reflected light.

[0004] However, the above-mentioned schemes suffer from low detection accuracy because they cannot be adapted to the complex surface characteristics of silicon carbide ingots. Summary of the Invention

[0005] This application provides a crystal ingot defect detection system and method to improve detection accuracy.

[0006] In a first aspect, this application provides a crystal ingot defect detection system, the system comprising:

[0007] A multi-directional illumination module is configured to provide illumination light from at least two different directions to the surface of the ingot to be inspected, so that both the flat and tilted areas of the ingot can be effectively illuminated during the inspection process.

[0008] The image acquisition module has an acquisition optical path facing the surface to be detected, and is configured to acquire optical signals that are modulated by the surface of the ingot and enter the acquisition optical path under the illumination of the multi-directional illumination module, and convert the optical signals into image data;

[0009] The processing module is configured to receive and process the image data to identify and output defect information on the surface of the ingot.

[0010] In one possible implementation, the processing module is further configured to: when a warped region is identified on the surface of the ingot, control the multi-directional illumination module to provide illumination light suitable for illuminating the warped region.

[0011] In one possible implementation, the system further includes a first motion module, wherein the processing module is specifically configured to:

[0012] The first motion module is controlled to drive the crystal ingot to perform step-by-step rotation or translation.

[0013] At each step position, the image acquisition module is controlled to acquire one or more frames of local surface images corresponding to the current field of view;

[0014] The images of the local surfaces obtained after scanning the entire surface of the crystal ingot are stitched together to generate complete image data of the crystal ingot surface.

[0015] In one possible implementation, the multi-directional lighting module includes at least two independent lighting units, which are fixedly arranged and configured to provide illumination to the ingot surface from different spatial positions and lighting angles.

[0016] or,

[0017] The multi-directional lighting module includes a lighting unit and a second motion module;

[0018] The second motion module is configured to change the relative position of the illumination unit and the crystal ingot surface, so that the illumination unit can illuminate the crystal ingot surface from different directions.

[0019] In one possible implementation, the at least two independent lighting units include at least one first lighting unit and at least one second lighting unit;

[0020] The first lighting unit is configured to provide illumination from the outer edge of the ingot towards the center of the ingot, so as to primarily illuminate the inclined or flat edge area of ​​the ingot through different lighting angles;

[0021] The second lighting unit is configured to provide illumination from the central region of the ingot toward the edge of the ingot, so as to primarily illuminate the warped edge region of the ingot.

[0022] In one possible implementation, the multi-directional illumination module includes a laser source, a beam shaping unit, a beam splitting unit, and at least two beam reflecting units;

[0023] The beam shaping unit is configured to modulate the laser into a uniform spot of a preset size and shape, and the beam splitting unit is configured to split the uniform spot into at least two sub-beams whose on / off state or intensity can be independently controlled.

[0024] The beam reflecting unit is configured to adjust the propagation direction of each of the sub-beams, so that at least two sub-beams are emitted from different spatial directions;

[0025] Accordingly, the processing module is also configured to, during the detection process, independently control the opening and closing or intensity combination of the at least two sub-beams according to the morphological characteristics of the current scanning area, so as to adaptively illuminate the flat or tilted areas of the ingot.

[0026] In one possible implementation, the processing module is further configured to:

[0027] Based on the received preset detection mode command, and / or based on the defect features obtained from real-time analysis of the image data, the current defect type is determined;

[0028] Based on the current defect type and the pre-stored correspondence between defect types and illumination wavelengths, the multi-directional illumination module is controlled to provide illumination at a specific wavelength.

[0029] In one possible implementation, the multi-directional illumination module further includes a light homogenizing unit disposed on the propagation path of the illumination light for homogenizing the illumination light.

[0030] In one possible implementation, the processing module is specifically configured as follows:

[0031] Extract feature information related to surface defects from the complete image data;

[0032] The feature information is input into a preset defect classification model or matched with a preset defect feature library to determine the defect information; wherein the defect classification model is optimized based on a defect image training set, and the defect information includes at least one of defect type, defect location, and defect size.

[0033] Secondly, this application provides a method for detecting defects in crystal ingots, the method comprising:

[0034] Illumination light from at least two different directions is provided to the surface of the ingot to be inspected so that both the flat and tilted areas of the ingot can be effectively illuminated during the inspection process.

[0035] The optical signal modulated on the illuminated surface of the ingot is acquired, and the optical signal is converted into image data;

[0036] The image data is received and processed to identify and output defect information on the surface of the ingot.

[0037] Thirdly, this application provides a crystal ingot defect detection device, the device comprising:

[0038] The lighting control module is used to provide illumination light from at least two different directions to the surface of the ingot to be inspected, so that both the flat and tilted areas of the ingot can be effectively illuminated during the inspection process.

[0039] The acquisition and control module is used to acquire the modulated optical signal on the illuminated surface of the ingot and convert the optical signal into image data;

[0040] The defect identification module is used to receive and process the image data to identify and output defect information on the surface of the ingot.

[0041] Fourthly, this application provides an electronic device, including at least one processor and a memory communicatively connected to the processor;

[0042] The memory stores computer-executed instructions;

[0043] The processor executes computer execution instructions stored in the memory to implement the method as described in the second aspect.

[0044] Fifthly, this application provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, are used to implement the method described in the second aspect.

[0045] In a sixth aspect, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method described in the second aspect.

[0046] This application provides a crystal ingot defect detection system and method. The crystal ingot defect detection system of this application provides illumination light from at least two different directions to the surface of the crystal ingot to be inspected through a multi-directional illumination module. An image acquisition module acquires the optical signals modulated by the crystal ingot surface and enters the acquisition optical path, converting them into image data. A processing module receives and processes the image data to identify and output defect information. Based on this system, multi-directional illumination can specifically avoid specular reflection interference from highly reflective surfaces of the crystal ingot through differentiated angle layouts, while simultaneously covering illumination blind spots in areas with large undulations and steeply angled edges. Combined with uniform light processing, it further improves illumination uniformity, enabling clear signal responses from defects of different locations and types, effectively solving the problem of poor adaptability of fixed illumination methods. Furthermore, the improved defect signal contrast brought about by multi-directional illumination can effectively reduce the risk of missed or false detections, ultimately achieving a significant improvement in detection accuracy. Attached Figure Description

[0047] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0048] Figure 1 This is a schematic diagram illustrating an application scenario of an ingot defect detection system provided in an embodiment of this application;

[0049] Figure 2 A schematic diagram of the structure of an ingot defect detection system provided in this application embodiment. Figure 1 ;

[0050] Figure 3 A schematic diagram of the structure of an ingot defect detection system provided in this application embodiment. Figure 2 ;

[0051] Figure 4A A structural example of an ingot defect detection system provided in this application embodiment. Figure 1 ;

[0052] Figure 4B A structural example of an ingot defect detection system provided in this application embodiment. Figure 2 ;

[0053] Figure 4C A structural example of an ingot defect detection system provided in this application embodiment. Figure 3 ;

[0054] Figure 5 This is a schematic diagram of a crystal ingot defect detection system provided in an embodiment of this application;

[0055] Figure 6 Figure 4 shows a structural example of an ingot defect detection system provided in this application embodiment;

[0056] Figure 7 A schematic flowchart of a crystal ingot defect detection method provided in an embodiment of this application;

[0057] Figure 8 This is a schematic diagram of the structure of an ingot defect detection device provided in an embodiment of this application;

[0058] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0059] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0060] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0061] Silicon carbide (SiC), as a new generation of semiconductor material, has become a core supporting material in fields such as power electronic devices, 5G communication, electric vehicles and renewable energy systems due to its excellent properties such as high thermal conductivity, high breakdown electric field and high electron mobility.

[0062] Silicon carbide ingots are a key raw material for the fabrication of silicon carbide wafers. The density of surface defects (such as scratches, cracks, micropipes, stacking faults, and dislocations) directly determines the wafer yield and the performance reliability of end devices. Because ingots need to be grown under high temperature and high pressure at approximately 2000°C, and the material itself has high hardness and high chemical stability, various microscopic and macroscopic defects are easily generated during the growth process. These defects may cause short circuits, leakage currents, or mechanical fractures in devices during subsequent processing. Therefore, ingot defect detection has become an indispensable core link in the quality control of the silicon carbide industry chain.

[0063] Currently, defect detection in silicon carbide ingots mainly relies on optical imaging and surface morphology analysis techniques. Mainstream methods include phase-deflection detection, spectral confocal detection, and bright-field reflection imaging. Phase-deflection detection projects coded stripe light onto the ingot surface, collects the phase information of the reflected light, and reconstructs the three-dimensional morphology using algorithms. Spectral confocal detection, based on the principle of dispersive optics, utilizes the wavelength-height correspondence after focusing a broadband light source to achieve high-precision imaging. Bright-field reflection imaging illuminates the ingot surface with light at a fixed angle and identifies defects by imaging the reflected light.

[0064] However, the above-mentioned scheme cannot adapt to the complex surface characteristics of silicon carbide ingots, such as high reflectivity, large undulations, and large edge tilt angles, resulting in insufficient contrast of defect signals and easy omission of edge areas, thus resulting in low detection accuracy.

[0065] Therefore, embodiments of this application provide a crystal ingot defect detection system and method to solve the above-mentioned problems. Specifically, the system of this application proposes to set up a multi-directional illumination module, an image acquisition module, and a processing module. The multi-directional illumination module is configured to provide illumination light from at least two different directions to the surface of the crystal ingot to be inspected. The image acquisition module has an acquisition optical path facing the surface to be inspected, and is used to acquire optical signals that have been modulated by the crystal ingot surface and enter the acquisition optical path and convert them into image data. The processing module receives and processes the image data to identify and output defect information.

[0066] It is understood that the ingot defect detection system of this application is applicable to quality inspection scenarios in any silicon carbide ingot mass production line. For example, Figure 1 This is a schematic diagram illustrating an application scenario of an ingot defect detection system provided in an embodiment of this application, such as... Figure 1 As shown, the ingot defect detection system of this application can be used in the ingot pretreatment detection scenario of high-power silicon carbide wafers for electric vehicles, and can be used to achieve accurate detection of scratches on the surface of silicon carbide ingots, cracks in the edge warping area and microtube defects.

[0067] Currently, in this scenario, silicon carbide ingots have high surface reflectivity, edge tilt angles of over 30°, and surface undulations exceeding 1mm. Traditional detection methods are prone to missing detections in the edge warping area and have weak signal capture capabilities for micro-defects such as microtubes, directly affecting the yield of subsequent wafer dicing.

[0068] When the crystal ingot defect detection system of this application is used for detection, the multi-directional illumination module provides illumination light to different directions such as the outer edge of the crystal ingot and the center to the edge. The illumination light facing the outer edge is adapted to the large tilt angle area at a low angle, and the illumination light facing the center to the edge covers the flat area and the inner surface of the warped edge. The image acquisition module collects the defect scattering signal under each illumination direction and converts it into image data. The processing module analyzes the image data to output defect information.

[0069] In the above process, the multi-directional illumination module avoids the interference of specular reflection from highly reflective surfaces through differentiated angle layout, making the scattered signals of the defect area clearly presented. At the same time, it fully covers the illumination blind spots of large undulation surfaces and large edge tilt areas, solving the problem of missed detection in traditional solutions. Multi-directional illumination enhances the signal contrast of different defects, and together with the image acquisition and processing module, it effectively improves the identification accuracy of defects such as microtubes and cracks.

[0070] It should be understood that the number of illumination directions provided by the multi-directional illumination module can be adjusted to two or more according to the ingot inspection requirements, and the illumination directions can be achieved through a preset fixed layout; the image acquisition module can use either an area scan camera or a line scan camera, both of which can meet the requirements for acquiring optical signals modulated by the ingot surface; the processing module can analyze the image data using conventional methods of basic feature extraction and comparison, and this embodiment does not limit this. Furthermore, the application scenarios of this system can also include the inspection of silicon carbide ingots for 5G communication devices, the inspection of silicon carbide ingots for renewable energy systems, and any other scenario requiring defect detection of silicon carbide ingots with high reflectivity, large undulations, and large edge tilt angles; this embodiment does not limit this.

[0071] The following detailed description, with reference to the accompanying drawings, outlines some embodiments of the ingot defect detection system of this application. Where the embodiments do not conflict, the following embodiments and features thereof can be combined with each other.

[0072] This application provides a crystal ingot defect detection system. Figure 2 A schematic diagram of the structure of an ingot defect detection system provided in this application embodiment. Figure 1 ,like Figure 2 As shown, the ingot defect detection system of this application embodiment includes a multi-directional illumination module, an image acquisition module, and a processing module.

[0073] Specifically, in this embodiment, the multi-directional illumination module is configured to provide illumination light from at least two different directions to the surface of the ingot to be inspected, so that both the flat and tilted areas of the ingot can be effectively illuminated during the inspection process.

[0074] The illumination light can be LED light, laser light, etc., and short-wavelength light such as blue light or violet light can be selected to enhance the defect scattering signal according to the detection requirements. Furthermore, it should be understood that a flat area refers to the internal area of ​​the ingot surface with a relatively gentle morphology, without obvious undulations or tilt angles; a tilted area refers to the edge of the ingot and areas near the edge with a larger surface tilt angle. Effective illumination means that the illumination light can cover the target area, so that the defect produces a clear scattering signal, avoiding signal loss or overexposure due to improper illumination angle, and providing a sufficient signal basis for defect identification.

[0075] Specifically, in this embodiment, the image acquisition module has an acquisition optical path facing the surface to be detected, and the image acquisition module is configured to acquire optical signals that are modulated by the crystal surface and enter the acquisition optical path under the illumination of the multi-directional illumination module, and convert the optical signals into image data.

[0076] The acquisition optical path is the optical transmission path between the image acquisition module and the surface of the ingot to be inspected. In this embodiment, the optical transmission path is equipped with a high depth-of-field, low-distortion telecentric lens to adapt to the undulation characteristics of the ingot surface >1mm, ensuring that defects at different locations can be clearly imaged. The selection of the telecentric lens needs to balance system structure and resolution: too long a working distance will result in an excessively large lens size, which is prone to structural interference and is not conducive to system integration; too short a working distance will not meet the inspection field of view requirements and may interfere with the illumination source. Therefore, for a field of view of 20-30mm, a high-field telecentric lens is selected as the objective lens, with a working distance of about 120-150mm, a depth of field of 2-3mm, and a resolution of about 15-20μm. This combination of parameters can balance field of view coverage, imaging clarity, and system integration, adapting to the undulation characteristics of the ingot surface >1mm, obtaining clear images without high-precision motion control, and avoiding structural interference caused by an excessively large lens size.

[0077] It should be understood that the optical signal modulated by the crystal ingot surface and entering the acquisition optical path refers to the light signal carrying surface morphology and defect information formed after the illumination light shines on the crystal ingot surface and is scattered by defect areas or reflected by flat areas. Converting the optical signal into image data means that the detector in the image acquisition module converts the light signal into digital image information that can be recognized by the processing module.

[0078] Specifically, the processing module in this embodiment is configured to receive and process image data to identify and output defect information on the surface of the ingot. This defect information includes key parameters such as the type of defect (e.g., scratches, cracks, microtubules, stacking faults, dislocations, etc.), its location, and size, providing a basis for ingot quality assessment.

[0079] In this embodiment, the processing module undertakes the core control function, which is responsible for controlling the multi-directional illumination module to provide illumination light from at least two different directions, and at the same time controlling the image acquisition module to perform image acquisition operations.

[0080] In this embodiment, at least two different directions of illumination light are not turned on at the same time. For example, the processing module first controls the multi-directional illumination module to turn on the illumination light in the first direction, and after the image acquisition module completes the image acquisition of the corresponding area, it switches to the illumination light in the second direction and synchronously controls the image acquisition module to acquire the image data under the illumination direction, thereby completing the illumination and acquisition of different directions of the entire surface in sequence.

[0081] More specifically, in this embodiment, the illumination light of the multi-directional illumination module is specifically LED strip light; the image acquisition module is specifically an area array camera or line array camera equipped with a telecentric objective lens, which has high resolution and low distortion characteristics and can accurately capture defect image information under different illumination directions.

[0082] It should be understood that the multi-directional illumination module and the image acquisition module can also achieve illumination control and acquisition control respectively through other control modules independent of the processing module; at the same time, the illumination light of the multi-directional illumination module can also be other types of visible light, and the image acquisition module can also be other photodetectors such as CCD. This embodiment does not limit these aspects.

[0083] In addition, in practical applications, at least two different directions of illumination can be turned on at the same time. As long as the illumination angles and positions are arranged reasonably to ensure that the defect signals generated by the illumination from different directions in the image acquisition module interfere with each other within a preset range and can be clearly distinguished, this embodiment does not limit this.

[0084] In this embodiment, the multi-directional illumination module provides the image acquisition module with effective illumination from multiple directions, adaptable to the complex surface of the crystal ingot. Based on this illumination condition, the image acquisition module and the processing module can work together to identify defects by utilizing the difference in the modulation characteristics of incident light between defects and smooth areas on the crystal ingot surface.

[0085] Specifically, the appropriate imaging mode can be flexibly selected based on the surface characteristics of the ingot under test (such as reflectivity and roughness) and the type of defect to be inspected. For example, in one embodiment, a bright-field reflection imaging mode can be used, which collects the light signal after the illumination light is reflected from the surface of the ingot and uses the reflectivity or morphological changes caused by defects to achieve identification; in another embodiment, a dark-field scattering imaging mode can be used, which, by configuring appropriate filters or polarizing elements, suppresses specular reflection light from flat areas and mainly collects the scattered light signal caused by defects (such as cracks and scratches), thereby significantly improving the contrast of micro-defects.

[0086] It is understood that, regardless of the imaging mode mentioned above or other feasible optical imaging modes used, their effective implementation depends on the adaptive illumination provided by the multi-directional illumination module, which can cover the entire surface of the ingot (including flat and tilted areas). The scope of protection of this application is not limited to the selection of a specific imaging mode.

[0087] As a further design feature, the processing module is also configured to control the multi-directional lighting module to provide illumination suitable for illuminating the warped area when a warped area is identified on the surface of the ingot.

[0088] Specifically, the warped area refers to the area on the surface near the edge of the crystal ingot that bulges from the direction closer to the center of the crystal ingot towards the direction away from the center, with the height gradually increasing. Its core characteristic is that the surface is bulging and the inner surface forms a hidden space, in which defects are easily hidden. On the other hand, the tilted area refers to the area on the edge of the crystal ingot and the nearby surface that gradually decreases in height from the direction closer to the center of the crystal ingot towards the direction away from the center, without obvious bulging. It only presents a simple tilted shape and has no hidden space.

[0089] In this embodiment, the processing module analyzes the initial image data acquired by the image acquisition module, and comprehensively identifies the location and range of the warped edge area based on the surface height change trend (the height increases against the trend near the edge), the distribution of hidden shadow areas in the image, and the surface tilt angle calculation results.

[0090] More specifically, the processing module first performs grayscale analysis on the initial image data. The inner surface of the warped edge region will form a significant low-grayscale shadow area because light cannot directly illuminate it, which is significantly different from the uniform grayscale distribution of the edge tilted region. At the same time, by converting the physical coordinates corresponding to the image pixels, the surface height value at different positions on the edge of the ingot is calculated. When the height value of multiple consecutive pixels shows a trend of gradually increasing from the center to the edge, and the difference in height change exceeds a preset threshold (e.g., greater than 0.5mm), it is initially determined to be a candidate region for warped edge. Then, the tilt angle algorithm is used to calculate the surface tilt angle of the candidate region. If the tilt angle value is between 20° and 50° (adapted to the typical tilt angle range of the warped edge region), and the tilt angle difference with the adjacent tilted region is greater than 15°, then the region is finally confirmed as a warped edge region, and its start and end coordinates are locked to determine the coverage area.

[0091] Furthermore, after the processing module identifies the warped edge area, it sends a targeted control signal to the multi-directional lighting module to adjust the position and angle of the lighting module so that the lighting light can accurately illuminate the hidden space on the inner surface of the warped edge area, ensuring that the defect in this area can generate a clear scattering signal. Then, the image acquisition module synchronously acquires the corresponding image data and transmits it to the processing module. The processing module combines the image data under this lighting direction to complete the identification of the defect in the warped edge area and output the information.

[0092] It should be understood that in practical applications, the warped edge area can also be identified by presetting the standard position and size range of the warped edge area corresponding to the ingot model, or by using a laser ranging module to detect the trajectory of the height change on the ingot surface in real time. This embodiment does not limit this.

[0093] In this embodiment, by adding illumination light to illuminate the warped edge area, the hidden space of the warped edge area can be accurately covered, solving the problem of missed defects caused by the inability of the illumination light to reach the hidden space in the traditional detection scheme. At the same time, the targeted illumination direction can enhance the signal contrast between the defects in the warped edge area and the background, making the micro-cracks, micro-tubes and other minute defects hidden on the inner surface clearly visible, further improving the coverage detection capability for complex surface defects of the ingot, making the detection range more comprehensive and the detection results more accurate.

[0094] The ingot defect detection system provided in this embodiment addresses the high reflectivity of ingots by employing a multi-directional illumination module that provides illumination from at least two different directions. This differentiated angle arrangement avoids overexposure or signal loss caused by single-fixed-angle illumination, ensuring clear capture of scattered signals from defect areas. Furthermore, considering the large surface undulations and steep edges of the ingot, the illumination from different directions can cover flat and inclined areas respectively, eliminating blind spots associated with single illumination and ensuring effective illumination of defects in easily missed areas such as inclined surfaces and edges. Simultaneously, the response characteristics of different directional illumination to different types of defects enhance the signal contrast between defects and the background, providing a sufficient foundation for the image acquisition module to accurately acquire the light signal modulated by the ingot surface. After the image acquisition module converts the light signal into image data, the processing module can accurately identify defect information by analyzing the image data.

[0095] Therefore, the system of this application can accurately adapt to the complex surface characteristics of crystal ingots, effectively reduce the risk of missed detection and false detection, and achieve simultaneous improvement in detection accuracy and efficiency without relying on multi-lens switching or complex optical path reconstruction. Moreover, the interference between illuminations from different directions is small, further ensuring the stability and reliability of defect identification.

[0096] As a preferred example, the processing module is further configured to: determine the current defect type based on the received preset detection mode command and / or based on the defect features obtained from real-time analysis of image data; and control the multi-directional illumination module to provide illumination at a specific wavelength according to the current defect type and the pre-stored correspondence between defect types and illumination wavelengths.

[0097] Specifically, the preset detection mode command is a control command that the user sets in advance based on the production process of the crystal ingot to be tested, the known defect distribution pattern, or specific detection requirements, and sends it to the processing module. The command clearly includes relevant identifiers of the target defect type, such as specific mode identifiers for microtube-specific testing and comprehensive defect screening.

[0098] In one implementation, when the processing module determines the current defect type solely through this instruction, after receiving the instruction, the processing module first parses it, extracts the target defect type identifier, and then directly determines the defect type corresponding to the identifier as the current defect type to be detected. For example, when receiving a scratch detection mode instruction, the current defect type is directly determined to be scratch. When receiving a stacking fault and dislocation joint detection instruction, it is determined that both stacking fault and dislocation defect types need to be detected simultaneously.

[0099] Specifically, in another implementation, when determining the current defect type based solely on real-time analysis of image data, the processing module first performs preprocessing operations such as noise reduction, contrast enhancement, and edge extraction on the initial image data transmitted by the image acquisition module. Then, it extracts key information such as geometric parameters, grayscale distribution characteristics, and scattering signal intensity of the suspected defect area through a preset algorithm. This information includes the length, width, shape, and area of ​​the defect, as well as the grayscale difference and grayscale uniformity between the defect area and the background.

[0100] The processing module then compares the extracted features with the internally stored standard feature library of various defects. When the feature of the suspected defect area meets the preset requirements for the overlap of the standard feature of a certain type of defect, the current defect type can be determined to be that type of defect. For example, if the extracted feature is round or elliptical, has a small diameter and high scattering signal intensity, and is highly consistent with the feature of microtubes in the standard feature library, the current defect type can be determined to be microtubes.

[0101] Specifically, in another implementation, when determining the current defect type by combining the two mentioned above, the processing module first receives and parses the preset detection mode instruction, obtains the target defect type range contained in the instruction, and uses this as the initial judgment basis. For example, the initial range corresponding to the edge defect detection instruction is common edge defects such as cracks and chipped edges.

[0102] The processing module then preprocesses and extracts defect features from the real-time acquired image data to obtain the specific features of suspected defect areas. These features are then precisely compared only within the defect standard feature library within the initial judgment range. If the overlap after comparison reaches a preset threshold, the defect type corresponding to the match is directly determined as the current defect type. If the overlap does not meet the threshold, the processing module expands the comparison range to the full defect standard feature library for re-comparison, while marking the suspected defect area as pending confirmation and retaining the relevant image data. For example, if the preset instruction is edge defect detection, and the extracted features highly match the standard features of cracks, the current defect type is determined to be a crack. If it does not match any defect features within the initial range, but matches the stacking fault features after expanding the range, it is determined to be a stacking fault defect.

[0103] Furthermore, the processing module has pre-stored a correspondence between defect types and illumination wavelengths based on the scattering response characteristics of different defects to different wavelengths of light. For example, microscopic defects such as microtubes and stacking faults are adapted to blue light or violet light, obvious defects such as scratches and cracks are adapted to blue light or visible light, and dislocations and triangular defects are adapted to violet light.

[0104] After determining the current defect type, the processing module retrieves the corresponding relationship to obtain the optimal illumination wavelength, and then sends a control signal containing the target wavelength parameter and switching instructions to the multi-directional illumination module. Upon receiving the signal, the multi-directional illumination module activates the corresponding wavelength light source and deactivates other wavelength light sources through an internal mechanism that switches between different wavelength LED beads or filters, thereby providing illumination at a specific wavelength. If multiple defect types need to be detected, the module prioritizes selecting a common optimal wavelength suitable for all target defects, or alternately switches the wavelengths corresponding to each defect according to a preset timing sequence for illumination acquisition.

[0105] It should be understood that, in addition to wired transmission, the preset detection mode command can also be sent via wireless communication methods such as Bluetooth and Wi-Fi. The defect feature extraction can use deep learning algorithms to replace traditional algorithms. The threshold for feature matching can be dynamically adjusted according to image clarity and lighting intensity. The wavelength adjustment of the multi-directional lighting module can also be fine-tuned by adjusting the light source driving current. This embodiment does not limit this.

[0106] In addition, in practical applications, the wavelength can be adjusted and optimized in real time. That is, the processing module controls the multi-directional illumination module to switch different candidate wavelengths in sequence, and controls the image acquisition module to acquire defect images at each wavelength. By comparing the scattering signal intensity of the defect area and the contrast between the defect and the background at different wavelengths, the optimal wavelength can be selected. Alternatively, the wavelength parameters can be determined by receiving the wavelength parameters manually input by the user based on experience or on-site detection results. This embodiment does not limit this method.

[0107] In this embodiment, the above process enables precise matching of illumination wavelength and defect type. Different defects exhibit varying scattering responses to different wavelengths of light. Targeted wavelength selection maximizes defect signal intensity, addressing the problem of poor defect identification performance with traditional fixed-wavelength illumination. Simultaneously, it supports multiple defect type determination methods, adapting to different scenarios such as batch detection of known defect types and comprehensive detection of unknown defect types, thus meeting diverse detection needs.

[0108] In addition, precise wavelength matching can effectively improve the distinction between defect areas and background, reduce the problem of defect signal blurring and loss caused by wavelength mismatch, and reduce the risk of false detection and missed detection. The preset detection mode can quickly start targeted detection, and real-time image analysis does not require manual input of a large number of parameters, thus balancing detection efficiency and versatility.

[0109] As a preferred example, the multi-directional lighting module also includes a light homogenizing unit, which is disposed on the propagation path of the lighting light and is used to homogenize the lighting light.

[0110] Specifically, the type of light-diffusing unit can be a conventional and efficient light-diffusing device such as a diffuser plate or frosted glass. These devices can eliminate the problem of uneven brightness of the illumination light through the scattering and refraction of light.

[0111] More specifically, the independent illumination unit can use an LED strip light source, whose light-emitting surface is homogenized by a light-diffusing unit (such as a diffuser plate or frosted glass), which can improve the uniformity of illumination and reduce the problem of large differences in brightness at different positions of the defect image due to uneven illumination, thereby improving the contrast of defect information; at the same time, it supports the output of short-wavelength light such as blue light, violet light, and even ultraviolet light. Short-wavelength light can enhance the defect scattering signal, and with the multi-directional illumination layout, it can further improve the resolution of defect detection.

[0112] It should be understood that the homogenizing unit needs to be connected in series in the light propagation path between the illumination source and the surface of the ingot to be inspected. Specifically, it can be integrated into the light-emitting end of the illumination source or set independently in the optical path between the illumination source and the ingot, ensuring that each beam of illumination light emitted from the illumination source is homogenized before illuminating the surface of the ingot. For example, in a system containing multiple illumination sources, each illumination source's light-emitting port is equipped with a corresponding homogenizing unit, so that illumination light from different directions, after homogenization, illuminates the flat area, inclined area, or warped edge area of ​​the ingot respectively.

[0113] In this embodiment, by setting up a uniform illumination unit, the uniformity of the illumination light can be effectively improved, avoiding inconsistent brightness of defect images at different locations on the crystal ingot surface due to differences in brightness of the original illumination light. This solves the problem of large differences in brightness in different areas of the defect image. When the uniform illumination light shines on the crystal ingot surface, both visible defects in flat areas and hidden defects in tilted or warped areas can obtain stable illumination conditions, ensuring uniform intensity of the scattered signal generated by the defects and reducing the loss or blurring of defect signals caused by local overexposure or underexposure. At the same time, uniform illumination light can also improve the contrast between the defect area and the background, providing a stable optical foundation for the image acquisition module to accurately acquire signals and the processing module to accurately identify defects. Combined with the differentiated layout of multi-directional illumination, the detection accuracy of the system for various defects is further improved. Moreover, the uniform illumination unit has a simple structure and is easy to integrate, without increasing the system complexity, thus balancing detection effect and practicality.

[0114] As a further detailed description of the ingot defect detection system of this application, Figure 3 A schematic diagram of the structure of an ingot defect detection system provided in this application embodiment. Figure 2 .like Figure 3 As shown, the ingot defect detection system in this embodiment also includes a first motion module.

[0115] Based on this, the processing module is specifically configured to: control the first motion module to drive the crystal ingot to perform step rotation or translation; at each step position, control the image acquisition module to acquire one or more frames of local surface images corresponding to the current field of view; and stitch together the local surface images obtained after scanning the entire surface of the crystal ingot to generate complete image data of the crystal ingot surface.

[0116] Specifically, in this embodiment, the first motion module comprises a drive motor (such as a servo motor) and a position sensor. The first motion module is used to drive the motion platform to move. The motion platform is used to carry the crystal ingot and achieve precise movement. The drive motor provides power for the movement, and the position sensor is used to provide real-time feedback on the step position information of the crystal ingot to ensure motion accuracy.

[0117] In this embodiment, the stepping method is implemented as follows: the processing module pre-sets the stepping distance or rotation angle based on the ingot size and the field of view of the image acquisition module (e.g., 20-30mm field of view). For example, the overlap rate between the translation stepping distance and the field of view width is set to 10%-20%, and the rotation stepping angle is set to 5°-10°. The drive motor drives the motion platform and the ingot to complete one stepping movement according to the control signal of the processing module. After the position sensor detects that the ingot has reached the preset stepping position, it sends a position signal to the processing module. After receiving the signal, the processing module controls the image acquisition module to acquire a local surface image of the current field of view, and then enters the next stepping cycle until the entire surface of the ingot is scanned.

[0118] In this embodiment, during the image stitching process, the processing module extracts feature points (such as defect edges and surface texture features) from each local surface image, determines the overlapping area of ​​adjacent images through a feature point matching algorithm, and then uses image fusion technology to eliminate brightness differences and stitching traces in the overlapping area. Finally, all local images are integrated according to a preset spatial position relationship to generate image data that completely covers the surface of the crystal ingot.

[0119] It should be understood that in practical applications, the first motion module can also drive the multi-directional illumination module or the image acquisition module to move. For example, the first motion module can drive the image acquisition module to perform translational scanning along the surface of the crystal ingot, or drive the multi-directional illumination module to adjust the illumination angle as the scanning process progresses. This can also achieve image acquisition of the entire surface of the crystal ingot. This embodiment does not limit this.

[0120] In addition, in practical applications, the first motion module can drive the crystal ingot or the acquisition module to move continuously, while the image acquisition module continuously acquires images at a preset frequency, corrects the image offset during the motion process through a motion compensation algorithm, and then completes the stitching; alternatively, the first motion module may not be included, and the crystal ingot may be manually rotated or moved, and the images of each area may be manually acquired in conjunction with the image acquisition module before stitching. This embodiment does not limit this.

[0121] During the above process, the first motion module drives the crystal ingot to rotate or translate in a stepping manner, which can accurately control the movement trajectory and position of the crystal ingot, ensuring that the image acquisition module fully covers the surface of the crystal ingot according to the preset path, avoiding the problem of missed scanning of areas due to manual operation or insufficient motion accuracy, and adapting to the needs of silicon carbide crystal ingots with large surface undulations and the need for full-range detection.

[0122] Furthermore, the coordinated use of stepping motion and image acquisition allows the image acquisition module to obtain clear and stable local images at each location, reducing motion blur interference with defect signals and providing a high-quality image foundation for subsequent stitching and defect identification. The generation of complete image data not only visually presents the distribution and overall condition of all defects on the ingot surface, facilitating comprehensive quality assessment by staff, but also provides a complete image dataset for training the processing module's algorithm, further optimizing defect identification accuracy. Compared to traditional overall imaging methods, it eliminates the need for ultra-large field-of-view lenses, reducing the system's performance requirements and hardware costs, thus balancing comprehensiveness, accuracy, and cost-effectiveness.

[0123] Based on the above settings, in this embodiment, the processing module is specifically configured to: extract feature information related to surface defects from complete image data; input the feature information into a preset defect classification model or match it with a preset defect feature library to determine defect information; wherein, the defect classification model is optimized based on the defect image training set, and the defect information includes at least one of defect type, defect location, and defect size.

[0124] Specifically, in this embodiment, the processing module first preprocesses the complete image data, including noise reduction, contrast enhancement, and edge smoothing, to eliminate interference caused by splicing marks and uneven lighting. Then, it extracts the contour features of suspected defect areas using an edge detection algorithm (such as the Canny algorithm), and obtains the geometric parameters of the defects through morphological processing, including length, width, area, and shape contour (such as linear, circular, or irregular shapes). Simultaneously, it analyzes grayscale features such as the grayscale difference between the defect area and the background, and the distribution of scattered signal intensity, comprehensively forming a set of feature information related to surface defects, fully covering the physical morphology and optical response characteristics of the defects.

[0125] In one implementation, when determining defect information using a pre-defined defect classification model, the processing module standardizes the extracted feature information, converting it into an input format recognizable by the model, and then inputs it into the pre-trained defect classification model. This model, built on deep learning algorithms (such as CNN convolutional neural networks), can automatically learn the mapping relationship between different defect features and defect types and sizes. Through calculations in convolutional, pooling, and fully connected layers within the model, it directly outputs the defect type (such as scratches, cracks, microtubules, stacking faults, etc.), precise location coordinates, and specific size parameters (such as length XX μm, width XX μm).

[0126] The training process for the defect classification model is as follows: First, a large number of defect images of different types and sizes of silicon carbide ingots are collected to construct a defect image training set covering various defects such as scratches, cracks, microtubules, stacking faults, and dislocations. The images in the training set are then labeled, specifying the type, location, and size of the defect in each image. The labeled training set is then divided into a training subset and a validation subset. The training subset is input into the initial model for iterative training, continuously adjusting the model parameters through backpropagation to minimize the error between the prediction results and the labeled information. During training, the validation subset is used to verify the model's performance in real time. When the model's recognition accuracy and positioning accuracy reach a preset threshold (e.g., accuracy ≥ 95%), training stops, resulting in the final usable defect classification model. Furthermore, new defect images collected from actual detection can be continuously added to the training set to iteratively optimize the model and further improve recognition stability.

[0127] In another implementation, when determining defect information using a pre-set defect feature library, the processing module first retrieves the internally stored defect feature library. This library stores standard feature parameters for various known defects (such as scratches, cracks, microtubules, etc.), including standard geometric parameter ranges, standard grayscale feature intervals, and standard scattering signal intensity thresholds. Subsequently, the extracted defect feature information to be detected is compared one by one with the standard features of various defects in the feature library, and the feature matching degree is calculated. When the matching degree exceeds a preset threshold (e.g., 85%), the defect type and size determination rules corresponding to the standard feature are applied to the defect to be detected. Combined with the spatial coordinates of the complete image data, the specific type, location, and size of the defect to be detected are determined.

[0128] In practical applications, a hybrid approach of preliminary screening using a feature library and precise judgment using a classification model can also be adopted to determine defect information. That is, firstly, several candidate defect types with high matching degree are screened out through the defect feature library, and then the feature information is input into the classification model for precise judgment only within the candidate type range. This improves recognition efficiency and ensures judgment accuracy. This embodiment does not limit this approach.

[0129] In the aforementioned process, through systematic feature extraction and precise matching / model judgment, automated and intelligent identification of defect information is achieved. Compared with traditional manual judgment methods, this effectively improves detection efficiency and adapts to the high-speed inspection requirements of mass production lines. The defect classification model is trained on a large amount of data and can accurately capture the subtle feature differences of different defects. Combined with the full surface coverage characteristics of complete image data, it effectively avoids misjudgment or omission of microscopic and hidden defects, effectively improving the accuracy and positioning precision of defect identification.

[0130] Meanwhile, the comprehensive information output, including defect type, location, and size, provides accurate data support for ingot quality grading and subsequent processing optimization, further improving the quality control level of the silicon carbide industry chain. Furthermore, the model can be continuously optimized through data supplementation and iteration, and the feature library can be updated and expanded according to new defect types, giving the system good adaptability and scalability to meet the detection needs of different batches and different defect types.

[0131] It should be understood that in practical applications, defect analysis can be performed on each local surface image immediately after it is acquired. The specific process can refer to the analysis process based on complete image data: first, extract defect feature information from a single local image, and then determine the defect information in the local image by using a defect classification model or matching with a defect feature library. After all local images have been acquired and analyzed, integrate the defect information of all local images to form a defect report of the entire surface of the ingot. This embodiment does not limit this.

[0132] The following section provides a detailed description of the specific implementation of the multi-directional lighting module in this application.

[0133] As one possible implementation, the multi-directional lighting module includes at least two independent lighting units, which are fixedly arranged and configured to provide illumination to the ingot surface from different spatial locations and lighting angles.

[0134] Specifically, in this embodiment, at least two independent lighting units include at least one first lighting unit and at least one second lighting unit; the first lighting unit is configured to provide illumination from the outer edge of the ingot towards the center of the ingot, so as to mainly illuminate the inclined or flat area of ​​the edge of the ingot through different lighting angles; the second lighting unit is configured to provide illumination from the center area of ​​the ingot towards the edge of the ingot, so as to mainly illuminate the warped edge area of ​​the ingot.

[0135] More specifically, in this embodiment, the multi-directional lighting module includes at least two first lighting units, one for illuminating a flat area and the other for illuminating a tilted area. The two first lighting units are positioned 10-20mm outside the edge of the ingot. One unit has an illumination angle (angle with the ingot surface) of 20°-50° and a center height of 100-120mm (relative to the height of the ingot's plane), illuminating the flat area. The other unit has an illumination angle of 10°-30° and is located below the first lighting unit, with a center height of 15-30mm, adapting to the tilting characteristics of the tilted area (where the surface height decreases from the center to the edge). In this embodiment, the second lighting unit is positioned above the central area of ​​the ingot, with an illumination angle of 20°-50° and a center height of 100-120mm, covering the hidden space of the inner surface of the tilted area (where the surface height increases from the center to the edge), avoiding blind spots.

[0136] In this embodiment, the independent lighting unit can be an LED strip light source, whose light-emitting surface is homogenized to ensure lighting uniformity, and supports the output of short-wavelength light such as blue light and purple light to enhance the defect scattering signal.

[0137] It should be understood that the layout, position, angle, and other parameters of the aforementioned lighting units are not randomly set, but rather determined by adhering to the core principle of highlighting defects and suppressing interference, combined with the characteristics of the crystal ingot surface and the optical principles of detection. The primary principle is to avoid the direct reflection light receiving range of the image acquisition module. The crystal ingot surface has high reflectivity; if the illumination light is directly reflected from the crystal ingot surface and enters the camera, it will produce strong glare, masking the defect signal. Therefore, the angle setting must ensure that the propagation path of the specularly reflected light is completely offset from the receiving angle of the camera lens.

[0138] Secondly, it is necessary to adapt to the surface tilt angle of different areas of the ingot. The surface normal direction of the edge tilt area and the warped edge area are different. The illumination angle must be matched with the tilt angle of the corresponding area to ensure that the light can be effectively incident on the surface where the defect is located, so that the defect produces obvious scattering signals or shadow contrast, and avoids the formation of detection blind spots due to the inability of light to reach.

[0139] Meanwhile, the layout location needs to take into account both the illumination coverage and signal strength. A distance of 10-20mm from the outer edge and a layout above the central area can ensure that the illumination light uniformly covers the target area and that the intensity of the defect scattering signal is sufficient, providing a clear signal basis for subsequent image acquisition and feature extraction, and ultimately achieving efficient differentiation between defects and background.

[0140] As an example, Figure 4A A structural example of an ingot defect detection system provided in this application embodiment. Figure 1 , Figure 4BA structural example of an ingot defect detection system provided in this application embodiment. Figure 2 .

[0141] like Figure 4A As shown in the figure, lighting unit 1, lighting unit 2, and lighting unit 3 are independent lighting units (1 and 2 are the first lighting unit, and 3 is the second lighting unit). Independent lighting units 1 and 2 provide illumination from the outer edge of the ingot (first lighting unit layout) towards the center, while lighting unit 3 provides illumination from the center area towards the edge (second lighting unit layout). The multi-directional illumination covers the flat area, the inclined edge area, and the warped edge area of ​​the ingot. The image acquisition module acquires the light signal modulated by the ingot surface, and the processing module finally outputs the detection result (pass / pending confirmation / fail).

[0142] In this example, the multi-directional lighting module also includes a light source controller, which controls the opening, closing, brightness adjustment, and wavelength switching of the independent lighting units, and can flexibly adjust the lighting parameters according to the needs of different detection areas.

[0143] Furthermore, in this example, each independent lighting unit is equipped with a light-diffusing unit (such as a diffuser plate), which can convert the direct light from the LED light source into uniformly diffused light, further reducing reflective interference on the crystal surface and improving lighting uniformity. During the detection process, the processing module controls the first motion module to drive the motion platform to move, thereby realizing the relative position change between the crystal and the lighting unit, ensuring full surface coverage without dead angles.

[0144] like Figure 4B As shown, this example is in Figure 4A Based on the existing system, an independent lighting unit 4 is added to further expand the lighting direction, forming a multi-dimensional, all-angle lighting network. The layout of multiple lighting units can adapt to the full surface coverage of larger ingots, avoiding the problem of insufficient lighting in edge areas due to excessively large ingot diameters. The first motion module works in conjunction with the multi-directional lighting module. The motion module drives the ingot to move step by step, while the lighting module provides precise lighting simultaneously, enabling precise lighting and image acquisition in different areas of the ingot, especially suitable for the inspection needs of large-sized silicon carbide ingots.

[0145] It should be understood that, Figure 5 This is a schematic diagram illustrating the principle of an ingot defect detection system provided in an embodiment of this application. Figure 5 As shown in the figure, lighting unit 1, lighting unit 2, and lighting unit 3 are lighting units at different locations, each corresponding to targeted lighting for different defect locations. In the figure, n represents the normal direction of the flat ingot surface, i.e., the horizontal direction; i represents the direction of the lighting source; and h represents the height of the center position of the lighting source.

[0146] Specifically, such as Figure 5As shown in (1), the illumination unit 1 (first illumination unit) corresponds to the relatively flat area of ​​the inner surface of the crystal ingot (i.e., the flat area). It is illuminated from the outside of the edge of the crystal ingot at an angle α (20°-50°). The center height h of the light source is about 100-150mm. This angle and height combination can make full use of the light source, avoid blocking the light and causing the image to be too dark, and ensure that the defects in the flat area produce clear scattering signals.

[0147] like Figure 5 As shown in (2), the lighting unit 2 (first lighting unit) corresponds to the edge of the crystal ingot and the area with a large surface tilt angle (i.e., the tilted area). The lighting position is close to the edge of the crystal ingot, the center height h of the light source is about 15-30mm, and the lighting angle α is set to 10°-30°. This can avoid local dark or overexposed images and enhance the contrast of defects in the large tilt angle area at the edge.

[0148] like Figure 5 As shown in (3), the lighting unit 3 (second lighting unit) corresponds to the warped edge area (i.e., the warped edge area) near the edge of the crystal ingot and with a large tilt angle towards the center. It illuminates from the center to the edge at an angle α (20°-50°). The center height h of the light source is about 100-150mm to ensure that the light can illuminate the inner surface of the warped edge and capture the hidden defect signal.

[0149] Meanwhile, in this example, since the surface undulation of the crystal ingot is >1mm, the objective lens of the image acquisition module is a high field of view deep center lens with specific parameters adapted to the 20-30mm field of view requirement, depth of field 2-3mm, working distance about 220mm, and resolution about 10μm. This reduces the system's requirements for high-precision motion control while ensuring that defects at different locations can be clearly imaged.

[0150] In this example, either an area scan camera or a line scan camera can be used to receive the image formed by the objective lens and convert it into image data. Correspondingly, in addition to controlling the rotation / translation of the motion platform, the processing module stores and classifies the acquired image information for training and optimizing the defect recognition algorithm, further improving the accuracy of defect recognition on the actual production line.

[0151] As another possible implementation, the multi-directional lighting module includes a lighting unit and a second motion module; the second motion module is configured to change the relative position of the lighting unit and the crystal ingot surface, so that the lighting unit can illuminate the crystal ingot surface from different directions.

[0152] Specifically, the lighting unit can be selected from one or more integrated light sources (such as integrated multi-wavelength LED light sources). When a single lighting unit is used, the second motion module can drive it to move along the circumference or radial direction of the ingot, or adjust the lighting angle to achieve lighting coverage in different directions. When multiple lighting units are used, the second motion module can drive some units to move to adapt to different ingot sizes.

[0153] In practical applications, the second motion module can also reuse the existing first motion module (i.e., the motion platform that carries the crystal ingot). By coordinating the stepping motion of the crystal ingot with the angle of the fixed lighting unit, the relative position change between the lighting direction and the surface of the crystal ingot can be indirectly achieved without the need for additional hardware, thus reducing system complexity and cost.

[0154] As an example, Figure 4C A structural example of an ingot defect detection system provided in this application embodiment. Figure 3 ,like Figure 4C As shown, for the three independent lighting units (lighting unit 1, lighting unit 2, and lighting unit 3), the first motion module has dual control functions. On the one hand, it controls the motion platform to rotate or translate the ingot to achieve full-surface scanning. On the other hand, according to the algorithm instructions of the processing module, it changes the lighting angle of the lighting unit in real time during the detection process. For example, when a tilted area is detected at the edge of the ingot, the algorithm determines through image analysis that the current lighting angle cannot effectively highlight the defect signal. Then, it precisely adjusts the tilt angle of the lighting unit through the motion control system until the contrast between the defect and the background reaches the optimal level.

[0155] The above-described configurations allow the multi-directional illumination module to adapt to the surface characteristics of crystal ingots, such as high reflectivity, large undulations, and large edge tilt angles. The fixed, independent illumination unit approach enables simultaneous illumination of multiple areas, resulting in high detection efficiency. The approach combined with a motion module offers greater flexibility, adapting to the detection needs of crystal ingots of different sizes and surface morphologies. Both methods eliminate blind spots inherent in single-illumination systems, enhance defect signal contrast, and provide a stable optical foundation for subsequent image acquisition and defect identification, while also ensuring the system's practicality and scalability.

[0156] As another possible implementation, the multi-directional illumination module includes a laser source, a beam shaping unit, a beam splitting unit, and at least two beam reflecting units; the beam shaping unit is configured to modulate the laser into a uniform spot of a preset size and shape, the beam splitting unit is configured to divide the uniform spot into at least two sub-beams whose on / off state or intensity can be independently controlled; the beam reflecting units are configured to adjust the propagation direction of each sub-beam so that at least two sub-beams are emitted from different spatial directions.

[0157] Correspondingly, the processing module is also configured to independently control the opening and closing or intensity combination of at least two sub-beams according to the morphological characteristics of the current scanning area during the detection process, so as to adaptively illuminate the flat or tilted areas of the ingot.

[0158] Specifically, when implementing multi-directional illumination based on the above architecture, a highly coherent laser beam is first output from the laser source. The beam shaping unit (such as a combination of beam expander and homogenizer) modulates the beam into a light spot with a size that matches the detection field of view and a uniform energy distribution, thus meeting the industrial detection requirements for beam stability and uniformity.

[0159] The light spot then enters a beam splitting unit (such as a polarizing beam splitter or beam splitter) and is split into at least two sub-beams with independently adjustable intensity. Each sub-beam corresponds to a beam reflecting unit (such as an adjustable-angle reflector). By adjusting the angle of the reflector, different sub-beams are directed toward the ingot surface in different spatial directions (such as a high angle corresponding to a flat area of ​​the ingot and a low angle corresponding to a tilted area).

[0160] During the image acquisition phase, the processing module independently controls the activation, deactivation, or intensity of the corresponding sub-beam based on the topographic features of the current scanning area (such as flat or tilted areas). For example, when scanning a flat area, the high-angle sub-beam is activated and its intensity is increased; when scanning a tilted area, the low-angle sub-beam is switched to and its intensity is adjusted to ensure that the defects in the target area generate clear scattering signals. Then, the image acquisition module synchronously acquires the light signal modulated by the crystal surface to complete the acquisition of the defect image.

[0161] It should be understood that, in practical applications, the beam shaping unit can use a microstructured mirror instead of the combination of a beam expander and homogenizer. This total internal reflection design can resist the effects of thermal effects and is suitable for high-power laser scenarios. The beam splitting unit can use an acousto-optic modulator to achieve rapid switching of sub-beams and continuous adjustment of light intensity, improving dynamic adaptation efficiency. The beam reflection unit can use an electrically driven rotating mirror instead of a manually adjusted mirror to achieve automated and precise control of the sub-beam direction, reducing the cost of manual intervention. This embodiment does not limit this aspect.

[0162] As an example, Figure 6 Figure 4 shows a structural example of an ingot defect detection system provided in an embodiment of this application. Figure 6 As shown, the laser beam output from the laser source is processed into a uniform spot by the shaping unit, split into multiple sub-beams by the beam splitting unit, and then the propagation direction is adjusted by the reflection unit, so that they are directed onto the surface of the crystal ingot from different spatial angles. The processing module controls the opening and closing and intensity of different sub-beams based on the current area morphology fed back by the motion platform. The image acquisition module collects defect signals and converts them into image data by the camera. Finally, the processing module outputs the detection results. At the same time, the motion control system synchronously controls the movement of the platform to achieve adaptive illumination and scanning of the entire surface of the crystal ingot.

[0163] The aforementioned setup, through a laser light source and beam-splitting reflection architecture, enables precise and independent control of the direction and intensity of sub-beams. Compared to independent LED illumination units, the high directionality and brightness of lasers further enhance the scattering signals from microscopic defects (such as microtubes and stacking faults), improving the sensitivity of defect identification. Simultaneously, the independent on / off switching and intensity adjustment of sub-beams can more accurately adapt to the morphological characteristics of different regions of the ingot, avoiding signal interference caused by ineffective illumination. Furthermore, the coherence of the laser light source, combined with beam shaping, yields a more uniform illumination spot, reducing glare interference from highly reflective surfaces of the ingot and further improving the contrast of defect images. This architecture balances the accuracy and flexibility of illumination, making it particularly suitable for scenarios requiring high precision in microscopic defect detection. Moreover, compared to solutions with multiple independent light sources, it offers higher hardware integration and facilitates system miniaturization.

[0164] The crystal ingot defect detection system provided in this application adapts to different morphologies of flat areas, inclined edges, and warped edges of crystal ingots through the differentiated design of multi-directional illumination modules. Combined with short-wavelength light sources, uniform light processing, and high-field deep-center lenses, it effectively suppresses glare interference from highly reflective surfaces, enhances the signal contrast of micro-defects, and achieves independent opening and closing of sub-beams and fine intensity control by using a laser beam splitting and reflection architecture or a multi-independent illumination unit scheme, effectively eliminating the detection blind zone of single illumination.

[0165] Simultaneously, the processing module and motion control system work together to automate and precisely adjust the detection process, and the acquired image data can be used for continuous training and optimization of the algorithm model. In addition, the system supports the selection of area array or line array cameras and can reuse motion modules to reduce hardware costs. It is suitable for the full-surface, blind-angle-free detection requirements of ingots of different sizes, while taking into account the detection accuracy, efficiency and economics of industrial applications, providing efficient and reliable intelligent technical support for the quality control of ingot production.

[0166] This application also provides a method for detecting defects in crystal ingots. Figure 7 This is a flowchart illustrating a method for detecting defects in an ingot, as provided in an embodiment of this application. Figure 7 As shown, the method of this application includes:

[0167] S701. Provide illumination light from at least two different directions to the surface of the ingot to be inspected, so that both the flat and tilted areas of the ingot can be effectively illuminated during the inspection process.

[0168] S702. Acquire the modulated optical signal on the illuminated crystal ingot surface and convert the optical signal into image data.

[0169] S703 Receives and processes image data to identify and output defect information on the surface of the crystal ingot.

[0170] Specifically, the method of this application is executed based on the ingot defect detection system proposed in any of the aforementioned system embodiments, and the processing module therein can be the main execution body. The processing module controls the multi-directional illumination module to project illumination light from different directions such as the outer edge of the ingot and the central area, adapting to the morphological characteristics of flat areas, inclined edge areas and warped edge areas. At the same time, the processing module activates the light homogenizing unit to homogenize the illumination light output from the multi-directional light source to ensure illumination uniformity.

[0171] In this embodiment, the processing module controls the image acquisition module to convert the difference in reflected light signals between the defective region and the smooth surface into digital image data, adapting to the undulation characteristics of the ingot surface greater than 1 mm. The image data is then preprocessed and features extracted. By matching with a preset defect classification model or feature library, the defect type, location, and size are identified. The acquired image data can also be used to supplement the training set for algorithm optimization.

[0172] In the above process, the illumination light of the multi-directional illumination module used is short-wavelength LED light such as blue light or violet light, or laser light, to enhance the contrast of defect signals. In addition, for details regarding the adjustment of the illumination angle (10°-30° to adapt to the edge tilted area, 20°-50° to adapt to the flat and warped edge areas), the rotation / translation scanning coordination of the motion platform, and the selection of optical path parameters, please refer to the aforementioned system embodiment, which will not be repeated here.

[0173] It should be understood that the subject executing the method of this application may also be any independent electronic device other than the processing module in the aforementioned system embodiments. This electronic device can interact with the multi-directional illumination module, image acquisition module and processing module in the ingot defect detection system to implement the method of this application, and this application does not limit it.

[0174] The method described in this application can specifically avoid specular reflection interference from the highly reflective surface of crystal ingots, while simultaneously covering the illumination blind spots of large undulation surfaces and areas with large edge tilt angles. Combined with uniform illumination processing, it further improves illumination uniformity, enabling clear signal responses from defects of different locations and types, effectively solving the problem of poor adaptability of fixed illumination methods. The improved defect signal contrast brought about by multi-directional illumination effectively reduces the risk of missed and false detections, ultimately achieving a significant improvement in detection accuracy. The entire method requires no lens switching; it can be adapted to the undulation characteristics of crystal ingot surfaces greater than 1mm when paired with a high-field, deep-center lens, reducing the system's requirements for high-precision motion control while maintaining detection efficiency and stability. Continuous training and optimization of the algorithm model using image data further enhances the system's ability to identify complex defects.

[0175] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0176] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0177] The above embodiments introduce a method for detecting crystal defects from the perspective of process flow. The following embodiments introduce a crystal defect detection device from the perspective of virtual module or virtual unit. For details, please refer to the following embodiments.

[0178] This application also provides an ingot defect detection device. Figure 8 This is a schematic diagram of the structure of an ingot defect detection device provided in an embodiment of this application, as shown below. Figure 8 As shown, in this embodiment, the ingot defect detection device may include:

[0179] The lighting control module 81 is used to provide illumination light from at least two different directions to the surface of the crystal ingot to be inspected, so that both the flat and tilted areas of the crystal ingot can be effectively illuminated during the inspection process.

[0180] The acquisition and control module 82 is used to acquire the optical signal modulated on the illuminated crystal ingot surface and convert the optical signal into image data;

[0181] The defect identification module 83 is used to receive and process image data to identify and output defect information on the surface of the crystal ingot.

[0182] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.

[0183] This application provides an electronic device. Figure 9 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application, such as... Figure 9 As shown, Figure 9 The illustrated electronic device includes at least one processor 91 and a memory 92. The processor 91 and the memory 92 are connected, for example, via a bus 93. Optionally, the electronic device may also include a transceiver 94. It should be noted that in practical applications, the transceiver 94 is not limited to one, and the structure of this electronic device does not constitute a limitation on the embodiments of this application.

[0184] Processor 91 may be a central processing unit (CPU), a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 91 may also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0185] Bus 93 may include a pathway for transmitting information between the aforementioned components. Bus 93 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Bus 93 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0186] The memory 92 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0187] The memory 92 stores computer execution instructions for implementing the present application's solution, and its execution is controlled by the processor 91. The processor 91 executes the computer execution instructions stored in the memory 92 to implement the content shown in the foregoing method embodiments.

[0188] This application also provides a computer-readable storage medium, which may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk. Specifically, the computer-readable storage medium stores computer-executable instructions, which are used to implement the methods in the above embodiments.

[0189] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the technical solution of the above method embodiments. Its implementation principle and technical effects are similar, and will not be repeated here.

[0190] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0191] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0192] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A crystal ingot defect detection system, characterized in that, The system includes: A multi-directional illumination module is configured to provide illumination light from at least two different directions to the surface of the ingot to be inspected, so that both the flat and tilted areas of the ingot can be effectively illuminated during the inspection process. The image acquisition module has an acquisition optical path facing the surface to be detected, and is configured to acquire optical signals that are modulated by the surface of the ingot and enter the acquisition optical path under the illumination of the multi-directional illumination module, and convert the optical signals into image data; The processing module is configured to receive and process the image data to identify and output defect information on the surface of the ingot.

2. The system according to claim 1, characterized in that, The processing module is further configured to: when it is identified that there is a warped edge region on the surface of the ingot, control the multi-directional lighting module to provide illumination light suitable for illuminating the warped edge region.

3. The system according to claim 1, characterized in that, The system further includes a first motion module, and the processing module is specifically configured as follows: The first motion module is controlled to drive the crystal ingot to perform step-by-step rotation or translation. At each step position, the image acquisition module is controlled to acquire one or more frames of local surface images corresponding to the current field of view; The images of the local surfaces obtained after scanning the entire surface of the crystal ingot are stitched together to generate complete image data of the crystal ingot surface.

4. The system according to any one of claims 1-3, characterized in that, The multi-directional lighting module includes at least two independent lighting units, which are fixedly arranged and configured to provide illumination to the surface of the ingot from different spatial positions and lighting angles. or, The multi-directional lighting module includes a lighting unit and a second motion module; The second motion module is configured to change the relative position of the illumination unit and the crystal ingot surface, so that the illumination unit can illuminate the crystal ingot surface from different directions.

5. The system according to claim 4, characterized in that, The at least two independent lighting units include at least one first lighting unit and at least one second lighting unit; The first lighting unit is configured to provide illumination from the outer edge of the ingot towards the center of the ingot, so as to primarily illuminate the inclined or flat edge area of ​​the ingot through different lighting angles; The second lighting unit is configured to provide illumination from the central region of the ingot toward the edge of the ingot, so as to primarily illuminate the warped edge region of the ingot.

6. The system according to any one of claims 1-3, characterized in that, The multi-directional illumination module includes a laser source, a beam shaping unit, a beam splitting unit, and at least two beam reflecting units. The beam shaping unit is configured to modulate the laser into a uniform spot of a preset size and shape, and the beam splitting unit is configured to split the uniform spot into at least two sub-beams whose on / off state or intensity can be independently controlled. The beam reflecting unit is configured to adjust the propagation direction of each of the sub-beams, so that at least two sub-beams are emitted from different spatial directions; Accordingly, the processing module is also configured to, during the detection process, independently control the opening and closing or intensity combination of the at least two sub-beams according to the morphological characteristics of the current scanning area, so as to adaptively illuminate the flat or tilted areas of the ingot.

7. The system according to any one of claims 1-3, characterized in that, The processing module is also configured to: Based on the received preset detection mode command, and / or based on the defect features obtained from real-time analysis of the image data, the current defect type is determined; Based on the current defect type and the pre-stored correspondence between defect types and illumination wavelengths, the multi-directional illumination module is controlled to provide illumination at a specific wavelength.

8. The system according to any one of claims 1-3, characterized in that, The multi-directional lighting module also includes a light homogenizing unit, which is disposed on the propagation path of the lighting light and is used to homogenize the lighting light.

9. The system according to claim 3, characterized in that, The processing module is specifically configured as follows: Extract feature information related to surface defects from the complete image data; The feature information is input into a preset defect classification model or matched with a preset defect feature library to determine the defect information; wherein the defect classification model is optimized based on a defect image training set, and the defect information includes at least one of defect type, defect location, and defect size.

10. A method for detecting defects in a crystal ingot, characterized in that, The method includes: Illumination light from at least two different directions is provided to the surface of the ingot to be inspected so that both the flat and tilted areas of the ingot can be effectively illuminated during the inspection process. The optical signal modulated on the illuminated surface of the ingot is acquired, and the optical signal is converted into image data; The image data is received and processed to identify and output defect information on the surface of the ingot.

Images