An intelligent rapid detection system and method for LED chip appearance

Abstract

The present invention discloses an intelligent rapid detection system and method for the appearance of an LED chip, and relates to the field of chip detection technology. The system constructs a light source combination matrix L through a light source control module, accurately sets the light source state, pitch angle and azimuth angle, and realizes multi-angle irradiation of the LED chip. The image acquisition module uses an industrial camera with a polarization filter to synchronously capture images at different angles, and obtains high-quality standard images through preprocessing such as perspective correction and distortion correction. In the feature extraction and fusion module, the system extracts key features such as multi-angle reflection index R (A, B), local surface texture parameters T (x, y) and light source polarization angle P (A, B), and generates a fused image Ifused through comprehensive calculation. This detection system can more accurately identify subtle defects on the chip surface, significantly improve detection accuracy and consistency, and reduce the time and error of manual detection.

Description

Technical Field

[0001] The present invention relates to the technical field of chip detection, and in particular to an intelligent rapid detection system and method for the appearance of an LED chip. Background Art

[0002] The LED chip appearance intelligent rapid detection system belongs to the field of industrial automation detection, especially the application scope of intelligent visual detection technology. In modern manufacturing, the appearance quality of LED chips directly affects their performance and market competitiveness. Therefore, efficient and accurate appearance detection has become a vital part of the LED chip production process. The detection system uses multi-angle light source analysis, image processing and intelligent algorithms to detect subtle defects on the surface of LED chips to ensure high-quality output of products.

[0003] In the Chinese invention application with application publication number CN112964635B, a chip detection method and system are disclosed, the chip detection method includes: controlling the target light emitted by the light source to irradiate the chip to be detected arranged on one side of the carrier substrate through the detection substrate, so that the side of the detection substrate close to the chip to be detected forms a wedge interference fringe corresponding to each chip to be detected; collecting the wedge interference fringe corresponding to each chip to be detected by the image acquisition device to form an interference fringe image corresponding to each chip to be detected; matching the interference fringe image with the preset image to obtain the detection result. The chip detection method of the present invention can not only detect whether the chip is placed correctly, but also detect multiple aspects including whether the chip is uneven in the vertical direction, and the detection is accurate and precise.

[0004] In combination with the prior art, the above application still has the following deficiencies:

[0005] At present, due to the limitations of existing detection technology, LED chips are prone to missed detection or false detection during the manufacturing process. Especially when there are complex textures or light reflection characteristics on the chip surface, traditional detection systems cannot fully analyze and process this information, making it difficult to detect subtle cracks, scratches, corrosion or other defects on the surface. If these defects are not detected and controlled in time, they will directly affect the luminous efficiency and service life of the chip, and then affect the quality and market reputation of downstream products. As the market demand for high-quality LED products increases, these detection deficiencies pose a severe challenge to manufacturers' production efficiency and economic benefits. Summary of the invention

[0006] In view of the deficiencies in the prior art, the present invention provides an intelligent rapid detection system and method for the appearance of an LED chip, which solves the problems mentioned in the background technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: including a light source control module, an image acquisition module, a feature extraction and fusion module, a defect recognition module and a comprehensive detection and analysis module;

[0008] The light source control module integrates the light source control device by constructing a light source combination matrix L on a computer light source control platform to set the shooting light source in multiple angles and directions;

[0009] The image acquisition module is used to successively acquire the appearance images of the LED chips under different light source combinations according to the light source combination matrix L, and pre-process the acquired appearance images of the LED chips to obtain a standard appearance picture set;

[0010] The feature extraction and fusion module is used to extract features from the annotated appearance picture set, obtain a picture feature set, and then construct a related algorithm formula, extract the picture feature set and input it into the related algorithm formula, output a multi-angle reflection index R (A, B), a local surface texture parameter T (x, y) and a light source polarization angle P (A, B), and then perform a comprehensive calculation on the multi-angle reflection index R (A, B), the local surface texture parameter T (x, y) and the light source polarization angle P (A, B) to obtain a fused image Ifused;

[0011] The defect recognition module constructs a related defect algorithm formula, inputs the fused image Ifused into the related defect algorithm formula, and outputs an edge detection score Sedge, a texture analysis score Stexture, and a reflection anomaly detection score Sreflection;

[0012] The comprehensive detection and analysis module obtains a comprehensive score Scom by comprehensively calculating the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection, and compares the preset qualified threshold T with the comprehensive score Scom to evaluate and analyze the detection situation.

[0013] Preferably, the light source control module is integrated on a computer control platform by constructing a light source combination matrix L and a light source control device, and sets the light source state l, the pitch angle A and the azimuth angle B of the shooting light source for control, and the light source control device includes a digital signal processing DSP, a pulse width modulation device, an automatic control motor and an automatic rotating bracket;

[0014] The light source is turned on and off by a digital signal processor DSP, the brightness of the light source is adjusted by a pulse width modulation device, the light source automatic control motor is used to change the pitch angle of the light source, and the automatic rotating bracket is used to adjust the azimuth angle of the light source;

[0015] The matrix form of the light source combination matrix L is as follows:

[0016] ;

[0017] In the formula, l ij Indicates the light source status of the jth light source in the i-th group. The specific value of l is binary 0 and 1. 0 means the light source is off, and 1 means the light source is on. ij represents the pitch angle of the jth light source in the i-th group, B ij It represents the azimuth angle of the jth light source in the i-th group. The meaning of the formula is: the light source combination matrix L is used to accurately control the combination and configuration of light sources, and send instructions to the light source control device on the computer control platform to adjust the light source.

[0018] Preferably, the image acquisition module includes an image acquisition unit and an image preprocessing unit;

[0019] The image acquisition unit works synchronously with the light source control device by using an industrial camera with a polarization filter, and sets a software trigger mechanism for the industrial camera. The trigger mechanism triggers the industrial camera to shoot when the light source is adjusted successively by the light source combination matrix L, and captures the appearance image of the LED chip under different light source angles in real time;

[0020] The image preprocessing unit is integrated with the industrial camera through the WIFI wireless transmission protocol, and the appearance image of the LED chip captured and photographed in real time is transmitted to the central processor, and the appearance image of the LED chip is preprocessed, and then the preprocessed appearance image of the LED chip is summarized to obtain a standard appearance picture set, and the preprocessing method includes image correction and image noise removal;

[0021] The image correction includes perspective correction, radiometric distortion correction and image registration to eliminate the collective distortion and deviation caused by the same light source angle and industrial camera position;

[0022] The image noise removal is performed by replacing each pixel of the horizontal axis x and the vertical axis y in the image with the average value of all pixels in the field, and smoothing the image to remove random noise in the appearance image of the LED chip.

[0023] Preferably, the feature extraction and fusion module includes a feature extraction unit, a feature calculation unit and a feature fusion unit;

[0024] The feature extraction unit is used to extract features from the standard image set obtained after preprocessing to obtain a picture feature set;

[0025] The image feature set includes reflected light intensity Iref, incident light intensity Iinc, horizontal axis electric field component Ex, and vertical axis electric field component Ey;

[0026] The reflected light intensity Iref is extracted by converting the RGB values ​​of the pixels of the standard image set into gray values ​​for analysis. The specific extraction method is as follows: , where I(x, y) represents the brightness value at the pixel point (x, y);

[0027] The incident light intensity Iinc is obtained by extracting the characteristics of the light source power P and the area A on the chip surface irradiated by the light. The specific extraction method is: , where the light source power P is obtained from the specifications of the light source equipment, and the area A of the chip surface irradiated by the light is obtained by the camera imaging method;

[0028] The horizontal electric field component Ex and the vertical electric field component Ey are captured by a polarization filter to obtain the polarized light intensity of the standard image set, and then feature extraction is performed. The specific extraction method is: , where Ipz(x, y) represents the polarized light intensity measured at the pixel point (x, y) of the standard image set;

[0029] The feature extraction unit includes a multi-angle reflection analysis unit, a local surface texture analysis unit and a light source polarization analysis unit;

[0030] The multi-angle reflection analysis unit is used to construct a multi-angle reflection algorithm formula, extract the reflected light intensity Iref and the incident light intensity Iinc in the image feature set and input them into the multi-angle reflection algorithm formula, calculate and output the multi-angle reflection index R (A, B);

[0031] The multi-angle reflection index R (A, B) is calculated and obtained by the following multi-angle reflection algorithm formula;

[0032] ;

[0033] Where Iref(A, B) represents the intensity of reflected light at angle (A, B), and Iinc(A, B) represents the intensity of incident light at angle (A, B).

[0034] Preferably, the local surface texture analysis unit constructs a local texture calculation formula by using the variance of the second-order differential of the image, calculates and outputs the local surface texture parameter T (x, y) of the standard image set;

[0035] The local surface texture parameter T (x, y) is calculated and obtained by the following local texture calculation formula;

[0036] ;

[0037] Where Var represents the variance of the second-order differential of the image.

[0038] Preferably, the light source polarization analysis unit constructs a light source polarization angle algorithm formula, extracts the horizontal axis electric field component Ex and the vertical axis electric field component Ey in the image feature set, inputs them into the light source polarization angle algorithm formula, and calculates and outputs the light source polarization angle P (A, B);

[0039] The light source polarization angle P (A, B) is obtained by the following light source polarization angle calculation formula;

[0040] ;

[0041] Wherein, Ey(A, B) represents the component of the electric field in the horizontal axis x direction in the light determined by the angle (A, B), Ex(A, B) represents the component of the electric field in the vertical axis y direction in the light determined by the angle (A, B), and arctan represents the inverse tangent function;

[0042] The feature fusion unit is used to comprehensively calculate the acquired multi-angle reflection index R (A, B), local surface texture parameter T (x, y) and light source polarization angle P (A, B) to obtain a fused image Ifused;

[0043] The fused image Ifused is obtained by calculation using the following algorithm formula;

[0044] ;

[0045] Where M represents the total number of fusion items, representing multi-angle reflection characteristics, multi-position texture parameters and multi-polarization angles in the fusion process, m represents the fusion item index, Am represents the pitch angle of the light source under the m-th measurement condition, Bm represents the azimuth angle of the light source under the m-th measurement condition, xm represents the horizontal axis pixel coordinate in the image under the m-th measurement condition, and ym represents the vertical axis pixel coordinate in the image under the m-th measurement condition.

[0046] Preferably, the defect recognition module includes an edge detection unit, a texture analysis unit and a reflection anomaly detection unit;

[0047] The edge detection unit calculates the edge strength of the fused image Ifused by using the Canny detection algorithm, constructs an edge detection score algorithm formula, inputs the detected edge strength into the edge detection score algorithm formula, performs edge score calculation to obtain an edge detection score Sedge;

[0048] The edge detection score Sedge is calculated by the following algorithm formula:

[0049] ;

[0050] Where N represents the number of edge points, Gx represents the gradient of the fused image in the horizontal axis x direction, Gy represents the gradient of the fused image in the vertical axis y direction, EdgeRegion represents the edge region in the fused image, Indicates texture intensity;

[0051] The texture analysis unit extracts the energy coefficients of the fused image Ifused at different scales by using wavelet transform, extracts the contrast C, entropy E and texture uniformity H parameters by using the gray level co-occurrence matrix GLCM, and constructs a texture score algorithm formula, inputs the contrast C, entropy E and uniform texture H into the texture score algorithm formula, and calculates and outputs the texture analysis score Stexture;

[0052] The texture analysis score Stexture is calculated and obtained by the following texture score algorithm formula;

[0053] ;

[0054] In the formula, K represents the total number of detection areas, C k represents the contrast of the kth detection area, E k represents the entropy of the kth detection area, H k represents the texture uniformity of the kth detection area;

[0055] The reflection anomaly detection unit is used to construct a reflection score algorithm formula, and then input the obtained multi-angle reflection index R (A, B) into the reflection score algorithm formula to calculate and output the reflection anomaly detection score Sreflection;

[0056] The texture analysis score is calculated by the following reflection score algorithm formula;

[0057] ;

[0058] Where Rre(A, B) represents the reference multi-angle reflectance index, represents the reflection characteristic deviation of the kth detection area.

[0059] Preferably, the comprehensive detection and analysis module includes a comprehensive score calculation unit and a comprehensive evaluation unit;

[0060] The comprehensive score calculation unit is used to comprehensively calculate the edge detection score Sedge, the texture analysis score Stexture and the reflection abnormality detection score Sreflection to obtain a comprehensive score Scom;

[0061] The comprehensive score Scom is calculated by the following algorithm formula:

[0062] ;

[0063] Wherein, a1, a2 and a3 represent the preset weight values ​​of the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection, respectively, and their specific values ​​are set by the user.

[0064] Preferably, the comprehensive evaluation unit compares and evaluates the qualified threshold T set by the user with the comprehensive score Scom, analyzes the appearance of the LED chip to quickly detect the qualification, and generates intelligent early warning information. The specific evaluation content is as follows;

[0065] When the comprehensive score Scom is less than the qualified threshold T, it means that the test result is qualified and no warning information needs to be generated;

[0066] When the comprehensive score Scom ≥ the qualified threshold T, it means that the test result is unqualified. At this time, an early warning message is generated to prompt that the appearance of the LED chip currently being tested is abnormal.

[0067] An intelligent and rapid detection method for the appearance of an LED chip comprises the following steps:

[0068] S1. By constructing a light source combination matrix L on a computer light source control platform and integrating it with a light source control device, the shooting light source is set at multiple angles and in multiple directions;

[0069] S2. According to the light source combination matrix L, the appearance images of the LED chips under different light source combinations are collected one by one, and the collected appearance images of the LED chips are preprocessed to obtain a standard appearance picture set;

[0070] S3, extract features from the annotated appearance picture set to obtain a picture feature set, then construct a related algorithm formula, extract the picture feature set and input it into the related algorithm formula, output the multi-angle reflection index R (A, B), the local surface texture parameter T (x, y) and the light source polarization angle P (A, B), and then perform comprehensive calculation on the multi-angle reflection index R (A, B), the local surface texture parameter T (x, y) and the light source polarization angle P (A, B) to obtain a fused image Ifused;

[0071] S4, construct a related defect algorithm formula, input the fused image Ifused into the related defect algorithm formula, and output the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection;

[0072] S5. Comprehensively calculate the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection to obtain a comprehensive score Scom, and compare the preset qualified threshold T with the comprehensive score Scom to evaluate and analyze the detection situation.

[0073] The present invention provides a system and method for intelligent and rapid detection of LED chip appearance. It has the following beneficial effects:

[0074] (1) The system significantly improves the accuracy and efficiency of detection by integrating five modules: light source control, image acquisition, feature extraction and fusion, defect recognition, and comprehensive detection analysis. The light source control module constructs a light source combination matrix L on the computer light source control platform, accurately sets the light source state l, pitch angle A, and azimuth angle B, and ensures that the LED chip is illuminated from multiple angles and directions. The image acquisition module uses an industrial camera with a polarization filter to synchronously capture images under different light source angles, and eliminates distortion and noise through image preprocessing such as perspective correction, radial distortion correction, and image registration to obtain a high-quality standard appearance picture set. In the feature extraction and fusion module, the system extracts key features such as multi-angle reflection index R (A, B), local surface texture parameters T (x, y), and light source polarization angle P (A, B) from these picture sets, and generates a fused image Ifused through comprehensive calculation of multi-angle reflection index R (A, B), local surface texture parameters T (x, y), and light source polarization angle P (A, B). This multi-angle, multi-dimensional detection method enables the system to more accurately identify subtle defects on the chip surface, greatly improving the accuracy and consistency of detection, while reducing the time and errors of manual inspection.

[0075] (2) The system adopts a modular design, which makes it highly flexible and adaptable. The light source control module flexibly controls the state, brightness, pitch angle and azimuth of the light source through digital signal processor DSP, pulse width modulation equipment, automatic control motor and automatic rotating bracket. The image acquisition module is closely integrated with the light source control equipment through the soft trigger mechanism of the industrial camera and the WIFI wireless transmission protocol, and captures and transmits images under different light source angles in real time. It also ensures the stability of image quality through pre-processing technologies such as perspective correction and radiation distortion correction. The feature extraction and fusion module further extracts and fuses the multi-angle reflection index R (A, B), local surface texture parameters T (x, y) and light source polarization angle P (A, B) to obtain the fused image Ifused, ensuring that the system can flexibly respond to different detection needs in various complex production environments. This modular and adjustable design enables the system to maintain efficient detection capabilities when dealing with different product specifications, production line rhythms and complex production environments, ensuring the stability of product quality.

[0076] (3) The system realizes intelligent evaluation and real-time warning of the detection results through the comprehensive detection and analysis module. The defect recognition module obtains the edge detection score Sedge, texture analysis score Stexture and reflection anomaly detection score Sreflection through multiple technologies such as edge detection, texture analysis and reflection anomaly detection. The comprehensive detection and analysis module calculates these scores comprehensively to generate a comprehensive score Scom. The system sets a qualified threshold T and compares the comprehensive score Scom with the qualified threshold T. When the comprehensive score Scom is lower than the qualified threshold T, the system automatically determines that the detection result is qualified; when the comprehensive score Scom is higher than or equal to the qualified threshold T, the system immediately generates intelligent warning information, indicating that the detected LED chip has an abnormality or defect in appearance. This intelligent warning function not only reduces the need for manual intervention, but also significantly improves the automation level of the production line, ensuring that problems in the production process can be identified and handled in a timely manner, thereby ensuring high-quality production of LED chips and reliable product performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 This is a schematic diagram of the process of an intelligent rapid detection system for LED chip appearance according to the present invention;

[0078] Figure 2 The present invention is a schematic diagram of the steps of a method for intelligently and rapidly detecting the appearance of an LED chip. DETAILED DESCRIPTION

[0079] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0080] Example 1: Please refer to Figure 1 , the present invention provides an intelligent rapid detection system for the appearance of LED chips. To achieve the above purpose, the present invention is implemented through the following technical solutions: including a light source control module, an image acquisition module, a feature extraction and fusion module, a defect recognition module and a comprehensive detection and analysis module;

[0081] The light source control module integrates the light source control device by constructing the light source combination matrix L on the computer light source control platform to set the shooting light source in multiple angles and directions;

[0082] The image acquisition module is used to successively acquire the appearance images of the LED chips under different light source combinations according to the light source combination matrix L, and pre-process the acquired appearance images of the LED chips to obtain a standard appearance picture set;

[0083] The feature extraction and fusion module is used to extract features from the annotated appearance picture set, obtain the picture feature set, and then construct the relevant algorithm formula. The extracted picture feature set is input into the relevant algorithm formula, and the multi-angle reflection index R (A, B), local surface texture parameter T (x, y) and light source polarization angle P (A, B) are output. The multi-angle reflection index R (A, B), local surface texture parameter T (x, y) and light source polarization angle P (A, B) are then comprehensively calculated to obtain the fused image Ifused;

[0084] The defect recognition module constructs the relevant defect algorithm formula, inputs the fused image Ifused into the relevant defect algorithm formula, and outputs the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection;

[0085] The comprehensive detection and analysis module obtains the comprehensive score Scom by comprehensively calculating the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection, and compares the preset qualified threshold T with the comprehensive score Scom to evaluate and analyze the detection situation.

[0086] In this embodiment, the system constructs a light source combination matrix L on a computer light source control platform through a light source control module, accurately sets the state, pitch angle and azimuth of the light source, and realizes multi-angle and multi-directional illumination of the LED chip. The image acquisition module successively acquires images according to the light source combination matrix L, and performs preprocessing to generate a high-quality standard appearance picture set. The feature extraction and fusion module further extracts the multi-angle reflection index R (A, B), the local surface texture parameter T (x, y) and the light source polarization angle P (A, B), and comprehensively calculates and generates a fused image Ifused. The defect recognition module uses the fused image Ifused through the relevant defect algorithm formula to output the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection. Finally, the comprehensive detection and analysis module obtains the comprehensive score Scom through comprehensive calculation, and compares it with the preset qualified threshold T to comprehensively evaluate the detection results. Through the collaborative work of each module, the system has been significantly improved in terms of detection speed, accuracy and stability. The multi-angle and multi-dimensional detection method enables the system to maintain efficient and reliable operation in a complex production environment. Compared with existing detection technologies, the system not only shortens detection time and reduces labor costs, but also significantly improves detection accuracy and reduces product quality risks caused by insufficient detection. In the future, the system is expected to further improve the detection capabilities of various complex defects by introducing more advanced intelligent algorithms and optimization technologies, and promote the development of the LED chip manufacturing industry to a higher level of automation and intelligence.

[0087] Example 2: This example is an explanation of Example 1. Figure 1 Specifically: the light source control module is integrated on a computer control platform by constructing a light source combination matrix L and a light source control device, and sets the light source state l, the pitch angle A and the azimuth angle B of the shooting light source for control. The light source control device includes a digital signal processing DSP, a pulse width modulation device, an automatic control motor and an automatic rotating bracket;

[0088] The light source is turned on and off by a digital signal processor DSP, the brightness of the light source is adjusted by a pulse width modulation device, the light source automatic control motor is used to change the pitch angle of the light source, and the automatic rotating bracket is used to adjust the azimuth angle of the light source;

[0089] The matrix form of the light source combination matrix L is as follows:

[0090] ;

[0091] In the formula, l ij Indicates the light source status of the jth light source in the i-th group. The specific value of l is binary 0 and 1. 0 means the light source is off, and 1 means the light source is on. ij represents the pitch angle of the jth light source in the i-th group, B ij It represents the azimuth angle of the jth light source in the i-th group. The meaning of the formula is: the light source combination matrix L is used to accurately control the combination and configuration of light sources, and send instructions to the light source control device on the computer control platform to adjust the light source.

[0092] In this embodiment, the light source control module of the system realizes precise control of the light source state, pitch angle and azimuth angle during the LED chip detection process by constructing the integration of the light source combination matrix L and the light source control device. The module uses a digital signal processor DSP to control the opening and closing of the light source, uses a pulse width modulation device to adjust the brightness of the light source, and flexibly adjusts the angle of the light source through automatic control of the motor and the rotating bracket. The application of the light source combination matrix L enables the system to flexibly set the light source at multiple angles and directions, thereby capturing reflection information at more angles. Compared with the traditional single-angle light source configuration method, the module has achieved significant improvements in the accuracy and flexibility of light source control, enabling the system to cover the surface of the LED chip more comprehensively, significantly improving the accuracy of detection and the ability to capture details, thereby greatly reducing the problems of missed detection or false detection caused by insufficient light or poor angles, and improving the overall detection effect and reliability.

[0093] Example 3: This example is an explanation of Example 2. Figure 1 ,Specifically: the image acquisition module includes an image acquisition unit and an image preprocessing unit;

[0094] The image acquisition unit works synchronously with the light source control device by using an industrial camera with a polarization filter, and sets a software trigger mechanism for the industrial camera. The trigger mechanism triggers the industrial camera to shoot when the light source is adjusted successively through the light source combination matrix L, and captures the appearance image of the LED chip under different light source angles in real time;

[0095] The image preprocessing unit is integrated with the industrial camera through the WIFI wireless transmission protocol, and the appearance image of the LED chip captured in real time is transmitted to the central processor, and the appearance image of the LED chip is preprocessed, and then the preprocessed appearance image of the LED chip is summarized to obtain a standard appearance picture set. The preprocessing method includes image correction and image noise removal;

[0096] Image correction includes perspective correction, radiometric distortion correction and image registration to eliminate the collective distortion and deviation caused by the same light source angle and industrial camera position;

[0097] Image noise removal is performed by replacing each pixel on the horizontal axis x and vertical axis y in the image with the average value of all pixels in the field, and smoothing the image to remove random noise in the LED chip appearance image.

[0098] In this embodiment, the system uses an industrial camera with a polarization filter through an image acquisition unit, works synchronously with the light source control device, and adjusts the light source combination matrix L one by one through a software trigger mechanism to capture the LED chip images under different light source angles in real time. The image preprocessing unit transmits these images to the central processor through WIFI wireless transmission, and performs preprocessing such as perspective correction, radiation distortion correction, image registration and noise removal, eliminating the distortion and noise caused by the light source angle and camera position, and finally generates a high-quality standard appearance picture set. This module reduces the distortion and random noise in the image, thereby improving the system's ability to detect tiny defects and making the overall detection process more efficient and reliable.

[0099] Example 4: This example is an explanation of Example 3. Figure 1 ,Specifically: the feature extraction and fusion module includes a feature extraction unit, a ,feature calculation unit and a feature fusion unit;

[0100] The feature extraction unit is used to extract features from the standard image set obtained after preprocessing to obtain a picture feature set;

[0101] The image feature set includes the reflected light intensity Iref, the incident light intensity Iinc, the electric field component Ex in the horizontal axis direction, and the electric field component Ey in the vertical axis direction;

[0102] The reflected light intensity Iref is extracted by converting the RGB values ​​of the pixels of the standard image set into grayscale values ​​for analysis. The specific extraction method is as follows: , where I(x, y) represents the brightness value at the pixel point (x, y);

[0103] The incident light intensity Iinc is obtained by extracting the features of the light source power P and the area A of the chip surface irradiated by the light. The specific extraction method is: , where the light source power P is obtained from the specifications of the light source equipment, and the area A of the chip surface irradiated by the light is obtained by the camera imaging method;

[0104] The electric field component Ex in the horizontal axis and the electric field component Ey in the vertical axis are captured by polarization filters to obtain the polarized light intensity of the standard image set, and then feature extraction is performed. The specific extraction method , where Ipz(x, y) represents the polarized light intensity measured at the pixel point (x, y) of the standard image set;

[0105] The feature extraction unit includes a multi-angle reflection analysis unit, a local surface texture analysis unit and a light source polarization analysis unit;

[0106] The multi-angle reflection analysis unit is used to construct a multi-angle reflection algorithm formula, extract the reflected light intensity Iref and the incident light intensity Iinc in the image feature set and input them into the multi-angle reflection algorithm formula to calculate and output the multi-angle reflection index R (A, B);

[0107] The multi-angle reflection index R (A, B) is calculated and obtained by the following multi-angle reflection algorithm formula;

[0108] ;

[0109] Where Iref(A, B) represents the intensity of reflected light at angle (A, B), and Iinc(A, B) represents the intensity of incident light at angle (A, B).

[0110] The local surface texture analysis unit constructs a local texture calculation formula by using the variance of the second-order differential of the image, and calculates and outputs the local surface texture parameters T (x, y) of the standard image set;

[0111] The local surface texture parameter T (x, y) is calculated by the following local texture calculation formula;

[0112] ;

[0113] Where Var represents the variance of the second-order differential of the image.

[0114] The light source polarization analysis unit constructs the light source polarization angle algorithm formula, extracts the horizontal axis electric field component Ex and the vertical axis electric field component Ey in the image feature set, inputs them into the light source polarization angle algorithm formula, and calculates and outputs the light source polarization angle P (A, B);

[0115] The light source polarization angle P (A, B) is obtained by the following light source polarization angle calculation formula;

[0116] ;

[0117] Wherein, Ey(A, B) represents the component of the electric field in the horizontal axis x direction in the light determined by the angle (A, B), Ex(A, B) represents the component of the electric field in the vertical axis y direction in the light determined by the angle (A, B), and arctan represents the inverse tangent function;

[0118] The feature fusion unit is used to comprehensively calculate the acquired multi-angle reflection index R (A, B), local surface texture parameter T (x, y) and light source polarization angle P (A, B) to obtain a fused image Ifused;

[0119] The fused image Ifused is obtained by calculating the following algorithm formula;

[0120] ;

[0121] Where M represents the total number of fusion items, representing multi-angle reflection characteristics, multi-position texture parameters and multi-polarization angles in the fusion process, m represents the fusion item index, Am represents the pitch angle of the light source under the m-th measurement condition, Bm represents the azimuth angle of the light source under the m-th measurement condition, xm represents the horizontal axis pixel coordinate in the image under the m-th measurement condition, and ym represents the vertical axis pixel coordinate in the image under the m-th measurement condition.

[0122] In this embodiment, the system extracts the reflected light intensity Iref, the incident light intensity Iinc, the electric field component Ex in the horizontal axis direction, and the electric field component Ey in the vertical axis direction from the preprocessed standard image set through the feature extraction unit, and performs in-depth calculation and processing of these key features through multi-angle reflection analysis, local surface texture analysis, and light source polarization analysis. Finally, the feature fusion unit comprehensively calculates the multi-angle reflection index R (A, B), the local surface texture parameter T (x, y), and the light source polarization angle P (A, B) to generate a fused image Ifused. This module has achieved significant improvements in the accuracy of feature extraction and the complexity of the fusion algorithm. Through multi-dimensional and multi-angle feature comprehensive analysis, the system can more accurately capture subtle defects and complex texture structures on the surface of the LED chip, thereby greatly improving the accuracy and reliability of detection, and providing a solid data foundation for subsequent defect identification and quality control.

[0123] Example 5: This example is an explanation of Example 4. Figure 1 ,Specifically: the defect recognition module includes an edge detection unit, a texture analysis unit and a reflection anomaly ,detection unit;

[0124] The edge detection unit calculates the edge strength of the fused image Ifused by using the Canny detection algorithm, and constructs an edge detection score algorithm formula. The detected edge strength is input into the edge detection score algorithm formula to perform edge score calculation to obtain the edge detection score Sedge.

[0125] The edge detection score Sedge is calculated by the following algorithm formula;

[0126] ;

[0127] Where N represents the number of edge points, Gx represents the gradient of the fused image in the horizontal axis x direction, Gy represents the gradient of the fused image in the vertical axis y direction, EdgeRegion represents the edge region in the fused image, Indicates texture intensity;

[0128] The texture analysis unit extracts the energy coefficients of the fused image Ifused at different scales by using wavelet transform, extracts the contrast C, entropy E and texture uniformity H parameters by using the gray level co-occurrence matrix GLCM, and constructs a texture score algorithm formula. The contrast C, entropy E and uniform texture H are input into the texture score algorithm formula to calculate and output the texture analysis score Stexture.

[0129] The texture analysis score Stexture is calculated by the following texture score algorithm formula;

[0130] ;

[0131] In the formula, K represents the total number of detection areas, C k represents the contrast of the kth detection area, E k represents the entropy of the kth detection area, H k represents the texture uniformity of the kth detection area;

[0132] The reflection anomaly detection unit is used to construct a reflection score algorithm formula, and then input the obtained multi-angle reflection index R (A, B) into the reflection score algorithm formula to calculate and output the reflection anomaly detection score Sreflection;

[0133] The texture analysis score is calculated using the following reflectance score algorithm formula;

[0134] ;

[0135] Where Rre(A, B) represents the reference multi-angle reflectance index, represents the reflection characteristic deviation of the kth detection area.

[0136] In this embodiment, the system uses the Canny detection algorithm to analyze the edge strength of the fused image Ifused through the edge detection unit, and accurately locates linear defects such as cracks and scratches on the chip surface by calculating the edge score Sedge. The texture analysis unit extracts parameters such as image contrast, entropy and texture uniformity through wavelet transform and gray level co-occurrence matrix GLCM, calculates the texture analysis score Stexture, and effectively identifies the roughness and unevenness of the chip surface. The reflection anomaly detection unit calculates the reflection anomaly detection score Sreflection by analyzing the deviation of the multi-angle reflection index R (A, B) from the reference value, thereby identifying the area with abnormal reflection characteristics. This module has achieved significant improvements in the accuracy, depth and comprehensiveness of defect detection. Through multi-dimensional feature analysis and score calculation, the system can more comprehensively capture various subtle defects on the chip surface, significantly improve the accuracy and reliability of detection, and ensure high standards of product quality.

[0137] Example 6: This example is an explanation of Example 1. Figure 1 ,Specifically: the comprehensive detection and analysis module includes a comprehensive score calculation unit and a comprehensive ,evaluation unit;

[0138] The comprehensive score calculation unit is used to comprehensively calculate the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection to obtain a comprehensive score Scom;

[0139] The comprehensive score Scom is calculated by the following algorithm formula;

[0140] ;

[0141] Wherein, a1, a2 and a3 represent the preset weight values ​​of the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection, respectively, and their specific values ​​are set by the user.

[0142] The comprehensive evaluation unit compares and evaluates the qualified threshold T set by the user with the comprehensive score Scom, analyzes the appearance of the LED chip for rapid qualification detection, and generates intelligent early warning information. The specific evaluation contents are as follows;

[0143] When the comprehensive score Scom is less than the qualified threshold T, it means that the test result is qualified and no warning information needs to be generated;

[0144] When the comprehensive score Scom ≥ the qualified threshold T, it means that the test result is unqualified. At this time, an early warning message is generated to prompt that the appearance of the LED chip currently being tested is abnormal.

[0145] In this embodiment, the comprehensive detection and analysis module of the system realizes a comprehensive evaluation of the surface quality of the LED chip by integrating the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection. The comprehensive score calculation unit generates a comprehensive score Scom by weighted calculation of these scores to ensure that the detection results can reflect the overall status of various features on the chip surface. The comprehensive evaluation unit quickly judges the qualification of the LED chip by comparing the comprehensive score Scom with the preset qualified threshold T, and instantly generates intelligent early warning information when an abnormality is found, indicating the existence of potential quality problems. Compared with the traditional single indicator detection method, this module has achieved significant improvements in the comprehensiveness, accuracy and intelligence level of the evaluation. Through the flexible setting of weights, the system can adjust the detection focus according to specific needs to ensure that the detection results are more in line with the actual application scenario. In the end, this module not only improves the accuracy and reliability of detection, but also greatly reduces the production risks caused by quality problems through the intelligent early warning function, laying a solid foundation for realizing efficient and automated production detection processes.

[0146] Example 7: Please refer to Figure 1 and Figure 2 , an intelligent and rapid detection method for the appearance of an LED chip, comprising the following steps:

[0147] S1. By constructing a light source combination matrix L on a computer light source control platform and integrating it with a light source control device, the shooting light source is set at multiple angles and in multiple directions;

[0148] S2. According to the light source combination matrix L, the appearance images of the LED chips under different light source combinations are collected one by one, and the collected appearance images of the LED chips are preprocessed to obtain a standard appearance picture set;

[0149] S3, extract features from the annotated appearance picture set to obtain a picture feature set, then construct a related algorithm formula, extract the picture feature set and input it into the related algorithm formula, output the multi-angle reflection index R (A, B), the local surface texture parameter T (x, y) and the light source polarization angle P (A, B), and then perform comprehensive calculation on the multi-angle reflection index R (A, B), the local surface texture parameter T (x, y) and the light source polarization angle P (A, B) to obtain a fused image Ifused;

[0150] S4, construct a related defect algorithm formula, input the fused image Ifused into the related defect algorithm formula, and output the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection;

[0151] S5. Comprehensively calculate the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection to obtain a comprehensive score Scom, and compare the preset qualified threshold T with the comprehensive score Scom to evaluate and analyze the detection situation.

[0152] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. An intelligent rapid detection system for LED chip appearance, characterized by: It includes light source control module, image acquisition module, feature extraction and fusion module, defect recognition module and comprehensive detection and analysis module; The light source control module controls the light source control device by constructing a light source combination matrix L on a computer light source control platform, and sets the shooting light source in multiple angles and directions; The light source control module constructs a light source combination matrix L, and then controls the light source state l, pitch angle A and azimuth angle B of the light source control device on a computer control platform according to the light source combination matrix L. The light source control device includes a digital signal processing DSP, a pulse width modulation device, an automatic control motor and an automatic rotating bracket; The matrix form of the light source combination matrix L is as follows: ； In the formula, l ij Indicates the light source status of the jth light source in the i-th group. The specific value of l is binary 0 and 1. 0 means the light source is off, and 1 means the light source is on. ij represents the pitch angle of the jth light source in the i-th group, B ij represents the azimuth angle of the jth light source in the i-th group; The image acquisition module is used to successively acquire the appearance images of the LED chips under different light source combinations according to the light source combination matrix L, and pre-process the acquired appearance images of the LED chips to obtain a standard appearance picture set; The image acquisition module includes an image acquisition unit and an image preprocessing unit; The image acquisition unit works synchronously with the light source control device by using an industrial camera with a polarization filter, and sets a software trigger mechanism for the industrial camera. The trigger mechanism triggers the industrial camera to shoot when the light source is adjusted successively by the light source combination matrix L, and captures the appearance image of the LED chip under different light source angles in real time; The image preprocessing unit is integrated with the industrial camera through the WIFI wireless transmission protocol, and the appearance image of the LED chip captured and photographed in real time is transmitted to the central processor, and the appearance image of the LED chip is preprocessed, and then the preprocessed appearance image of the LED chip is summarized to obtain a standard appearance picture set, and the preprocessing method includes image correction and image noise removal; The image correction includes perspective correction, radiometric distortion correction and image registration to eliminate the collective distortion and deviation caused by the same light source angle and industrial camera position; The image noise removal is performed by replacing each pixel of the horizontal axis x and the vertical axis y in the image with the average value of all pixels in the field, and smoothing the image to remove random noise in the appearance image of the LED chip; The feature extraction and fusion module is used to extract features from the annotated appearance picture set, obtain a picture feature set, and then construct a related algorithm formula, extract the picture feature set and input it into the related algorithm formula, output a multi-angle reflection index R (A, B), a local surface texture parameter T (x, y) and a light source polarization angle P (A, B), and then perform a comprehensive calculation on the multi-angle reflection index R (A, B), the local surface texture parameter T (x, y) and the light source polarization angle P (A, B) to obtain a fused image Ifused; The feature extraction and fusion module includes a feature extraction unit, a feature calculation unit and a feature fusion unit; The feature extraction unit is used to extract features from the standard image set obtained after preprocessing to obtain a picture feature set; The image feature set includes reflected light intensity Iref, incident light intensity Iinc, horizontal axis electric field component Ex, and vertical axis electric field component Ey; The reflected light intensity Iref is extracted by converting the RGB values ​​of the pixels of the standard image set into gray values ​​for analysis. The specific extraction method is as follows: , where I(x, y) represents the brightness value at the pixel point (x, y); The incident light intensity Iinc is obtained by extracting the characteristics of the light source power P and the area D on the chip surface irradiated by the light. The specific extraction method is: , where the light source power P is obtained from the specifications of the light source equipment, and the area D of the chip surface irradiated by the light is obtained by the camera imaging method; The horizontal electric field component Ex and the vertical electric field component Ey are captured by a polarization filter to obtain the polarized light intensity of the standard image set, and then feature extraction is performed. The specific extraction method is: , where Ipz(x, y) represents the polarized light intensity measured at the pixel point (x, y) of the standard image set; The feature extraction unit includes a multi-angle reflection analysis unit, a local surface texture analysis unit and a light source polarization analysis unit; The multi-angle reflection analysis unit is used to construct a multi-angle reflection algorithm formula, extract the reflected light intensity Iref and the incident light intensity Iinc in the image feature set and input them into the multi-angle reflection algorithm formula, calculate and output the multi-angle reflection index R (A, B); The multi-angle reflection index R (A, B) is calculated and obtained by the following multi-angle reflection algorithm formula; ； Where, Iref(A, B) represents the intensity of reflected light at angle (A, B), and Iinc(A, B) represents the intensity of incident light at angle (A, B); The defect recognition module constructs a related defect algorithm formula, inputs the fused image Ifused into the related defect algorithm formula, and outputs an edge detection score Sedge, a texture analysis score Stexture, and a reflection anomaly detection score Sreflection; The comprehensive detection and analysis module obtains a comprehensive score Scom by comprehensively calculating the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection, and compares the preset qualified threshold T with the comprehensive score Scom to evaluate and analyze the detection situation.

2. According to claim 1, the LED chip appearance intelligent rapid detection system is characterized by: The local surface texture analysis unit constructs a local texture calculation formula by using the variance of the second-order differential of the image, calculates and outputs the local surface texture parameters T (x, y) of the standard image set; The local surface texture parameter T (x, y) is calculated and obtained by the following local texture calculation formula; ； Where Var represents the variance of the second-order differential of the image.

3. The LED chip appearance intelligent rapid detection system according to claim 1, characterized in that: The light source polarization analysis unit constructs a light source polarization angle algorithm formula, extracts the horizontal axis electric field component Ex and the vertical axis electric field component Ey in the image feature set, inputs them into the light source polarization angle algorithm formula, calculates and outputs the light source polarization angle P (A, B); The light source polarization angle P (A, B) is obtained by the following light source polarization angle calculation formula; ； Wherein, Ey(A, B) represents the component of the electric field in the horizontal axis x direction in the light determined by the angle (A, B), Ex(A, B) represents the component of the electric field in the vertical axis y direction in the light determined by the angle (A, B), and arctan represents the inverse tangent function; The feature fusion unit is used to comprehensively calculate the acquired multi-angle reflection index R (A, B), local surface texture parameter T (x, y) and light source polarization angle P (A, B) to obtain a fused image Ifused; The fused image Ifused is obtained by calculation using the following algorithm formula; ； Where M represents the total number of fusion items, which represents multi-angle reflection characteristics, multi-position texture parameters and multi-polarization angles in the fusion process, m represents the fusion item index, and A m represents the elevation angle of the light source under the mth measurement condition, B m represents the azimuth angle of the light source under the mth measurement condition, x m Indicates the horizontal pixel coordinate in the image under the mth measurement condition, y m Represents the vertical pixel coordinates in the image under the mth measurement condition.

4. The LED chip appearance intelligent rapid detection system according to claim 3, characterized in that: The defect recognition module includes an edge detection unit, a texture analysis unit and a reflection anomaly detection unit; The edge detection unit calculates the edge strength of the fused image Ifused by using the Canny detection algorithm, constructs an edge detection score algorithm formula, inputs the detected edge strength into the edge detection score algorithm formula, performs edge score calculation to obtain an edge detection score Sedge; The edge detection score Sedge is calculated by the following algorithm formula: ； In the formula, N represents the number of edge points, G x Represents the gradient of the fused image in the horizontal x direction, G y It represents the gradient of the fused image in the y direction of the vertical axis, and EdgeRegion represents the edge area in the fused image. Indicates texture intensity; The texture analysis unit extracts the energy coefficients of the fused image Ifused at different scales by using wavelet transform, extracts the contrast C, entropy E and texture uniformity H parameters by using the gray level co-occurrence matrix GLCM, and constructs a texture score algorithm formula, inputs the contrast C, entropy E and uniform texture H into the texture score algorithm formula, and calculates and outputs the texture analysis score Stexture; The texture analysis score Stexture is calculated and obtained by the following texture score algorithm formula; ； In the formula, K represents the total number of detection areas, C k represents the contrast of the kth detection area, E k represents the entropy of the kth detection area, H k represents the texture uniformity of the kth detection area; The reflection anomaly detection unit is used to construct a reflection score algorithm formula, and then input the obtained multi-angle reflection index R (A, B) into the reflection score algorithm formula to calculate and output the reflection anomaly detection score Sreflection; The reflection anomaly detection score Sreflection is calculated and obtained by the following reflection score algorithm formula; ； Where Rre(A, B) represents the reference multi-angle reflectance index, represents the reflection characteristic deviation of the kth detection area.

5. The LED chip appearance intelligent rapid detection system according to claim 1, characterized in that: The comprehensive detection and analysis module includes a comprehensive score calculation unit and a comprehensive evaluation unit; The comprehensive score calculation unit is used to comprehensively calculate the edge detection score Sedge, the texture analysis score Stexture and the reflection abnormality detection score Sreflection to obtain a comprehensive score Scom; The comprehensive score Scom is calculated by the following algorithm formula: ； Wherein, a1, a2 and a3 represent the preset weight values ​​of the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection, respectively, and their specific values ​​are set by the user.

6. The LED chip appearance intelligent rapid detection system according to claim 5, characterized in that: The comprehensive evaluation unit compares and evaluates the qualified threshold T set by the user with the comprehensive score Scom, analyzes the appearance of the LED chip to quickly detect the qualification, and generates intelligent early warning information. The specific evaluation content is as follows; When the comprehensive score Scom is less than the qualified threshold T, it means that the test result is qualified and no warning information needs to be generated; When the comprehensive score Scom ≥ the qualified threshold T, it means that the test result is unqualified. At this time, an early warning message is generated to prompt that the appearance of the LED chip currently being tested is abnormal.

7. An intelligent and rapid detection method for the appearance of an LED chip, applied to an intelligent and rapid detection system for the appearance of an LED chip according to any one of claims 1 to 6, characterized in that: The following steps are involved: S1. By constructing a light source combination matrix L on a computer light source control platform and integrating it with a light source control device, the shooting light source is set at multiple angles and in multiple directions; S2. According to the light source combination matrix L, the appearance images of the LED chips under different light source combinations are collected one by one, and the collected appearance images of the LED chips are preprocessed to obtain a standard appearance picture set; S3, extract features from the annotated appearance picture set to obtain a picture feature set, then construct a related algorithm formula, extract the picture feature set and input it into the related algorithm formula, output the multi-angle reflection index R (A, B), the local surface texture parameter T (x, y) and the light source polarization angle P (A, B), and then perform comprehensive calculation on the multi-angle reflection index R (A, B), the local surface texture parameter T (x, y) and the light source polarization angle P (A, B) to obtain a fused image Ifused; S4, construct a related defect algorithm formula, input the fused image Ifused into the related defect algorithm formula, and output the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection; S5. Comprehensively calculate the edge detection score Sedge, the texture analysis score Stexture and the reflection anomaly detection score Sreflection to obtain a comprehensive score Scom, and compare the preset qualified threshold T with the comprehensive score Scom to evaluate and analyze the detection situation.

Images