Multi-parameter semiconductor defect testing method and system

Abstract

The invention relates to the technical field of semiconductors, in particular to a multi-parameter semiconductor defect testing method and system. The method comprises the following steps: firstly, constructing an LED test data set by collecting image data, electrical data, thermal data and mechanical data of an LED; secondly, respectively calculating an LED image test score, an electrical test score, a thermal test score and a mechanical test score according to the test data set; and finally, according to the LED image test score, the LED electrical test score, the LED thermal test score and the LED mechanical test score, generating an LED defect test score through a multi-parameter defect test, and dividing the LED defects into different defect grades. Through the multi-parameter defect testing method, the LED defect grade can be accurately obtained.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor technology, and in particular to a multi-parameter semiconductor defect testing method and system. Background Art

[0002] The multi-parameter detection method is a detection technology that has gradually emerged in the semiconductor field in recent years. It collects multiple physical quantities, such as electrical, optical and mechanical, at the same time to conduct a comprehensive analysis of the object being tested, and understand the performance of materials and devices more comprehensively and accurately. The multi-parameter detection method can make up for the limitations of single parameter detection, improve the sensitivity and reliability of detection, and is suitable for the characterization and diagnosis of complex systems.

[0003] During the manufacturing process of semiconductor devices, various defects are prone to occur, such as lattice dislocations, impurities, and interface traps, which can significantly affect the electrical performance and reliability of the device. Efficient and accurate detection and characterization of these defects are essential to improving the performance and reliability of semiconductor devices. Through effective defect detection, problems that occur during the manufacturing process can be discovered and corrected in a timely manner, preventing defective products from entering the market, thereby improving product quality and production efficiency.

[0004] Light-emitting diode (LED) is an important semiconductor device, and its working principle is to emit light through the combination of electrons and holes in semiconductor materials. Due to its advantages such as high efficiency, long life and environmental friendliness, LED is widely used in lighting, display, signal indication and other fields. The performance of LED directly affects its application effect and life, so in the manufacturing process, strict control of material purity, crystal quality and interface characteristics is very important. Most of the existing LED quality detection is mainly based on single parameter detection such as electrical characteristics, spectral characteristics and surface defects. For example, the Chinese patent with publication number CN110319925B proposes an LED detection device. The invention connects or separates the electrode of the LED with the test end to power on or off the LED, thereby alleviating the technical problems of inconvenient use and low detection efficiency of the detection device in the prior art. In addition, the Chinese patent with publication number CN108391343B provides an LED detection device, a detection system and a detection method, which detects the induced current of each of the inductive resistors and determines the backlight working state of the corresponding LED module according to the induced current. The above literatures all test LED quality based on a single parameter, which may not fully understand the multi-dimensional impact of defects, resulting in inaccurate LED defect detection results.

[0005] To this end, a multi-parameter semiconductor defect testing method and system are proposed. Summary of the invention

[0006] The object of the present invention is to provide a multi-parameter semiconductor defect testing method and system. First, by collecting the image data, electrical data, thermal data and mechanical data of the LED, an LED test data set is constructed. Secondly, according to the test data set, the LED image test score, electrical test score, thermal test score and mechanical test score are calculated respectively. Then, according to the LED image test score, the LED electrical test score, the LED thermal test score and the LED mechanical test score, the LED defect test score is generated through multi-parameter defect testing, and the LED defects are divided into different defect levels. The present invention can accurately obtain the LED defect level through the multi-parameter defect testing method.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A multi-parameter semiconductor defect testing method, comprising:

[0009] Collect LED image data, electrical data, thermal data and mechanical data to build LED test data set;

[0010] According to the LED test data set, an LED image test score, an LED electrical test score, an LED thermal test score, and an LED mechanical test score are obtained;

[0011] The LED image test score acquisition step includes: inputting the image in the LED image data into the LED image feature test model, acquiring the LED image feature vector, and generating the LED image test score according to the LED image feature vector;

[0012] According to the LED image test score, the LED electrical test score, the LED thermal test score and the LED mechanical test score, an LED defect test score is generated through multi-parameter defect testing; the formula of the LED defect test score is:

[0013] S=α1S im +α2S e +α3S t +α4S m ;

[0014] Among them, S is the LED defect test score, S im Score for LED image test, S e Score the LED electrical test, S t Score for LED thermal test, S m Score the LED mechanical test, α1, α2, α3 and α4 are expressed as defect test weights, α1+α2+α3+α4=1;

[0015] According to the LED defect test score, the LED defect grade is obtained.

[0016] Preferably, the process of generating the LED image test score includes:

[0017] The LED image feature test model includes an LED image input layer, an LED image preprocessing layer, an LED image feature extraction layer, an LED image defect recognition layer and an LED image test scoring data output layer;

[0018] The LED image input layer is used to input the image in the LED image data into the LED image feature test model; the LED image preprocessing layer is used to grayscale, denoise and contrast enhance the image in the LED image data; the LED image feature extraction layer is used to extract features from the preprocessed image, and the specific steps include:

[0019] The processed image is converted into a grayscale image, and a grayscale co-occurrence matrix is ​​calculated to generate a matrix describing the relationship between grayscale values ​​of adjacent pixels in the image; the image texture features are obtained by calculating the combination frequency of grayscale values ​​of adjacent pixels on the grayscale co-occurrence matrix; the image contrast features are obtained by performing weighted calculation on each element in the grayscale co-occurrence matrix to obtain the defect area; the outline of the defect area in the image is extracted by an edge detection algorithm, and the area and perimeter of the defect area are calculated based on the outline to evaluate the shape features of the defect area; the image texture features, the contrast features and the shape features are normalized to obtain the LED image feature vector;

[0020] The LED image defect recognition layer is used to input the features extracted from the LED image feature extraction layer into the trained defect test model to perform defect recognition and obtain a defect image;

[0021] The LED image test score data output layer outputs the LED image test score.

[0022] Preferably, the LED image test score is obtained according to the defect image, and the calculation formula is:

[0023]

[0024] Among them, S im is the LED image test score, N is the number of LED images to be tested, M is the number of elements in the LED image feature vector, S con,i is the contrast feature score of the i-th defect area, ω con is the contrast feature weight, S cir,i is the roundness characteristic score of the i-th defect area, ω cir is the roundness feature weight, Ai is the area of ​​the i-th defect region, A i,max is the area of ​​the i-th defect region, V j is the jth eigenvector influencing factor.

[0025] Preferably, the process of generating the LED electrical test score includes:

[0026] Collecting electrical data of the area to be tested in the LED, specifically including:

[0027] Connect the LED to a power source and a current source, and connect a digital multimeter, an oscilloscope, and a resistance tester to the test points of the LED;

[0028] The forward voltage, forward current, reverse voltage and reverse current of the LED are collected by the digital multimeter and the oscilloscope; the electrical signal waveform of the LED is tested by the oscilloscope, wherein the electrical signal waveform includes a switching response time and a noise dynamic characteristic;

[0029] Constructing a feature vector group through the collected data; normalizing the feature vector group to obtain an LED electrical feature vector group; the LED electrical feature vector group includes the forward voltage, the reverse voltage, the forward current, the reverse current, the forward resistance, the reverse resistance, the response time and the noise;

[0030] The LED electrical test score is calculated according to the LED electrical characteristic vector group, and the calculation formula is:

[0031]

[0032] Among them, S e Rating for LED electrical test, V f is the forward voltage, I f is the forward current, R is the standard resistance, V r is the reverse voltage, I r is the reverse current, κ is the noise factor, κ max is the maximum noise factor, T res is the switch response time, T res,max is the maximum switch response time, ω1, ω2, ω3, ω4 are weight coefficients, ω1+ω2+ω3+ω4=1.

[0033] Preferably, the process of generating the LED thermal test score includes:

[0034] Collecting thermal data of the area to be tested in the LED, specifically including:

[0035] Capturing thermal images of the LED surface using a thermal imaging camera to obtain temperature differences in different areas and existing local hot spots, and generating a temperature distribution map of the LED; extracting thermal characteristics of the LED by analyzing the temperature distribution in the image;

[0036] Normalizing the features to construct a LED thermal feature vector group; the LED thermal feature vector group includes average operating temperature, temperature rise rate, temperature uniformity and thermal conductivity estimation;

[0037] The LED thermal test score is calculated according to the LED thermal characteristic vector group, and the calculation formula is:

[0038]

[0039] Among them, S t Rating for LED thermal test, T avg is the average operating temperature, ΔT rise is the temperature rise rate, HS is the total number of hot spots on the LED surface that are significantly higher than the surrounding area, is the reciprocal of thermal conductivity, β1, β2, β3, β4 are weight coefficients, β1+β2+β3+β4=1.

[0040] Preferably, the process of generating the LED mechanical test score includes:

[0041] Collect mechanical data of the area to be tested in the LED, including:

[0042] By installing the LED on a vibration table and collecting response data at different frequencies and amplitudes, a first mechanical characteristic including a frequency response value and a vibration amplitude value is obtained; by installing the LED on an impact testing machine and collecting response data after impact, a second mechanical characteristic including a maximum displacement response and an impact energy value is obtained; by installing the lead of the LED on a tensile testing machine and collecting force and displacement data, a third mechanical characteristic including a stress value and a strain value is obtained;

[0043] Normalizing the first mechanical feature, the second mechanical feature, and the third mechanical feature, and merging them into an LED mechanical feature vector group;

[0044] The LED mechanical test score is calculated according to the LED mechanical feature vector group, and the calculation formula is:

[0045]

[0046] Among them, S m Scoring LED Mechanical Tests, ¶ j is the weight of the jth feature, xj is the jth eigenvector value, p ji is the element in the i-th row and j-th column of the principal component matrix P, σ j is the standard deviation of the jth feature;

[0047] The specific calculation method of the principal component matrix P is:

[0048]

[0049] Where n is the total number of features, X i is the eigenvector of the i-th feature, X is the eigenmean, Λ is the eigenvalue diagonal matrix, for The transposed matrix of .

[0050] Preferably, the LED defect levels include:

[0051] Set the defect test level threshold set TH={S1, S2, S3}. When the LED defect test score is less than S1, the corresponding LED defect level is level one; when the LED defect test score is in the interval [S1, S2), the corresponding LED defect level is level two; when the LED defect test score is in the interval [S2, S3], the corresponding LED defect level is level three; when the LED defect test score is greater than S3, the corresponding LED defect level is level four; the lower the level of the LED defect test score, the fewer LED defects; wherein the range of the LED defect test score S is [0, 1], and S1, S2 and S3 are 0.25, 0.5, and 0.75, respectively.

[0052] A multi-parameter semiconductor defect testing system, comprising:

[0053] Data acquisition unit: used to collect LED image data, electrical data, thermal data and mechanical data;

[0054] Defect test score generating unit: used to construct a corresponding feature vector group according to the collected features, and obtain the LED image test score, LED electrical test score, LED thermal test score and LED mechanical test score;

[0055] An LED defect test score generating unit: configured to generate an LED defect test score through a multi-parameter defect test according to the LED image test score, the LED electrical test score, the LED thermal test score and the LED mechanical test score;

[0056] LED defect test score level generating unit: used for obtaining the LED defect level according to the LED defect test score.

[0057] Preferably, the LED defect levels include:

[0058] Set the defect test level threshold set TH={S1, S2, S3}. When the LED defect test score is less than S1, the corresponding LED defect level is level one; when the LED defect test score is in the interval [S1, S2), the corresponding LED defect level is level two; when the LED defect test score is in the interval [S2, S3], the corresponding LED defect level is level three; when the LED defect test score is greater than S3, the corresponding LED defect level is level four; the lower the level of the LED defect test score, the fewer LED defects; wherein the range of the LED defect test score S is [0, 1], and S1, S2 and S3 are 0.25, 0.5, and 0.75, respectively.

[0059] Preferably, collecting electrical characteristics of the area to be tested in the LED specifically includes:

[0060] Connect the LED to a power source and a current source, and ensure that the connections are secure; connect a digital multimeter, an oscilloscope, and a resistance tester to the test points of the LED;

[0061] The forward voltage, forward current, reverse voltage and reverse current of the LED are collected by the digital multimeter and the oscilloscope; the electrical signal waveform of the LED is tested by the oscilloscope, wherein the electrical signal waveform includes a switching response time and a noise dynamic characteristic;

[0062] Constructing a feature vector group through the collected data; normalizing the feature vector group to obtain an LED electrical feature vector group; the LED electrical feature vector group includes the forward voltage, the reverse voltage, the forward current, the reverse current, the forward resistance, the reverse resistance, the response time and the noise;

[0063] The LED electrical test score is calculated according to the LED electrical characteristic vector group, and the calculation formula is:

[0064]

[0065] Among them, S e Rating for LED electrical test, V f is the forward voltage, I f is the forward current, R is the standard resistance, V r is the reverse voltage, I r is the reverse current, κ is the noise factor, κ max is the maximum noise factor, T res is the switch response time, T res,maxis the maximum switch response time, ω1, ω2, ω3, ω4 are weight coefficients, ω1+ω2+ω3+ω4=1.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] 1. The present invention inputs the collected LED image into the LED image feature test model, obtains the image texture features, image contrast features, defect areas and defect area shapes, and constructs the LED image feature vector. According to the obtained LED image feature vector, a defect image is obtained. Combining the feature vector of the LED image with the area and perimeter of the defect area, the LED image test score is obtained. This method can obtain the LED defect test score more accurately by considering the area of ​​the defect area, providing a data basis for the subsequent LED defect grade test.

[0068] 2. The present invention considers the influence of mechanical characteristics on defects when detecting LED defects; by obtaining the frequency response value, vibration amplitude value, maximum displacement response, impact energy value, stress value and strain value of the LED, the LED mechanical test score is obtained. This method integrates the performance of LED under different mechanical characteristics, helps to identify and evaluate potential defects, and provides a data basis for subsequent LED defect level testing.

[0069] 3. When calculating the LED defect grade, the present invention calculates the LED image test score, LED electrical test score, LED thermal test score and LED mechanical test score based on the LED image data, electrical data, thermal data and mechanical data, and obtains the LED defect test score through multi-parameter defect testing. This method comprehensively considers information from different modes of LED and can calculate LED defect results more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A schematic flow chart of a multi-parameter semiconductor defect testing method provided by an embodiment of the present invention;

[0071] Figure 2 A schematic diagram of the structure of a multi-parameter semiconductor defect testing system provided by an embodiment of the present invention;

[0072] Figure 3 A schematic diagram of the structure of a defect test score generating unit provided in an embodiment of the present invention;

[0073] Figure 4 A schematic diagram of the structure of an LED image feature test model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0074] 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.

[0075] The present invention provides a multi-parameter semiconductor defect testing method, which is applied to a multi-parameter semiconductor defect testing system. The specific method and system flowchart refer to Figure 1 and Figure 2 .

[0076] Embodiment 1

[0077] As an embodiment of the present invention, refer to Figure 1 S10 in the embodiment is applied to a data acquisition unit of a multi-parameter semiconductor defect testing system. In this embodiment, image data, electrical data, thermal data and mechanical data of an LED to be tested are collected to construct an LED test data set.

[0078] Furthermore, the LED test data set includes an LED image subset, and the LED image subset includes an LED image to be tested taken by a camera; the electrical data includes forward voltage, reverse voltage, forward current, leakage current, power and photoelectric conversion efficiency; the thermal data includes junction temperature, thermal resistance, heat capacity, thermal conductivity, steady-state and transient thermal characteristics; the above-mentioned mechanical data includes frequency response value, vibration amplitude value, maximum displacement response, impact energy value, stress value, strain value, Young's modulus value and fatigue life.

[0079] Reference Figure 1 S20 in the example is applied to a defect test score generation unit of a multi-parameter semiconductor defect test system, referring to Figure 3 The defect test score generating unit includes a plurality of modules. In this step, the LED image test score, the LED electrical test score, the LED thermal test score and the LED mechanical test score are obtained according to the LED test data set.

[0080] Furthermore, the LED image test score acquisition step includes: inputting the image in the LED image data into the LED image feature test model, acquiring the LED image feature vector, and generating the LED image test score according to the LED image feature vector. The specific process includes:

[0081] Reference Figure 4, the LED image feature test model includes an LED image input layer, an LED image preprocessing layer, an LED image feature extraction layer, an LED image defect recognition layer and an LED image test scoring data output layer;

[0082] The LED image input layer is used to input the image in the LED image data into the LED image feature test model; the LED image preprocessing layer is used to grayscale, denoise and contrast enhance the image in the LED image data; the LED image feature extraction layer is used to extract features from the preprocessed image, and the specific steps include:

[0083] The processed image is converted into a grayscale image, and a grayscale co-occurrence matrix is ​​calculated to generate a matrix describing the relationship between grayscale values ​​of adjacent pixels in the image; the image texture features are obtained by calculating the combination frequency of grayscale values ​​of adjacent pixels on the grayscale co-occurrence matrix; the image contrast features are obtained by performing weighted calculation on each element in the grayscale co-occurrence matrix to obtain the defect area; the outline of the defect area in the image is extracted by an edge detection algorithm, and the area and perimeter of the defect area are calculated based on the outline to evaluate the shape characteristics of the defect area;

[0084] The LED image defect recognition layer is used to input the features extracted from the LED image feature extraction layer into the trained defect test model for further defect recognition to obtain a defect image;

[0085] The LED image test score data output layer outputs the LED image test score.

[0086] In this embodiment, the LED to be tested is photographed by a microscope camera, and a total of 1000 LED surface images are collected to construct an LED image subset, and the images in the LED image subset are classified by the LED image feature test model. In this embodiment, the texture features, contrast features and shape features of the normal and defective images are obtained by analyzing the LED image subset, as shown in Table 1. In this embodiment, in terms of texture features, the higher uniformity of the normal image indicates that the image is more uniform and consistent, while the lower uniformity of the defective image reflects the uneven image. In the contrast feature, the lower contrast of the normal image shows that the overall contrast of the image is weak, while the higher contrast of the defective image shows that the image contrast is strong. In terms of contrast features, the smaller grayscale variance of the normal image shows that the grayscale value of the image changes less, while the larger grayscale variance of the defective image shows that the grayscale value changes more. The average brightness of the normal image is moderate, while the slightly higher brightness of the defective image shows the difference in brightness. In terms of shape features, the normal image has no defective area, indicating that the image surface has no obvious defects, while the defective image has obvious defective areas. The roundness of the defective area is a complete shape in the normal image, while the irregular shape of the defect is shown in the defective image.

[0087] Table 1 Image feature comparison table

[0088]

[0089] These features are extracted through a microscope camera and an image feature test model, reflecting in detail the differences between normal and defective LED images in texture, contrast, and shape characteristics, which helps to accurately classify and evaluate LED product defects.

[0090] Normalizing the image texture feature, the contrast feature and the shape feature to obtain the LED image feature vector;

[0091] Furthermore, the LED image test score is obtained, and the calculation formula is:

[0092]

[0093] Among them, S im is the LED image test score, N is the number of LED images to be tested, M is the number of elements in the LED image feature vector, S con,i is the contrast feature score of the i-th defect area, ω con is the contrast feature weight, S cir,i is the roundness characteristic score of the i-th defect area, ω cir is the roundness feature weight, A i is the area of ​​the i-th defect region, A i,max is the area of ​​the i-th defect region, V j is the jth eigenvector influencing factor.

[0094] In this embodiment, the defect situation of the LED surface is evaluated by calculating the LED image test score. The score comprehensively considers multiple key features of the image, including contrast features and roundness features, and effectively reflects the degree of defects on the LED surface. For example, the contrast feature score can evaluate the difference in grayscale values ​​in the image, while the roundness feature score can reflect whether the shape of the defect area is regular. Through the comprehensive analysis of these features, the quality of the LED image can be more comprehensively evaluated to ensure the accuracy of the test results.

[0095] Furthermore, the generation process of the LED electrical test score includes:

[0096] Collecting electrical characteristics of the area to be tested in the LED, specifically including:

[0097] Connect the LED to a power source and a current source, and ensure that the connections are secure; connect a digital multimeter, an oscilloscope, and a resistance tester to the test points of the LED;

[0098] The forward voltage, forward current, reverse voltage and reverse current of the LED are collected by the digital multimeter and the oscilloscope; the electrical signal waveform of the LED is tested by the oscilloscope, wherein the electrical signal waveform includes a switching response time and a noise dynamic characteristic;

[0099] Constructing a feature vector group through the collected data; normalizing the feature vector group to obtain an LED electrical feature vector group; the LED electrical feature vector group includes the forward voltage, the reverse voltage, the forward current, the reverse current, the forward resistance, the reverse resistance, the response time and the noise;

[0100] The LED electrical test score is calculated according to the LED electrical characteristic vector group, and the calculation formula is:

[0101]

[0102] Among them, S e Rating for LED electrical test, V f is the forward voltage, I f is the forward current, R is the standard resistance, V r is the reverse voltage, I r is the reverse current, κ is the noise factor, κ max is the maximum noise factor, T res is the switch response time, T res,max is the maximum switch response time, ω1, ω2, ω3, ω4 are weight coefficients, ω1+ω2+ω3+ω4=1.

[0103] In this embodiment, the electrical characteristics of the LED test area, including forward voltage, forward current, reverse voltage, reverse current, forward resistance, reverse resistance, response time and noise, are collected to construct an LED electrical characteristic vector group. The electrical performance of the LED is evaluated through the LED electrical characteristic vector group. At the same time, by considering dynamic characteristics such as switch response time and noise, the characteristics of the LED in practical applications are analyzed. Through a multi-dimensional, multi-feature comprehensive analysis method, the accuracy and reliability of the electrical test score are improved, providing a data basis for subsequent LED defect level testing.

[0104] Furthermore, the process of generating the LED thermal test score includes:

[0105] Collecting the thermal characteristics of the area to be tested in the LED, specifically including:

[0106] Capturing thermal images of the LED surface using a thermal imaging camera to obtain temperature differences in different areas and existing local hot spots, and generating a temperature distribution map of the LED; extracting thermal characteristics of the LED by analyzing the temperature distribution in the image;

[0107] Normalizing the features to construct a LED thermal feature vector group; the LED thermal feature vector group includes average operating temperature, temperature rise rate, temperature uniformity and thermal conductivity estimation;

[0108] The LED thermal test score is calculated according to the LED thermal characteristic vector group, and the calculation formula is:

[0109]

[0110] Among them, S t Rating for LED thermal test, T avg is the average operating temperature, ΔT rise is the temperature rise rate, HS is the total number of hot spots on the LED surface that are significantly higher than the surrounding area, is the reciprocal of thermal conductivity, β1, β2, β3, β4 are weight coefficients, β1+β2+β3+β4=1.

[0111] In this embodiment, the average operating temperature represents the overall thermal performance of the LED, while the heating rate reflects the temperature change of the LED from power-on to the steady state. The temperature uniformity reveals the uniformity of the surface temperature or the presence of local hot spots. By comparing these characteristic values ​​with the design specifications, thermal defects of the LED, such as insufficient heat dissipation or material problems, can be identified. Finally, based on the extracted thermal features and analysis results, the LED is defectively scored to evaluate its thermal performance and determine whether there are thermal defects.

[0112] Furthermore, the process of generating the LED mechanical test score includes:

[0113] By installing the LED on a vibration table and collecting response data at different frequencies and amplitudes, a first mechanical characteristic including a frequency response value and a vibration amplitude value is obtained; by installing the LED on an impact testing machine and collecting response data after impact, a second mechanical characteristic including a maximum displacement response and an impact energy value is obtained; by installing the lead of the LED on a tensile testing machine and collecting force and displacement data, a third mechanical characteristic including a stress value and a strain value is obtained;

[0114] Normalizing the first mechanical feature, the second mechanical feature, and the third mechanical feature, and merging them into an LED mechanical feature vector group;

[0115] The LED mechanical test score is calculated according to the LED mechanical feature vector group, and the calculation formula is:

[0116]

[0117] Among them, Sm Scoring LED Mechanical Tests, ¶ j is the weight of the jth feature, x j is the jth eigenvector value, p ji is the element in the i-th row and j-th column of the principal component matrix P, σ j is the standard deviation of the jth feature;

[0118] The specific calculation method of the principal component matrix P is:

[0119]

[0120] Where n is the total number of features, X i is the eigenvector of the ith feature, is the eigenmean, Λ is the eigenvalue diagonal matrix, for The transposed matrix of .

[0121] In this embodiment, by obtaining the frequency response value, vibration amplitude value, maximum displacement response, impact energy value, stress value and strain value of the LED, the LED mechanical test score is calculated based on these feature vector groups to evaluate the performance of the LED under different mechanical conditions. In particular, the multi-dimensional feature data is converted into main features through the principal component analysis method, which improves the accuracy and reliability of the score. This method integrates the performance of the LED under different mechanical characteristics, helps to identify and evaluate potential defects, and provides a data basis for subsequent LED defect level testing.

[0122] Reference Figure 1 S30 in the embodiment is applied to an LED defect test score generation unit in a multi-parameter semiconductor defect test system.

[0123] Further, according to the LED image test score, the LED electrical test score, the LED thermal test score and the LED mechanical test score, an LED defect test score is generated through multi-parameter defect testing; the formula of the LED defect test score is:

[0124] S=α1S im +α2S e +α3S t +α4S m ;

[0125] Among them, S is the LED defect test score, S im Score for LED image test, S e Score the LED electrical test, S t Score for LED thermal test, S mScore the LED mechanical test, α1, α2, α3 and α4 are expressed as defect test weights, α1+α2+α3+α4=1.

[0126] In this embodiment, by integrating the LED image test score, LED electrical test score, LED thermal test score and LED mechanical test score, a multi-parameter defect test can be performed to generate a comprehensive LED defect test score. This method integrates the test results of different fields and can comprehensively evaluate the quality and performance of LEDs. By considering image, electronic, thermal and mechanical characteristics, potential defects of LEDs can be identified.

[0127] Reference Figure 1 S40 in the embodiment is applied to an LED defect test scoring level generating unit in a multi-parameter semiconductor defect test system.

[0128] Furthermore, the LED defect levels include:

[0129] Set the defect test level threshold set TH={S1, S2, S3}. When the LED defect test score is less than S1, the corresponding LED defect level is level one; when the LED defect test score is in the interval [S1, S2), the corresponding LED defect level is level two; when the LED defect test score is in the interval [S2, S3], the corresponding LED defect level is level three; when the LED defect test score is greater than S3, the corresponding LED defect level is level four; the lower the level of the LED defect test score, the fewer LED defects; wherein the range of the LED defect test score S is [0, 1], and S1, S2 and S3 are 0.25, 0.5, and 0.75, respectively.

[0130] In this embodiment, four defect test levels are set in this step, and the LED defect test score level is determined according to the calculated LED defect score. The lower the LED defect test score level, the fewer LED defects there are. This method can calculate the LED defect test score level.

[0131] This embodiment evaluates the defect status of the LED by comprehensively considering the image, electrical, thermal and mechanical characteristics of the LED. First, the LED image is input into the image feature test model, the texture features, contrast features, defect areas and their shapes of the image are extracted, and the LED image feature vector is constructed. Based on these feature vectors, combined with the area and perimeter of the defect area, the LED image test score is calculated, thereby providing basic data for the defect level test. When calculating the LED defect level, this embodiment combines the image, electrical, thermal and mechanical test scores of the LED, and calculates the LED defect test score through a multi-parameter defect test method, thereby evaluating the quality and defect conditions of the LED. This detection method improves the accuracy of defect assessment and provides strong data support for the quality control and improvement of LEDs.

[0132] Embodiment 2

[0133] As an embodiment of the present invention, refer to Figure 1 S10 in the embodiment is applied to a data acquisition unit of a multi-parameter semiconductor defect testing system. In this embodiment, image data, electrical data, thermal data and mechanical data of an LED to be tested are collected to construct an LED test data set.

[0134] Furthermore, the LED test data set includes an LED image subset, and the LED image subset includes an LED image to be tested taken by a camera; the electrical data includes forward voltage, reverse voltage, forward current, leakage current, power and photoelectric conversion efficiency; the thermal data includes junction temperature, thermal resistance, heat capacity, thermal conductivity, steady-state and transient thermal characteristics; the above-mentioned mechanical data includes frequency response value, vibration amplitude value, maximum displacement response, impact energy value, stress value, strain value, Young's modulus value and fatigue life.

[0135] Reference Figure 1 S20 in the example is applied to a defect test score generation unit of a multi-parameter semiconductor defect test system, referring to Figure 3 The defect test score generating unit includes a plurality of modules. In this step, the LED image test score, the LED electrical test score, the LED thermal test score and the LED mechanical test score are obtained according to the LED test data set.

[0136] Furthermore, the LED image test score acquisition step includes: inputting the image in the LED image data into the LED image feature test model, acquiring the LED image feature vector, and generating the LED image test score according to the LED image feature vector. The specific process includes:

[0137] The LED image feature test model includes an LED image input layer, an LED image preprocessing layer, an LED image feature extraction layer, an LED image defect recognition layer and an LED image test scoring data output layer;

[0138] The LED image input layer is used to input the image in the LED image data into the LED image feature test model; the LED image preprocessing layer is used to grayscale, denoise and contrast enhance the image in the LED image feature; the LED image feature extraction layer is used to extract features from the preprocessed image, and the specific steps include:

[0139] The processed image is converted into a grayscale image, and a grayscale co-occurrence matrix is ​​calculated to generate a matrix describing the relationship between grayscale values ​​of adjacent pixels in the image; the image texture features are obtained by calculating the combination frequency of grayscale values ​​of adjacent pixels on the grayscale co-occurrence matrix; the image contrast features are obtained by performing weighted calculation on each element in the grayscale co-occurrence matrix to obtain the defect area; the outline of the defect area in the image is extracted by an edge detection algorithm, and the area and perimeter of the defect area are calculated based on the outline to evaluate the shape characteristics of the defect area;

[0140] The LED image defect recognition layer is used to input the features extracted from the LED image feature extraction layer into the trained defect test model to perform defect recognition and obtain a defect image;

[0141] The LED image test score data output layer outputs the LED image test score.

[0142] In this embodiment, the LED to be inspected is photographed by a microscope camera, and a total of 589 LED surface images are collected to construct an LED image subset, and the images in the LED image subset are classified by the LED image feature test model. In this embodiment, the image texture features, contrast features, and shape features of normal and defective types are obtained by analyzing the LED image subset, as shown in Table 2.

[0143] Normalizing the image texture feature, the contrast feature and the shape feature to obtain the LED image feature vector;

[0144] Furthermore, the LED image test score is obtained, and the calculation formula is:

[0145]

[0146] Among them, S im is the LED image test score, N is the number of LED images to be tested, M is the number of elements in the LED image feature vector, S con,i is the contrast feature score of the i-th defect area, ωcon is the contrast feature weight, S cir,i is the roundness characteristic score of the i-th defect area, ω cir is the roundness feature weight, A i is the area of ​​the i-th defect region, A i,max is the area of ​​the i-th defect region, V j is the j-th eigenvector influencing factor.

[0147] Table 2 Image feature comparison table

[0148]

[0149] Furthermore, the generation process of the LED electrical test score includes:

[0150] Collecting electrical characteristics of the area to be tested in the LED, specifically including:

[0151] Connect the LED to a power source and a current source, and ensure that the connections are secure; connect a digital multimeter, an oscilloscope, and a resistance tester to the test points of the LED;

[0152] The forward voltage, forward current, reverse voltage and reverse current of the LED are collected by the digital multimeter and the oscilloscope; the electrical signal waveform of the LED is tested by the oscilloscope, wherein the electrical signal waveform includes a switching response time and a noise dynamic characteristic;

[0153] Constructing a feature vector group through the collected data; normalizing the feature vector group to obtain an LED electrical feature vector group; the LED electrical feature vector group includes the forward voltage, the reverse voltage, the forward current, the reverse current, the forward resistance, the reverse resistance, the response time and the noise;

[0154] The LED electrical test score is calculated according to the LED electrical characteristic vector group, and the calculation formula is:

[0155]

[0156] Among them, S e Rating for LED electrical test, V f is the forward voltage, I f is the forward current, R is the standard resistance, V r is the reverse voltage, I r is the reverse current, κ is the noise factor, κ max is the maximum noise factor, T res is the switch response time, T res,max is the maximum switch response time, ω1, ω2, ω3, ω4 are weight coefficients, ω1+ω2+ω3+ω4=1.

[0157] Furthermore, the process of generating the LED thermal test score includes:

[0158] Collecting the thermal characteristics of the area to be tested in the LED, specifically including:

[0159] Capturing thermal images of the LED surface using a thermal imaging camera to obtain temperature differences in different areas and existing local hot spots, and generating a temperature distribution map of the LED; extracting thermal characteristics of the LED by analyzing the temperature distribution in the image;

[0160] Normalizing the features to construct a LED thermal feature vector group; the LED thermal feature vector group includes average operating temperature, temperature rise rate, temperature uniformity and thermal conductivity estimation;

[0161] The LED thermal test score is calculated according to the LED thermal characteristic vector group, and the calculation formula is:

[0162]

[0163] Among them, S t Rating for LED thermal test, T avg is the average operating temperature, ΔT rise is the temperature rise rate, HS is the total number of hot spots on the LED surface that are significantly higher than the surrounding area, is the reciprocal of thermal conductivity, β1, β2, β3, β4 are weight coefficients, β1+β2+β3+β4=1.

[0164] Furthermore, the process of generating the LED mechanical test score includes:

[0165] By installing the LED on a vibration table and collecting response data at different frequencies and amplitudes, a first mechanical characteristic including a frequency response value and a vibration amplitude value is obtained; by installing the LED on an impact testing machine and collecting response data after impact, a second mechanical characteristic including a maximum displacement response and an impact energy value is obtained; by installing the lead of the LED on a tensile testing machine and collecting force and displacement data, a third mechanical characteristic including a stress value and a strain value is obtained;

[0166] Normalizing the first mechanical feature, the second mechanical feature, and the third mechanical feature, and merging them into an LED mechanical feature vector group;

[0167] The LED mechanical test score is calculated according to the LED mechanical feature vector group, and the calculation formula is:

[0168]

[0169] Among them, S m Scoring LED Mechanical Tests, ¶ j is the weight of the jth feature, x j is the jth eigenvector value, p ji is the element in the i-th row and j-th column of the principal component matrix P, σ j is the standard deviation of the jth feature;

[0170] The specific calculation method of the principal component matrix P is:

[0171]

[0172] Where n is the total number of features, X i is the eigenvector of the i-th feature, X is the eigenmean, Λ is the eigenvalue diagonal matrix, for The transposed matrix of .

[0173] Further, according to the LED image test score, the LED electrical test score, the LED thermal test score and the LED mechanical test score, an LED defect test score is generated through multi-parameter defect testing; the formula of the LED defect test score is:

[0174] S=α1S im +α2S e +α3S t +α4S m ;

[0175] Among them, S is the LED defect test score, S im Score for LED image test, S e Score the LED electrical test, S t Score for LED thermal test, S m Score the LED mechanical test, α1, α2, α3 and α4 are expressed as defect test weights, α1+α2+α3+α4=1.

[0176] Furthermore, the LED defect levels include:

[0177] The defect test level threshold set TH={S1, S2, S3} is set. When the LED defect test score is less than S1, the corresponding LED defect level is level one; when the LED defect test score is in the interval [S1, S2), the corresponding LED defect level is level two; when the LED defect test score is in the interval [S2, S3], the corresponding LED defect level is level three; when the LED defect test score is greater than S3, the corresponding LED defect level is level four; the lower the level of the LED defect test score, the fewer LED defects; wherein the range of S is [0, 1], and S1, S2 and S3 are 0.25, 0.5, and 0.75, respectively.

[0178] In this embodiment, a multi-parameter semiconductor defect testing method is provided. First, an LED test data set is constructed by collecting image data, electrical data, thermal data and mechanical data of an LED. They are used to generate LED image test scores, LED electrical test scores, LED thermal test scores and LED mechanical test scores, respectively, to ensure a comprehensive analysis of the LED from multiple dimensions. In terms of image testing, the LED image is processed by an LED image feature test model to accurately extract texture, contrast and shape features from the image. The electrical test collects data such as voltage, current, resistance and other data of the LED through a series of electrical instruments, and constructs a feature vector group. The accurate measurement of these electrical features helps to identify the electrical performance problems of the LED. The thermal test obtains the temperature distribution data of the LED through a thermal imaging camera to evaluate its thermal performance, especially the temperature rise rate and thermal conductivity, which is crucial for identifying possible thermal damage. The mechanical test evaluates the performance of the LED under different mechanical stresses through experiments such as vibration, impact and stretching, and detects its mechanical stability and durability. Combining these scoring data, a multi-parameter defect test method is used to generate an LED defect test score, thereby comprehensively evaluating the overall quality and defect conditions of the LED. This method divides LED defect test scores into different levels by setting defect test level thresholds, thereby systematically describing the severity of LED defects and improving the comprehensiveness and accuracy of detection from a multi-dimensional and systematic defect detection method.

[0179] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-parameter semiconductor defect testing method, characterized in that: include: Collect LED image data, electrical data, thermal data and mechanical data to build LED test data set; According to the LED test data set, an LED image test score, an LED electrical test score, an LED thermal test score, and an LED mechanical test score are obtained; The LED image test score acquisition step includes: inputting the image in the LED image data into the LED image feature test model, acquiring the LED image feature vector, and generating the LED image test score according to the LED image feature vector; According to the LED image test score, the LED electrical test score, the LED thermal test score and the LED mechanical test score, an LED defect test score is generated through multi-parameter defect testing; the formula of the LED defect test score is: S=α1S im +α2S e +α3S t +α4S m ; Among them, S is the LED defect test score, S im Score for LED image test, S e Score LED electrical test, S t Score for LED thermal test, S m Score the LED mechanical test, α1, α2, α3 and α4 are expressed as defect test weights, α1+α2+α3+α4=1; According to the LED defect test score, the LED defect grade is obtained.

2. A multi-parameter semiconductor defect testing method according to claim 1, characterized in that: The generation process of the LED image test score includes: The LED image feature test model includes an LED image input layer, an LED image preprocessing layer, an LED image feature extraction layer, an LED image defect recognition layer and an LED image test scoring data output layer; The LED image input layer is used to input the image in the LED image data into the LED image feature test model; the LED image preprocessing layer is used to grayscale, denoise and contrast enhance the image in the LED image data; the LED image feature extraction layer is used to extract features from the preprocessed image, and the specific steps include: The processed image is converted into a grayscale image, and a grayscale co-occurrence matrix is ​​calculated to generate a matrix describing the relationship between grayscale values ​​of adjacent pixels in the image; the image texture features are obtained by calculating the combination frequency of grayscale values ​​of adjacent pixels on the grayscale co-occurrence matrix; the image contrast features are obtained by performing weighted calculation on each element in the grayscale co-occurrence matrix to obtain the defect area; the outline of the defect area in the image is extracted by an edge detection algorithm, and the area and perimeter of the defect area are calculated based on the outline to evaluate the shape features of the defect area; the image texture features, the contrast features and the shape features are normalized to obtain the LED image feature vector; The LED image defect recognition layer is used to input the features extracted from the LED image feature extraction layer into the trained defect test model to perform defect recognition and obtain a defect image; The LED image test score data output layer outputs the LED image test score.

3. A multi-parameter semiconductor defect testing method according to claim 2, characterized in that: According to the defect image, the LED image test score is obtained, and the calculation formula is: Among them, S im is the LED image test score, N is the number of LED images to be tested, M is the number of elements in the LED image feature vector, S con,i is the contrast feature score of the i-th defect area, ω con is the contrast feature weight, S cir,i is the roundness characteristic score of the i-th defect area, ω cir is the roundness feature weight, A i is the area of ​​the i-th defect region, A i,max is the area of ​​the i-th defect region, V j is the jth eigenvector influencing factor.

4. A multi-parameter semiconductor defect testing method according to claim 1, characterized in that: The process of generating the LED electrical test score includes: Collecting electrical data of the area to be tested in the LED, specifically including: Connect the LED to a power source and a current source, and connect a digital multimeter, an oscilloscope, and a resistance tester to the test points of the LED; The forward voltage, forward current, reverse voltage and reverse current of the LED are collected by the digital multimeter and the oscilloscope; the electrical signal waveform of the LED is tested by the oscilloscope, wherein the electrical signal waveform includes a switching response time and a noise dynamic characteristic; Constructing a feature vector group through the collected data; normalizing the feature vector group to obtain an LED electrical feature vector group; the LED electrical feature vector group includes the forward voltage, the reverse voltage, the forward current, the reverse current, the forward resistance, the reverse resistance, the response time and the noise; The LED electrical test score is calculated according to the LED electrical characteristic vector group, and the calculation formula is: Among them, S e Rating for LED electrical test, V f is the forward voltage, I f is the forward current, R is the standard resistance, V r is the reverse voltage, I r is the reverse current, κ is the noise factor, κ max is the maximum noise factor, T res is the switch response time, T res,max is the maximum switch response time, ω1, ω2, ω3, ω4 are weight coefficients, ω1+ω2+ω3+ω4=1.

5. The multi-parameter semiconductor defect testing method according to claim 1, characterized in that: The process of generating the LED thermal test score includes: Collecting thermal data of the area to be tested in the LED, specifically including: Capturing thermal images of the LED surface using a thermal imaging camera to obtain temperature differences in different areas and existing local hot spots, and generating a temperature distribution map of the LED; extracting thermal characteristics of the LED by analyzing the temperature distribution in the image; Normalizing the features to construct a LED thermal feature vector group; the LED thermal feature vector group includes average operating temperature, temperature rise rate, temperature uniformity and thermal conductivity estimation; The LED thermal test score is calculated according to the LED thermal characteristic vector group, and the calculation formula is: Among them, S t Rating for LED thermal test, T avg is the average operating temperature, ΔT rise is the temperature rise rate, HS is the total number of hot spots on the LED surface that are significantly higher than the surrounding area, is the reciprocal of thermal conductivity, β1, β2, β3, β4 are weight coefficients, β1+β2+β3+β4=1.

6. A multi-parameter semiconductor defect testing method according to claim 1, characterized in that: The process of generating the LED mechanical test score includes: Collect mechanical data of the area to be tested in the LED, including: By installing the LED on a vibration table and collecting response data at different frequencies and amplitudes, a first mechanical characteristic including a frequency response value and a vibration amplitude value is obtained; by installing the LED on an impact testing machine and collecting response data after impact, a second mechanical characteristic including a maximum displacement response and an impact energy value is obtained; by installing the lead of the LED on a tensile testing machine and collecting force and displacement data, a third mechanical characteristic including a stress value and a strain value is obtained; Normalizing the first mechanical feature, the second mechanical feature, and the third mechanical feature, and merging them into an LED mechanical feature vector group; The LED mechanical test score is calculated according to the LED mechanical feature vector group, and the calculation formula is: Among them, S m Scoring LED Mechanical Tests, ¶ j is the weight of the jth feature, x j is the jth eigenvector value, p ji is the element in the i-th row and j-th column of the principal component matrix P, σ j is the standard deviation of the jth feature; The specific calculation method of the principal component matrix P is: Where n is the total number of features, X i is the eigenvector of the ith feature, is the eigenmean, Λ is the eigenvalue diagonal matrix, for The transposed matrix of .

7. The multi-parameter semiconductor defect testing method according to claim 1, characterized in that: The LED defect levels include: Set the defect test level threshold set TH={S1, S2, S3}. When the LED defect test score is less than S1, the corresponding LED defect level is level one; when the LED defect test score is in the interval [S1, S2), the corresponding LED defect level is level two; when the LED defect test score is in the interval [S2, S3], the corresponding LED defect level is level three; when the LED defect test score is greater than S3, the corresponding LED defect level is level four; the lower the level of the LED defect test score, the fewer LED defects; wherein the range of the LED defect test score S is [0, 1], and S1, S2 and S3 are 0.25, 0.5, and 0.75, respectively.

8. A multi-parameter semiconductor defect testing system, characterized in that: include: Data acquisition unit: used to collect LED image data, electrical data, thermal data and mechanical data; Defect test score generating unit: used to construct a corresponding feature vector group according to the collected data, and obtain the LED image test score, LED electrical test score, LED thermal test score and LED mechanical test score; An LED defect test score generating unit: configured to generate an LED defect test score through a multi-parameter defect test according to the LED image test score, the LED electrical test score, the LED thermal test score and the LED mechanical test score; LED defect test score level generating unit: used for obtaining the LED defect level according to the LED defect test score.

9. The multi-parameter semiconductor defect testing system according to claim 8, characterized in that: The LED defect levels include: Set the defect test level threshold set TH={S1, S2, S3}. When the LED defect test score is less than S1, the corresponding LED defect level is level one; when the LED defect test score is in the interval [S1, S2), the corresponding LED defect level is level two; when the LED defect test score is in the interval [S2, S3], the corresponding LED defect level is level three; when the LED defect test score is greater than S3, the corresponding LED defect level is level four; the lower the level of the LED defect test score, the fewer LED defects; wherein the range of the LED defect test score S is [0, 1], and S1, S2 and S3 are 0.25, 0.5, and 0.75, respectively.

10. The multi-parameter semiconductor defect testing system according to claim 8, characterized in that: Collecting electrical characteristics of the area to be tested in the LED, specifically including: Connect the LED to a power source and a current source, and ensure that the connections are secure; connect a digital multimeter, an oscilloscope, and a resistance tester to the test points of the LED; The forward voltage, forward current, reverse voltage and reverse current of the LED are collected by the digital multimeter and the oscilloscope; the electrical signal waveform of the LED is tested by the oscilloscope, wherein the electrical signal waveform includes a switching response time and a noise dynamic characteristic; Constructing a feature vector group through the collected data; normalizing the feature vector group to obtain an LED electrical feature vector group; the LED electrical feature vector group includes the forward voltage, the reverse voltage, the forward current, the reverse current, the forward resistance, the reverse resistance, the response time and the noise; The LED electrical test score is calculated according to the LED electrical characteristic vector group, and the calculation formula is: Among them, S e Rating for LED electrical test, V f is the forward voltage, I f is the forward current, R is the standard resistance, V r is the reverse voltage, I r is the reverse current, κ is the noise factor, κ max is the maximum noise factor, T res is the switch response time, T res,max is the maximum switch response time, ω1, ω2, ω3, ω4 are weight coefficients, ω1+ω2+ω3+ω4=1.

Images