AI detection method and equipment for appearance defects of small quartz crystal resonator

Abstract

The invention discloses an AI detection method and equipment for appearance defects of a small quartz crystal resonator, and relates to the technical field of appearance detection of quartz crystal resonators, postures of the quartz crystal resonator are automatically detected through equipment such as a photoelectric sensor and a correlation type optical fiber sensor, and the front, back, longitudinal and transverse states of a product can be quickly distinguished. Products which do not accord with standard postures are automatically analyzed, and the postures are adjusted by adopting a corresponding blowing mode, so that the products quickly reach the standard postures suitable for detection, manual intervention is avoided, the feeding and posture adjusting efficiency is greatly improved, and the production cost is reduced. The method for extracting the defect area through gray scale comparison and the defect judgment method based on gray scale threshold segmentation are combined for metal surface defect detection, high-precision and high-efficiency detection of metal surface defects is achieved, and the method is particularly suitable for automatic production of precise elements such as quartz crystal resonators and the like. And a reliable defect detection solution is provided for the field of electronic manufacturing.

Description

Technical Field

[0001] The present invention relates to the technical field of quartz crystal resonator appearance inspection, and in particular to an AI inspection method and device for appearance defects of a small quartz crystal resonator. Background Art

[0002] Small quartz crystal resonators are the core frequency control components of electronic devices, and their appearance defects directly affect product performance and reliability. Traditional detection methods mainly rely on manual visual inspection or mechanical contact detection, which have problems such as low efficiency, strong subjectivity, and easy missed detection. With the increase in demand for automated production, machine vision-based detection technology has gradually become popular, but it still faces many technical bottlenecks. Therefore, an AI detection method and equipment for appearance defects of small quartz crystal resonators is needed.

[0003] Prior art, such as the invention application patent with announcement number: CN114844482A, discloses a chip for high-frequency crystal resonator and its production control system, which relates to the field of quartz crystal resonator technology, and includes a parameter detection module, a production tracking module and an alarm module; the parameter detection module performs quality detection on each chip based on multiple detection tasks, and sends the corresponding signal and detection data to the host computer with a timestamp; the detection tasks include size detection, appearance detection and resistance detection; wherein the detection data all carry equipment identification; the production tracking module is used to obtain the detection data with the same equipment identification generated by the parameter detection module for production tracking analysis. If the production deviation YD is greater than the deviation threshold, it is determined that the corresponding production equipment has a trend of producing unqualified chips, and a production abnormality signal is generated to remind the management personnel of the mobile terminal to inspect the production equipment and improve the chip quality and pass rate.

[0004] In view of the above scheme, the inventors of the present application have found that the above technology has at least the following technical problems: 1. The surface of quartz crystal is prone to mirror reflection. The traditional grayscale threshold segmentation method will fail to segment at the junction of reflective edges, such as the junction of metal coating and glass packaging, due to grayscale mutation, and the normal area will be mistakenly judged as a defect or the real defect in the reflective area will be missed. For sub-millimeter scratches, dents or pinholes, traditional algorithms are prone to miss detection or misjudge them as background noise due to insufficient feature extraction capabilities. When multiple defects are superimposed, traditional methods cannot effectively separate and identify composite features, resulting in confusing detection results.

[0005] 2. Traditional algorithms rely on manually designed features, and the processing time for a single frame is long, which makes it difficult to meet the real-time requirements of high-speed production lines. For different types of quartz crystals, thresholds, filter parameters, etc. need to be manually adjusted, and the debugging cycle is long, which cannot meet the needs of rapid production changes.

[0006] 3. Slight changes in light intensity or angle on existing production lines can significantly affect the grayscale distribution of images. Traditional methods require frequent calibration, and traditional algorithms lack the ability to uniformly process the reflective properties of different materials. The natural texture on the surface of quartz crystals is easily misjudged as a defect by traditional algorithms. For defects with depth information such as depressions and bumps, traditional two-dimensional vision is difficult to accurately quantify and requires reliance on offline three-dimensional detection equipment, resulting in a fragmented detection process.

[0007] 4. Traditional methods require manual labeling of defect samples, and the detection rules need to be manually summarized, which is highly subjective and difficult to deal with new defects. When a new process or new defect type is introduced to the production line, the algorithm logic needs to be redesigned. The sizes and shapes of different types of quartz crystals vary greatly (such as cylindrical and SMD packaging). Traditional methods require customized detection logic for each model, and cannot fuse multi-sensor data for joint detection, which limits the comprehensiveness of defect detection. Summary of the invention

[0008] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to provide an AI detection method and device for appearance defects of a small quartz crystal resonator.

[0009] To solve the above technical problems, the present invention adopts the following technical solutions: In the first aspect, the present invention provides an AI detection method for appearance defects of a small quartz crystal resonator, including: Step 1, product loading and posture adjustment: When each quartz crystal resonator in the current batch of the target enterprise is transmitted to the detection equipment, the product posture corresponding to each quartz crystal resonator is analyzed, and each quartz crystal resonator that meets the standards is recorded as a standard quartz crystal resonator, and then the posture of each standard quartz crystal resonator is adjusted.

[0010] Step 2, image acquisition and preprocessing: When the front side of each standard quartz crystal resonator is automatically transmitted to the No. 1 camera position, the front quality parameters corresponding to the front side of each standard quartz crystal resonator are obtained, and each standard quartz crystal resonator is automatically transmitted to the No. 2 camera position to obtain the back quality parameters corresponding to the back side of each standard quartz crystal resonator.

[0011] Step 3: Defect detection and identification: According to the front quality parameters corresponding to the front of each standard quartz crystal resonator and the back quality parameters corresponding to the back of the back, identify whether the appearance quality of each standard quartz crystal resonator is qualified. If the appearance quality of a standard quartz crystal resonator is unqualified, classify the standard quartz crystal resonator into the corresponding NG material box.

[0012] In a second aspect, the present invention provides a small quartz crystal resonator appearance defect AI detection device, including: a product loading and posture adjustment module: used to analyze the product posture corresponding to each quartz crystal resonator after each quartz crystal resonator in the current batch of the target enterprise is transmitted to the detection equipment, and record each quartz crystal resonator that meets the standard as a standard quartz crystal resonator, and then adjust the posture of each standard quartz crystal resonator.

[0013] Image acquisition and preprocessing module: used to automatically transmit the front of each standard quartz crystal resonator to the No. 1 camera position to obtain the front quality parameters corresponding to the front of each standard quartz crystal resonator, and then automatically transmit each standard quartz crystal resonator to the No. 2 camera position to obtain the back quality parameters corresponding to the back of each standard quartz crystal resonator.

[0014] Defect detection and identification module: It is used to identify whether the appearance quality of each standard quartz crystal resonator is qualified according to the front quality parameters corresponding to the front of each standard quartz crystal resonator and the back quality parameters corresponding to the back of the resonator. If the appearance quality of a standard quartz crystal resonator is unqualified, the standard quartz crystal resonator will be classified into the corresponding NG material box.

[0015] The beneficial effects of the present invention are as follows: 1. The embodiment of the present invention can quickly distinguish the front, back, longitudinal and lateral states of the product by automatically detecting the posture of the quartz crystal resonator through devices such as photoelectric sensors and reflected optical fiber sensors. For products that do not meet the standard posture, the posture is automatically analyzed and adjusted using the corresponding blowing mode, so that the product can quickly reach the standard posture suitable for detection, avoiding manual intervention, greatly improving the efficiency of loading and posture adjustment, and saving time for subsequent detection processes. The posture of the standard quartz crystal resonator is accurately adjusted using high-precision position sensors, pressure sensors, and high-precision visual inspection systems. During the adjustment process, the adsorption force, electrostatic distribution and intensity are monitored and controlled in real time to ensure that the lateral deviation of the product is less than 0.05mm, so that the product can be imaged in a standard posture in the camera detection area, reducing the detection error caused by inaccurate posture and improving the accuracy of the detection results.

[0016] 2. In the embodiment of the present invention, the front and back images of the quartz crystal resonator are collected at the camera position 1 and the camera position 2 respectively, and the images are preprocessed. By separating the reflective and non-reflective areas, multiple key parameters such as grayscale difference, gradient amplitude, reflective area, average grayscale, defect area, circularity, pin position deviation and oxidation area ratio are extracted. These parameters can fully reflect the characteristics of the product surface and possible defects, and provide rich information for subsequent defect detection and identification. The collected images are processed by Gaussian filtering noise reduction technology, which can remove noise interference in the image and make the image clearer, which is conducive to the subsequent extraction and analysis of image features, and further improves the accuracy of detection. Through the grayscale value comparison method of the standard image, the area with large changes is independently separated into areas suitable for grayscale threshold processing, and then refined for detection. The method of extracting defect areas by grayscale contrast and the defect judgment method based on grayscale threshold segmentation are combined for metal surface defect detection, which realizes high-precision and high-efficiency detection of metal surface defects, which is particularly suitable for the automated production of precision components such as quartz crystal resonators, and provides a reliable defect detection solution for the field of electronic manufacturing.

[0017] 3. The embodiment of the present invention can quantitatively evaluate the appearance quality of the product by calculating the appearance quality evaluation value corresponding to each standard quartz crystal resonator and comparing it with the set standard value. This quantitative evaluation method avoids the error of subjective judgment and makes the test results more objective and accurate. The characteristic parameters of different areas are analyzed by multiple models such as the reflective area evaluation value analysis model, the non-reflective area evaluation value analysis model, the pin evaluation value analysis model and the defect evaluation value analysis model, which comprehensively considers the defects of different parts of the product and improves the comprehensiveness and accuracy of defect detection.

[0018] 4. The embodiment of the present invention compares the appearance quality assessment value of the unqualified product with the appearance quality assessment value interval corresponding to each quartz crystal resonator defect type in the database, which can accurately determine the defect type of the product and classify it into the corresponding NG material box. This precise classification method is conducive to the subsequent analysis and processing of defective products, and improves the efficiency of production management. The entire inspection process, from product loading, posture adjustment, image acquisition, defect detection to defect classification, has realized automated operation, reduced manual participation, reduced labor costs, and improved the stability and consistency of detection. It is suitable for quartz crystal resonator appearance defect detection on large-scale production lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 The present invention is a flowchart of the steps for implementing the method.

[0021] Figure 2 This is a schematic diagram of the connection of the equipment modules of the present invention. DETAILED DESCRIPTION

[0022] 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.

[0023] Embodiments of the present invention Figure 1 As shown, an AI detection method for appearance defects of small quartz crystal resonators includes: Step 1, product loading and posture adjustment: When each quartz crystal resonator in the current batch of the target enterprise is transmitted to the detection equipment, the product posture corresponding to each quartz crystal resonator is analyzed, and each quartz crystal resonator that meets the standard is recorded as a standard quartz crystal resonator, and then the posture of each standard quartz crystal resonator is adjusted.

[0024] In a specific embodiment, the product posture corresponding to each quartz crystal resonator is analyzed, and the specific analysis process is as follows: A1. After each quartz crystal resonator in the current batch of the target enterprise is transmitted to the detection equipment, the photoelectric sensor of the hopper of the detection equipment detects the presence of the quartz crystal resonator, triggers the vibration motor to start, and then automatically sets the vibration frequency of the vibration motor corresponding to each quartz crystal resonator. The vibration frequency of the vibration motor is calculated according to the formula: .

[0025] A2. When each quartz crystal resonator falls into the vibration plate in the detection equipment through the vibration of the vibration motor, two sets of Keyence through-beam fiber optic sensors are installed at the discharge port of the vibration plate in the detection equipment. When each quartz crystal resonator passes through the detection area of ​​the two sets of Keyence through-beam fiber optic sensors, the Keyence through-beam fiber optic sensors will receive two sets of corresponding optical signals and convert them into electrical signals for output. The signal strengths of the two sets of Keyence through-beam fiber optic sensors corresponding to each quartz crystal resonator are recorded as and .

[0026] A3. The signal strength of the two Keyence fiber optic sensors corresponding to each quartz crystal resonator is and The signal strengths of the two Keyence fiber optic sensors corresponding to a quartz crystal resonator are compared with those of the preset quartz crystal resonator. and If the signal strengths of the two Keyence fiber optic sensors corresponding to a quartz crystal resonator are both greater than the preset signal strengths of the quartz crystal resonator, the quartz crystal resonator is judged to be positive. On the contrary, if the signal strengths of the two Keyence fiber optic sensors corresponding to a quartz crystal resonator are greater than the preset signal strengths of the quartz crystal resonator, the quartz crystal resonator is judged to be positive. and If not all of the signal strengths are greater than the preset signal strength corresponding to the quartz crystal resonator, it is determined that the quartz crystal resonator is the back side.

[0027] A4. The signal strength of the two Keyence fiber optic sensors corresponding to each quartz crystal resonator is and Perform difference calculation, the signal strength difference is equal to and The absolute value of the difference is calculated, and the signal strength difference output by the two groups of Keyence opposed-beam fiber optic sensors corresponding to each quartz crystal resonator is compared with the signal strength difference corresponding to the preset quartz crystal resonator. If the signal strength difference output by the two groups of Keyence opposed-beam fiber optic sensors corresponding to a certain quartz crystal resonator is greater than or equal to the signal strength difference corresponding to the preset quartz crystal resonator, the quartz crystal resonator is determined to be longitudinal. Conversely, if the signal strength difference output by the two groups of Keyence opposed-beam fiber optic sensors corresponding to a certain quartz crystal resonator is less than the signal strength difference corresponding to the preset quartz crystal resonator, the quartz crystal resonator is determined to be transverse.

[0028] A5. If a quartz crystal resonator is facing forward, it is directly transmitted to the glass plate in the detection equipment. If a quartz crystal resonator is facing backward or vertically, the blowing pattern corresponding to the quartz crystal resonator is analyzed, and the quartz crystal resonator is blown to the bottom of the vibration plate according to the corresponding blowing pattern, so that the quartz crystal resonator faces forward and horizontally, and then transmitted to the glass plate in the detection equipment. The quartz crystal resonators corresponding to the glass plate transmitted to the detection equipment are judged as quartz crystal resonators that meet the standards.

[0029] It should be noted that the blowing mode is a composite control strategy based on pneumatic posture adjustment and dynamic matching of models. The posture of the product is changed by the impact of the airflow to ensure that it enters the inspection station in a standard posture (front and horizontal). The core design points are as follows: Back blowing: The airflow acts vertically on the back of the product, and the recoil force is used to flip the product 180°. Longitudinal blowing: The airflow is applied along the longitudinal axis of the product, and the longitudinal posture is adjusted to the horizontal through the rotational torque. Pulsed airflow (0.2-0.5MPa) is used to ensure that the posture is changed without damaging the product. The airflow action time is controlled at 5-10ms, which is synchronized with the movement cycle of the vibration plate.

[0030] In a specific embodiment, the blowing pattern corresponding to the quartz crystal resonator is analyzed, and the specific analysis process is as follows: B1. Obtain the model corresponding to the current batch of quartz crystal resonators, and compare the model corresponding to the current batch of quartz crystal resonators with the model corresponding to the back blowing pattern of each quartz crystal resonator in the database. If the model corresponding to the current batch of quartz crystal resonators is the same as the model corresponding to the back blowing pattern of a quartz crystal resonator in the database, then the back blowing pattern of the quartz crystal resonator in the database is recorded as the back blowing pattern corresponding to the quartz crystal resonator.

[0031] B2. Compare the model corresponding to the current batch of quartz crystal resonators with the model corresponding to the longitudinal blowing mode of each quartz crystal resonator in the database. If the model corresponding to the current batch of quartz crystal resonators is the same as the model corresponding to the longitudinal blowing mode of a quartz crystal resonator in the database, then the longitudinal blowing mode of the quartz crystal resonator in the database is recorded as the longitudinal blowing mode corresponding to the quartz crystal resonator.

[0032] In a specific embodiment, the posture adjustment of each standard quartz crystal resonator is carried out as follows: after each standard quartz crystal resonator is transferred to the guide module through the glass plate in the detection equipment, a high-precision position sensor is used to detect its initial position, including information in the lateral, longitudinal and height directions. The control system drives the mechanical structure of the guide module to align the vacuum suction nozzle with a diameter of 3mm with the product center of each standard quartz crystal resonator, start the vacuum system to generate adsorption force, and the pressure sensor monitors the adsorption force in real time to stably control it within the range of 0.05-0.1MPa. The start model is The electrostatic generator of Simco-Ion7850 applies 2kV static electricity on the surface of the guide rail, uses a high-precision visual inspection system to monitor the product posture in real time, and calculates its lateral deviation value relative to the ideal posture. Based on the deviation value, the control system adjusts the static electricity distribution and intensity in real time, corrects the product posture through Coulomb force, and repeats the inspection and adjustment steps until the lateral deviation of the product is less than 0.05mm. When the product posture fully meets the standard, the vacuum system and the electrostatic generator are turned off, each standard quartz crystal resonator is released, and each standard quartz crystal resonator is transported to the camera inspection area for appearance defect detection.

[0033] It should be noted that when the standard quartz crystal resonator is transmitted through the guide module, a Keyence LV-4100 industrial camera (resolution 2592×1944, frame rate 30fps) is used with a telecentric lens to collect images in real time, and the lateral deviation of the product is calculated through the Otsu binarization and ellipse fitting algorithm. The control system drives the Simco-Ion7850 electrostatic generator to apply 0-2kV dynamic voltage to the 8 sets of independent electrodes on the guide rail, and uses the Coulomb force to correct the posture in real time; through the iterative adjustment of the PID controller (Kp=10kV / mm, Kd=5kVs / mm), combined with cubic spline interpolation sub-pixel edge detection and COMSOL optimized uniform electrostatic field (uniformity>95%), precise control of lateral deviation <0.05mm is achieved, the average adjustment time is 120ms, and the repeatability accuracy is ±0.015mm, ensuring that the product enters the inspection station with a standard posture.

[0034] Step 2, image acquisition and preprocessing: When the front side of each standard quartz crystal resonator is automatically transmitted to the No. 1 camera position, the front quality parameters corresponding to the front side of each standard quartz crystal resonator are obtained, and each standard quartz crystal resonator is automatically transmitted to the No. 2 camera position to obtain the back quality parameters corresponding to the back side of each standard quartz crystal resonator.

[0035] In a specific embodiment, the front quality parameters and back quality parameters corresponding to the front of each standard quartz crystal resonator are obtained, and the specific acquisition process is as follows: C1, when the front of each standard quartz crystal resonator is transmitted to the first camera position, a ring LED light source is used to illuminate the image at 45° obliquely, and after Gaussian filtering and noise reduction, the pixel-by-pixel grayscale difference is used to obtain the image. , pixel grayscale difference Greater than 50 marks reflective area, poor grayscale Less than 50 marks the non-reflective area, separates the reflective and non-reflective areas: extracts the gradient amplitude in the reflective area And reflective area , Otsu threshold segmentation is applied to calculate the average grayscale in the non-reflective area , defect area and circularity At the same time, the reflective area uses Canny edge detection to detect the pin outline, and combines sub-pixel edge detection to calculate the pin position deviation = / , Refers to the position offset on the horizontal X-axis. Refers to the position offset on the vertical Y axis; and based on the grayscale co-occurrence matrix, the low grayscale area ratio is analyzed to quantify the degree of oxidation. The final output includes the grayscale difference corresponding to the front of each standard quartz crystal resonator. , gradient amplitude , Reflective area , average grayscale , defect area , circularity , Pin position deviation and oxidation area ratio Key parameters, among which, Indicates the number corresponding to each standard quartz crystal resonator, , u is a positive integer.

[0036] C2. When the back of each standard quartz crystal resonator is transmitted to the second camera position, a ring LED light source is used to illuminate the image at a 45° angle. After Gaussian filtering and noise reduction, the image is obtained by pixel-by-pixel grayscale difference. , pixel grayscale difference Greater than 50 marks reflective area, poor grayscale Less than 50 marks the non-reflective area, separates the reflective and non-reflective areas: extracts the gradient amplitude in the reflective area And reflective area , Otsu threshold segmentation is applied to calculate the average grayscale in the non-reflective area , defect area and circularity At the same time, the reflective area uses Canny edge detection to detect the pin outline, and combines sub-pixel edge detection to calculate the pin position deviation = / , Refers to the position offset on the horizontal X-axis. Refers to the position offset on the vertical Y axis; and based on the grayscale co-occurrence matrix, the low grayscale area ratio is analyzed to quantify the degree of oxidation. The final output includes the grayscale difference corresponding to the back of each standard quartz crystal resonator. , gradient amplitude , Reflective area , average grayscale , defect area , circularity , Pin position deviation and oxidation area ratio Key parameters.

[0037] Step 3: Defect detection and identification: According to the front quality parameters corresponding to the front of each standard quartz crystal resonator and the back quality parameters corresponding to the back of the back, identify whether the appearance quality of each standard quartz crystal resonator is qualified. If the appearance quality of a standard quartz crystal resonator is unqualified, classify the standard quartz crystal resonator into the corresponding NG material box.

[0038] In a specific embodiment, the identification of whether the appearance quality of each standard quartz crystal resonator is qualified, the specific analysis process is as follows: analyze the appearance quality evaluation value corresponding to each standard quartz crystal resonator, and compare the appearance quality evaluation value corresponding to each standard quartz crystal resonator with the appearance quality evaluation value corresponding to the set standard quartz crystal resonator. If the appearance quality evaluation value corresponding to a standard quartz crystal resonator is greater than or equal to the appearance quality evaluation value corresponding to the set standard quartz crystal resonator, then the appearance quality of the standard quartz crystal resonator is identified as unqualified. Otherwise, if the appearance quality evaluation value corresponding to a standard quartz crystal resonator is less than the appearance quality evaluation value corresponding to the set standard quartz crystal resonator, then the appearance quality of the standard quartz crystal resonator is identified as qualified.

[0039] In a specific embodiment, the analysis of the appearance quality evaluation value corresponding to each standard quartz crystal resonator is as follows: obtaining the front defect evaluation value and the back defect evaluation value corresponding to each standard quartz crystal resonator, and recording the front defect evaluation value and the back defect evaluation value corresponding to each standard quartz crystal resonator as and ,in, Indicates the number corresponding to each standard quartz crystal resonator, , u is a positive integer, substitute into the calculation formula: The appearance quality evaluation value corresponding to each standard quartz crystal resonator is obtained. ,in, , They are the standard front defect evaluation value and the standard back defect evaluation value corresponding to the set quartz crystal resonator, , They are the weight factors corresponding to the set front defect evaluation value of the quartz crystal resonator and the weight factors corresponding to the standard back defect evaluation value, respectively. is a natural constant.

[0040] It should be noted that , Both are greater than 0 and less than 1.

[0041] It should also be noted that the setting of standard values ​​needs to be based on industry standards and enterprise process specifications, combined with statistical analysis of historical good product data. First, the distribution characteristics of the front / back defect evaluation values ​​of good products are determined through Gaussian fitting, and the μ±3σ range is taken as the reference interval. Then, the standard value is determined according to the product grade requirements. The weight factor needs to be dynamically adjusted according to the degree of influence of the defect on the product performance. The weight factor is determined by the analytic hierarchy process (AHP) combined with the failure mode analysis (FMEA).

[0042] In a specific embodiment, the front defect evaluation value and the back defect evaluation value corresponding to each standard quartz crystal resonator are obtained, and the specific acquisition process is as follows: D1, the grayscale difference corresponding to the front of each standard quartz crystal resonator is calculated. , gradient amplitude , Reflective area , and the grayscale difference corresponding to the back of each standard quartz crystal resonator , gradient amplitude , Reflective area As input information, they are respectively entered into the reflective area evaluation value analysis model. After calculation and analysis by the reflective area evaluation value analysis model, the front reflective area evaluation value corresponding to each standard quartz crystal resonator is finally output. and the back reflective area evaluation value .

[0043] It should be noted that the analysis process of the evaluation value of the front reflective area corresponding to each standard quartz crystal resonator is as follows: the grayscale difference corresponding to the front of each standard quartz crystal resonator is calculated. , gradient amplitude , Reflective area After normalization, substitute into the analysis formula: , get the front reflective area evaluation value corresponding to each standard quartz crystal resonator .

[0044] It should also be noted that the evaluation value of the back reflective area corresponding to each standard quartz crystal resonator is obtained by analyzing the evaluation value of the front reflective area corresponding to each standard quartz crystal resonator.

[0045] D2. The average grayscale corresponding to the front of each standard quartz crystal resonator , defect area , circularity , and the corresponding average grayscale on the back of each standard quartz crystal resonator , defect area , circularity As input information, they are entered into the non-reflective area evaluation value analysis model respectively. After calculation and analysis by the non-reflective area evaluation value analysis model, the front non-reflective area evaluation value corresponding to each standard quartz crystal resonator is finally output. and the back non-reflective area evaluation value .

[0046] It should be noted that the analysis process of the evaluation value of the front non-reflective area corresponding to each standard quartz crystal resonator is as follows: the average grayscale corresponding to the front of each standard quartz crystal resonator , defect area , circularity After normalization, substitute into the analysis formula: , get the evaluation value of the front non-reflective area corresponding to each standard quartz crystal resonator .

[0047] It should also be noted that the evaluation value of the back non-reflective area corresponding to each standard quartz crystal resonator is obtained by analyzing the evaluation value of the front non-reflective area corresponding to each standard quartz crystal resonator.

[0048] D3. The pin position deviation corresponding to the front side of each standard quartz crystal resonator and oxidation area ratio , and the corresponding pin position deviation on the back of each standard quartz crystal resonator and oxidation area ratio As input information, they are entered into the pin evaluation value analysis model respectively. After calculation and analysis by the pin evaluation value analysis model, the positive pin evaluation value corresponding to each standard quartz crystal resonator is finally output. and back pin evaluation value .

[0049] It should be noted that the analysis process of the front pin evaluation value corresponding to each standard quartz crystal resonator is as follows: the pin position deviation corresponding to the front of each standard quartz crystal resonator and oxidation area ratio Through normalization processing, substitute into the analysis formula , get the corresponding positive pin evaluation value of each standard quartz crystal resonator .

[0050] It should also be noted that the back pin evaluation value corresponding to each standard quartz crystal resonator is obtained by analyzing the evaluation value of the front pin corresponding to each standard quartz crystal resonator.

[0051] D4. The front reflective area evaluation value corresponding to the front of each standard quartz crystal resonator , front non-reflective area evaluation value and positive pin evaluation value , and the corresponding back reflective area evaluation value of each standard quartz crystal resonator , Evaluation value of non-reflective area on the back and back pin evaluation value As input information, the defect evaluation value analysis model is used respectively. After calculation and analysis by the defect evaluation value analysis model, the positive defect evaluation value corresponding to each standard quartz crystal resonator is finally output. and back defect evaluation value .

[0052] It should be noted that the analysis process of the front defect evaluation value corresponding to each standard quartz crystal resonator is as follows: the front reflective area evaluation value corresponding to the front of each standard quartz crystal resonator , front non-reflective area evaluation value and positive pin evaluation value Through normalization processing, substitute into the analysis formula , and obtain the positive defect evaluation value corresponding to each standard quartz crystal resonator .

[0053] It should also be noted that the back defect evaluation value corresponding to each standard quartz crystal resonator is obtained by analyzing the analysis process of the front defect evaluation value corresponding to each standard quartz crystal resonator.

[0054] It should also be noted that the back pin evaluation value corresponding to each standard quartz crystal resonator is obtained by analyzing the evaluation value of the front pin corresponding to each standard quartz crystal resonator.

[0055] In a specific embodiment, if the appearance quality of a standard quartz crystal resonator is unqualified, the standard quartz crystal resonator is classified into the corresponding NG material box, and the specific analysis process is as follows: the appearance quality assessment value corresponding to the standard quartz crystal resonator is compared with the appearance quality assessment value interval corresponding to each quartz crystal resonator defect type in the database. If the appearance quality assessment value corresponding to the standard quartz crystal resonator is within the appearance quality assessment value interval corresponding to a quartz crystal resonator defect type in the database, the quartz crystal resonator defect type in the database is recorded as the defect type corresponding to the standard quartz crystal resonator.

[0056] And the standard quartz crystal resonator is classified into the corresponding defect type NG material box according to the corresponding defect type.

[0057] It should be noted that after the appearance quality evaluation value of the standard quartz crystal resonator is calculated, the system matches it to the predefined defect type intervals (such as cracks, scratches, etc.) in the database in real time, and follows the priority rules of severe defects > oxidation > scratches > cracks > others to handle the interval overlap problem. Based on the MODBUS RTU protocol, the control system drives the robot arm or solenoid valve to push the product to the corresponding NG material box (such as CRACK_NG, SCRATCH_NG), and the push action response time is less than 50ms / piece. At the same time, the system records the evaluation value, defect label and detection time of each product, generates a daily report to statistically analyze the defect distribution, and dynamically adjusts the interval parameters through the misclassification feedback mechanism to ensure that the classification accuracy is ≥98%.

[0058] Embodiments of the present invention Figure 2 As shown, a small quartz crystal resonator appearance defect AI detection equipment includes: a product loading and posture adjustment module: used to analyze the product posture corresponding to each quartz crystal resonator after each quartz crystal resonator in the current batch of the target enterprise is transmitted to the detection equipment, and record each quartz crystal resonator that meets the standard as a standard quartz crystal resonator, and then adjust the posture of each standard quartz crystal resonator.

[0059] Image acquisition and preprocessing module: used to automatically transmit the front of each standard quartz crystal resonator to the No. 1 camera position to obtain the front quality parameters corresponding to the front of each standard quartz crystal resonator, and then automatically transmit each standard quartz crystal resonator to the No. 2 camera position to obtain the back quality parameters corresponding to the back of each standard quartz crystal resonator.

[0060] Defect detection and identification module: It is used to identify whether the appearance quality of each standard quartz crystal resonator is qualified according to the front quality parameters corresponding to the front of each standard quartz crystal resonator and the back quality parameters corresponding to the back of the resonator. If the appearance quality of a standard quartz crystal resonator is unqualified, the standard quartz crystal resonator will be classified into the corresponding NG material box.

[0061] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined in this specification, they shall all fall within the protection scope of the present invention.

Claims

1. A small quartz crystal resonator appearance defect AI detection method, characterized in that: include: Step 1: Product loading and posture adjustment: After the quartz crystal resonators in the current batch of the target enterprise are transmitted to the testing equipment, the product postures corresponding to the quartz crystal resonators are analyzed, and the quartz crystal resonators that meet the standards are recorded as standard quartz crystal resonators, and the postures of the standard quartz crystal resonators are adjusted; Step 2, image acquisition and preprocessing: when the front side of each standard quartz crystal resonator is automatically transmitted to the first camera position, the front side quality parameters corresponding to the front side of each standard quartz crystal resonator are obtained, and each standard quartz crystal resonator is automatically transmitted to the second camera position again to obtain the back side quality parameters corresponding to the back side of each standard quartz crystal resonator; Step 3: Defect detection and identification: According to the front quality parameters corresponding to the front of each standard quartz crystal resonator and the back quality parameters corresponding to the back of the back, identify whether the appearance quality of each standard quartz crystal resonator is qualified. If the appearance quality of a standard quartz crystal resonator is unqualified, classify the standard quartz crystal resonator into the corresponding NG material box.

2. The AI ​​detection method for appearance defects of a small quartz crystal resonator according to claim 1, characterized in that: The product posture corresponding to each quartz crystal resonator is analyzed, and the specific analysis process is as follows: A1. After each quartz crystal resonator in the current batch of the target enterprise is transmitted to the testing equipment, the photoelectric sensor of the hopper of the testing equipment detects the presence of the quartz crystal resonator, triggers the vibration motor to start, and then automatically sets the vibration motor vibration frequency corresponding to each quartz crystal resonator. The vibration motor vibration frequency is calculated according to the formula: ； A2. When each quartz crystal resonator falls into the vibration plate in the detection equipment through the vibration of the vibration motor, two sets of Keyence through-beam fiber optic sensors are installed at the discharge port of the vibration plate in the detection equipment. When each quartz crystal resonator passes through the detection area of ​​the two sets of Keyence through-beam fiber optic sensors, the Keyence through-beam fiber optic sensors will receive two sets of corresponding optical signals and convert them into electrical signals for output. The signal strengths of the two sets of Keyence through-beam fiber optic sensors corresponding to each quartz crystal resonator are recorded as and ; A3. The signal strength of the two Keyence fiber optic sensors corresponding to each quartz crystal resonator is and The signal strengths of the two Keyence fiber optic sensors corresponding to a quartz crystal resonator are compared with those of the preset quartz crystal resonator. and If the signal strengths of the two Keyence fiber optic sensors corresponding to a quartz crystal resonator are both greater than the preset signal strengths of the quartz crystal resonator, the quartz crystal resonator is judged to be positive. On the contrary, if the signal strengths of the two Keyence fiber optic sensors corresponding to a quartz crystal resonator are greater than the preset signal strengths of the quartz crystal resonator, the quartz crystal resonator is judged to be positive. and If not all of them are greater than the signal strength corresponding to the preset quartz crystal resonator, it is determined that the quartz crystal resonator is the back side; A4. The signal strength of the two Keyence fiber optic sensors corresponding to each quartz crystal resonator is and Perform difference calculation, the signal strength difference is equal to and The absolute value of the difference is calculated, and the signal strength difference output by the two groups of Keyence opposed-beam optical fiber sensors corresponding to each quartz crystal resonator is compared with the signal strength difference corresponding to the preset quartz crystal resonator. If the signal strength difference output by the two groups of Keyence opposed-beam optical fiber sensors corresponding to a certain quartz crystal resonator is greater than or equal to the signal strength difference corresponding to the preset quartz crystal resonator, the quartz crystal resonator is determined to be longitudinal. On the contrary, if the signal strength difference output by the two groups of Keyence opposed-beam optical fiber sensors corresponding to a certain quartz crystal resonator is less than the signal strength difference corresponding to the preset quartz crystal resonator, the quartz crystal resonator is determined to be transverse. A5. If a quartz crystal resonator is facing forward, it is directly transmitted to the glass plate in the detection equipment. If a quartz crystal resonator is facing backward or vertically, the blowing pattern corresponding to the quartz crystal resonator is analyzed, and the quartz crystal resonator is blown to the bottom of the vibration plate according to the corresponding blowing pattern, so that the quartz crystal resonator faces forward and horizontally, and then transmitted to the glass plate in the detection equipment. The quartz crystal resonators corresponding to the glass plate transmitted to the detection equipment are judged as quartz crystal resonators that meet the standards.

3. The AI ​​detection method for appearance defects of a small quartz crystal resonator according to claim 2, characterized in that: The specific analysis process of analyzing the blowing mode corresponding to the quartz crystal resonator is as follows: B1. Obtain the model corresponding to the current batch of quartz crystal resonators, and compare the model corresponding to the current batch of quartz crystal resonators with the model corresponding to the back blowing mode of each quartz crystal resonator in the database. If the model corresponding to the current batch of quartz crystal resonators is the same as the model corresponding to the back blowing mode of a quartz crystal resonator in the database, then the back blowing mode of the quartz crystal resonator in the database is recorded as the back blowing mode corresponding to the quartz crystal resonator; B2. Compare the model corresponding to the current batch of quartz crystal resonators with the model corresponding to the longitudinal blowing mode of each quartz crystal resonator in the database. If the model corresponding to the current batch of quartz crystal resonators is the same as the model corresponding to the longitudinal blowing mode of a quartz crystal resonator in the database, then the longitudinal blowing mode of the quartz crystal resonator in the database is recorded as the longitudinal blowing mode corresponding to the quartz crystal resonator.

4. The AI ​​detection method for appearance defects of a small quartz crystal resonator according to claim 3, characterized in that: The specific adjustment process of adjusting the posture of each standard quartz crystal resonator is as follows: After each standard quartz crystal resonator is transferred to the guide module through the glass plate in the detection equipment, its initial position is detected by a high-precision position sensor, including information in the lateral, longitudinal and height directions. The control system drives the mechanical structure of the guide module, aligns the vacuum nozzle with a diameter of 3mm with the product center of each standard quartz crystal resonator, starts the vacuum system to generate adsorption force, and the pressure sensor monitors the adsorption force in real time, and stably controls it within the range of 0.05-0.1MPa. The electrostatic generator model Simco-Ion7850 is started to apply 2kV static electricity on the surface of the guide rail. The product posture is monitored in real time by a high-precision visual inspection system, and its lateral deviation value relative to the ideal posture is calculated. According to the deviation value, the control system adjusts the static distribution and intensity in real time, corrects the product posture by Coulomb force, and repeats the detection and adjustment steps until the lateral deviation of the product is less than 0.05mm. When the product posture fully meets the standard, the vacuum system and the electrostatic generator are turned off, and each standard quartz crystal resonator is released, and each standard quartz crystal resonator is transported to the camera inspection area for appearance defect detection.

5. The AI ​​detection method for appearance defects of a small quartz crystal resonator according to claim 4, characterized in that: The front quality parameters and back quality parameters corresponding to the front of each standard quartz crystal resonator are obtained, and the specific acquisition process is as follows: C1. When the front of each standard quartz crystal resonator is transmitted to the No. 1 camera position, a ring LED light source is used to illuminate the image at a 45° angle. After Gaussian filtering and noise reduction, the image is obtained by pixel-by-pixel grayscale difference. , pixel grayscale difference Greater than 50 marks reflective area, poor grayscale Less than 50 marks the non-reflective area, separates the reflective and non-reflective areas: extracts the gradient amplitude in the reflective area And reflective area , Otsu threshold segmentation is applied to calculate the average grayscale in the non-reflective area , defect area and circularity At the same time, the reflective area uses Canny edge detection to detect the pin outline, and combines sub-pixel edge detection to calculate the pin position deviation = / , Refers to the position offset on the horizontal X-axis. Refers to the position offset on the vertical Y axis; and based on the grayscale co-occurrence matrix, the low grayscale area ratio is analyzed to quantify the degree of oxidation. The final output includes the grayscale difference corresponding to the front of each standard quartz crystal resonator. , gradient amplitude , Reflective area , average grayscale , defect area , circularity , Pin position deviation and oxidation area ratio Key parameters, among which, Indicates the number corresponding to each standard quartz crystal resonator, , u is a positive integer; C2. When the back of each standard quartz crystal resonator is transmitted to the second camera position, a ring LED light source is used to illuminate the image at a 45° angle. After Gaussian filtering and noise reduction, the image is obtained by pixel-by-pixel grayscale difference. , pixel grayscale difference Greater than 50 marks reflective area, poor grayscale Less than 50 marks the non-reflective area, separates the reflective and non-reflective areas: extracts the gradient amplitude in the reflective area And reflective area , Otsu threshold segmentation is applied to calculate the average grayscale in the non-reflective area , defect area and circularity At the same time, the reflective area uses Canny edge detection to detect the pin outline, and combines sub-pixel edge detection to calculate the pin position deviation = / , Refers to the position offset on the horizontal X-axis. Refers to the position offset on the vertical Y axis; and based on the grayscale co-occurrence matrix, the low grayscale area ratio is analyzed to quantify the degree of oxidation. The final output includes the grayscale difference corresponding to the back of each standard quartz crystal resonator. , gradient amplitude , Reflective area , average grayscale , defect area , circularity , Pin position deviation and oxidation area ratio Key parameters.

6. The AI ​​detection method for appearance defects of a small quartz crystal resonator according to claim 5, characterized in that: The specific analysis process of identifying whether the appearance quality of each standard quartz crystal resonator is qualified is as follows: Analyze the appearance quality assessment value corresponding to each standard quartz crystal resonator, and compare the appearance quality assessment value corresponding to each standard quartz crystal resonator with the appearance quality assessment value corresponding to the set standard quartz crystal resonator. If the appearance quality assessment value corresponding to a standard quartz crystal resonator is greater than or equal to the appearance quality assessment value corresponding to the set standard quartz crystal resonator, then the appearance quality of the standard quartz crystal resonator is identified as unqualified. Otherwise, if the appearance quality assessment value corresponding to a standard quartz crystal resonator is less than the appearance quality assessment value corresponding to the set standard quartz crystal resonator, then the appearance quality of the standard quartz crystal resonator is identified as qualified.

7. The AI ​​detection method for appearance defects of a small quartz crystal resonator according to claim 6, characterized in that: The analysis of the appearance quality evaluation values ​​corresponding to each standard quartz crystal resonator is as follows: Obtain the front defect evaluation value and the back defect evaluation value corresponding to each standard quartz crystal resonator, and record the front defect evaluation value and the back defect evaluation value corresponding to each standard quartz crystal resonator as and ,in, Indicates the number corresponding to each standard quartz crystal resonator, , u is a positive integer, substitute into the calculation formula: The appearance quality evaluation value corresponding to each standard quartz crystal resonator is obtained. ,in, , They are the standard front defect evaluation value and the standard back defect evaluation value corresponding to the set quartz crystal resonator, , They are the weight factors corresponding to the set front defect evaluation value of the quartz crystal resonator and the weight factors corresponding to the standard back defect evaluation value, respectively. is a natural constant.

8. The AI ​​detection method for appearance defects of a small quartz crystal resonator according to claim 7, characterized in that: The specific acquisition process of obtaining the front defect evaluation value and the back defect evaluation value corresponding to each standard quartz crystal resonator is as follows: D1, the grayscale difference corresponding to the front of each standard quartz crystal resonator , gradient amplitude , Reflective area , and the grayscale difference corresponding to the back of each standard quartz crystal resonator , gradient amplitude , Reflective area As input information, they are respectively entered into the reflective area evaluation value analysis model. After calculation and analysis by the reflective area evaluation value analysis model, the front reflective area evaluation value corresponding to each standard quartz crystal resonator is finally output. and the back reflective area evaluation value ; D2. The average grayscale corresponding to the front of each standard quartz crystal resonator , defect area , circularity , and the corresponding average grayscale on the back of each standard quartz crystal resonator , defect area , circularity As input information, they are entered into the non-reflective area evaluation value analysis model respectively. After calculation and analysis by the non-reflective area evaluation value analysis model, the front non-reflective area evaluation value corresponding to each standard quartz crystal resonator is finally output. and the back non-reflective area evaluation value ; D3. The pin position deviation corresponding to the front side of each standard quartz crystal resonator and oxidation area ratio , and the corresponding pin position deviation on the back of each standard quartz crystal resonator and oxidation area ratio As input information, they are entered into the pin evaluation value analysis model respectively. After calculation and analysis by the pin evaluation value analysis model, the positive pin evaluation value corresponding to each standard quartz crystal resonator is finally output. and back pin evaluation value ; D4. The front reflective area evaluation value corresponding to the front of each standard quartz crystal resonator , Evaluation value of the front non-reflective area and positive pin evaluation value , and the corresponding back reflective area evaluation value of each standard quartz crystal resonator , Evaluation value of non-reflective area on the back and back pin evaluation value As input information, the defect evaluation value analysis model is used respectively. After calculation and analysis by the defect evaluation value analysis model, the positive defect evaluation value corresponding to each standard quartz crystal resonator is finally output. and back defect evaluation value .

9. The AI ​​detection method for appearance defects of a small quartz crystal resonator according to claim 8, characterized in that: If the appearance quality of a standard quartz crystal resonator is unqualified, the standard quartz crystal resonator will be classified into the corresponding NG material box. The specific analysis process is as follows: Compare the appearance quality evaluation value corresponding to the standard quartz crystal resonator with the appearance quality evaluation value interval corresponding to each quartz crystal resonator defect type in the database; if the appearance quality evaluation value corresponding to the standard quartz crystal resonator is within the appearance quality evaluation value interval corresponding to a quartz crystal resonator defect type in the database, then record the quartz crystal resonator defect type in the database as the defect type corresponding to the standard quartz crystal resonator; And the standard quartz crystal resonator is classified into the corresponding defect type NG material box according to the corresponding defect type.

10. A small quartz crystal resonator appearance defect AI detection device that executes the small quartz crystal resonator appearance defect AI detection method according to any one of claims 1 to 9, characterized in that: include: Product loading and posture adjustment module: used to analyze the product postures of each quartz crystal resonator after each quartz crystal resonator in the current batch of the target enterprise is transmitted to the testing equipment, and record each quartz crystal resonator that meets the standards as a standard quartz crystal resonator, and then adjust the posture of each standard quartz crystal resonator; Image acquisition and preprocessing module: used to automatically transfer the front of each standard quartz crystal resonator to the first camera position, obtain the front quality parameters corresponding to the front of each standard quartz crystal resonator, and automatically transfer each standard quartz crystal resonator to the second camera position to obtain the back quality parameters corresponding to the back of each standard quartz crystal resonator; Defect detection and identification module: It is used to identify whether the appearance quality of each standard quartz crystal resonator is qualified according to the front quality parameters corresponding to the front of each standard quartz crystal resonator and the back quality parameters corresponding to the back of the resonator. If the appearance quality of a standard quartz crystal resonator is unqualified, the standard quartz crystal resonator will be classified into the corresponding NG material box.

Images