WO2020082714A1 - Laser total reflection-type 3c transparent component defect detection apparatus and method - Google Patents

Abstract

A laser total reflection-type 3C transparent component defect detection apparatus and method, the apparatus comprising a positioning drive apparatus, a laser emitter (1), a light-transmissive work platform (3), a detection camera, a deep learning operation unit, an ARM embedded controller, and an acousto-optic warning device, the laser emitter (1) being connected to the positioning drive apparatus, the positioning drive apparatus, the detection camera, the deep learning operation unit, and the acousto-optic warning device having a communication connection with the ARM embedded controller by means of a CAN boss, and the detection camera photographing images of the light-transmissive work platform (3).

Description

Laser total reflection type 3C transparent member defect detection device and method Technical field

The invention belongs to the technical field of product surface defect detection, in particular to a 3C industry transparent component product quality defect detection device and method.

Background technique

China is a big country of 3C product manufacturing, and the application of transparent components in the industry is also increasing, and the quality requirements of products are becoming higher and higher. However, for the detection of 3C transparent component defects, most of them still remain in the stage of relying on artificial naked eye recognition. There is a large labor intensity and serious optical pollution, which hurts the eyesight of the inspectors, and due to the inconsistency of the inspectors' experience, there is a missed inspection risks of. In addition, China's labor resources are depleted and costs continue to rise. There is an urgent need for the intelligent upgrade of 3C industry testing equipment to reduce the probability of part defects and improve the yield rate, thereby increasing corporate profits.

At present, most of the defect detection of transparent purchased parts on the market is mainly manual, a small number of automated detection equipment, mainly optical detection, through a one-time whole product illumination (visible light mainly), and then use the camera to collect images, Analyze and judge whether there is a defect, but because the 3C transparent member has a small size and the defect is not obvious, it is detected by conventional optical principles, and its recognition accuracy rate needs to be further improved. Therefore, there is an urgent need for a new detection method to improve the signal-to-noise ratio, make it easier to identify subtle product defects, improve the yield of related industries and reduce the cost, and thus contribute to the development of the 3C industry.

Summary of the invention

The technical problem to be solved by the present invention is to provide a laser total reflection type 3C transparent member defect detection device with stable detection performance, high detection quality and high detection efficiency.

In order to solve the above technical problems, the present invention adopts the following technical solutions:

A laser total reflection type 3C transparent member defect detection device, which includes a positioning drive device, a laser emitter, a light-transmitting workbench, a detection camera, a deep learning arithmetic unit, an ARM embedded controller, and an audible and visual alarm, The laser emitter is connected to the positioning drive device. The positioning drive device, the detection camera, the deep learning operation unit and the sound and light alarm device are connected to the ARM embedded controller through the CAN bus, and the detection camera is shot against the light-transmitting working machine to obtain the image.

The light-transmitting workbench is also provided with a total reflection upper auxiliary part and a total reflection lower auxiliary part for clamping the transparent member to be tested. The total reflection lower auxiliary part is provided on the light-transmitting workbench, and the transparent member to be tested is provided on the full There is an exposed gap portion between the upper reflective auxiliary member and the total reflective lower auxiliary member, and the total reflective upper auxiliary member and the transparent member to be measured.

The positioning drive device includes a microcontroller, an X-direction motor, an X-direction positioning screw, a Y-direction motor, a Y-direction positioning screw, an X-direction load platform, a Y-direction load platform and an angular positioning stepper motor, and an X-direction positioning wire The rod is connected to the X-direction motor and mounted on the X-direction load platform. The Y-direction load platform is screwed onto the X-direction positioning screw. The Y-direction screw is connected to the Y-direction motor and mounted on the Y-direction load platform. The positioning stepping motor is mounted on the Y-positioning screw through a connecting block, the laser emitter is connected to the drive shaft of the angular positioning stepping motor, and the X-direction motor, Y-direction motor and angular positioning stepping motor are respectively connected to the microcontroller. The microcontroller communicates with the ARM embedded controller through the CAN bus, and a grating ruler is connected to the microcontroller.

One end of the X-direction load platform and the Y-direction load platform are respectively provided with a zero position sensor in communication connection with a microcontroller.

The total reflection upper auxiliary member, the transparent member to be detected, and the total reflection lower auxiliary member meet the requirements of total reflection of laser light, that is, the incident angle C of the laser light incident on the transparent member to be detected satisfies

C≥sin ^-1 (n ₂ / n ₁ ),

Where n ₂ is the refractive index of the total reflection upper auxiliary member and total reflection lower auxiliary member, and n ₁ is the refractive index of the transparent member to be measured.

The laser emitter is provided with at least three laser heads of different specifications side by side for measuring transparent members to be measured with different thicknesses.

The detection camera is arranged below the light-transmitting workbench, and a sheet metal shell is arranged on the periphery of the detection camera.

A laser total reflection type 3C transparent component defect detection method, including the following steps:

Place the transparent member to be tested between the total reflection upper auxiliary part and the total reflection lower auxiliary part and place it on the light-transmitting workbench;

Move the laser emitter to a predetermined position, rotate and adjust the angle of the laser emitter, position the laser emitter to the state to be detected, then the laser emitter continuously emits laser light toward the transparent member to be tested, and the laser light is incident into the transparent member to be tested.

Detect the camera to take a picture to obtain an image, transfer the image to the ARM embedded controller for image preprocessing, and then transfer the image to the deep learning arithmetic unit to automatically identify the strength and position of the light leakage, thus automatically detecting the type of defect of the component to be detected And its location.

The refractive index n ₂ of the total reflection upper auxiliary member and the total reflection lower auxiliary member is 1/2 or less than the refractive index n ₁ of the transparent member to be measured, so that the laser incident angle C of the transparent member to be measured is less than 45 degrees .

The transparent member to be tested is a flat 3C transparent member or a curved 3C transparent member.

The invention has the following beneficial effects:

1) Based on the principle of optical total reflection, the 3C transparent member to be inspected is detected using a visible laser with strong directivity. Due to defects in the transparent member or on the surface, the laser will be refracted, reflected, and scattered, so that part of the laser does not meet the conditions of total reflection. , Leaking from the surface of the transparent member, captured by the detection camera, and taking pictures to provide basic data for later intelligent analysis.

2) In order to detect 3C transparent purchases of different thicknesses, laser heads of different specifications and sizes are used, so that automatic detection can take into account both the detection quality and the detection efficiency.

3) Using the principle of total reflection and its laser positioning auxiliary device, the patent of the present invention can not only detect the flat type 3C transparent member, but also detect the curved type 3C transparent member.

4) Using deep learning 3C transparent component intelligent detection technology to complete the precise and online automatic detection of transparent components, which overcomes the inconsistency of manual detection and makes the quality stable during the detection process.

BRIEF DESCRIPTION

Figure 1 is a schematic diagram of the connection principle of the device of the present invention;

2 is a schematic structural view of the positioning drive device of the present invention;

Figure 3 is a schematic diagram of the connection principle of the positioning drive device of the present invention;

Figure 4 is a schematic structural view of the light-transmitting workbench of the present invention;

Figure 5-1 is a schematic diagram of the optical path of the component to be tested without defects;

Figure 5-2 is a schematic diagram of the optical path with defects on the surface of the component to be tested;

Figure 5-3 is a schematic diagram of the optical path with defects inside the component to be tested;

Figure 6-1 is a schematic diagram of the optical path of the planar type test component;

Figure 6-2 is a schematic diagram of the optical path refraction of a curved transparent member to be measured;

Figures 7-1, 7-2 and 7-3 are schematic diagrams of the structure of the deep learning network;

Figure 8 is a schematic diagram of the principle of detection area division.

detailed description

To further understand the features, technical means, and specific objectives and functions of the present invention, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

As shown in FIGS. 1-4, the present invention discloses a laser total reflection type 3C transparent member defect detection device, which includes a positioning driving device, a laser emitter 1, a light-transmitting workbench 3, a detection camera, a depth Learning operation unit, ARM embedded controller and audible and visual alarm, laser emitter is connected with positioning drive device, positioning drive device, inspection camera, deep learning operation unit and sound and light alarm device communicate with ARM embedded controller via CAN bus Connect and detect the camera to take the image against the light-transmitting work machine. The positioning drive device positions the laser emitter so that the laser emitter emits laser light at an accurate position, ensuring that the laser light accurately enters the transparent member to be measured.

The light-transmitting workbench 3 is also provided with a total reflection upper auxiliary part 41 and a total reflection lower auxiliary part 42 for clamping the transparent member to be tested. The total reflection lower auxiliary part 41 is provided on the light-transmitting workbench 3 and is to be tested The transparent member 5 is provided between the total reflection upper auxiliary member 41 and the total reflection lower auxiliary member 41, and there is a bare gap portion between the total reflection upper auxiliary member 41 and the transparent member 5 to be measured. The materials of the total reflection upper auxiliary part and the total reflection lower auxiliary part are made of flexible wear-resistant light-transmitting materials, and the hardness is lower than that of the 3C transparent member to be tested; different types of supporting total reflection upper auxiliary member and total reflection Lower auxiliary parts, so as to achieve full coverage of the 3C transparent member to be inspected by optical inspection. The total reflection upper auxiliary member, the transparent member to be detected, and the total reflection lower auxiliary member meet the requirements of total reflection of laser light, that is, the incident angle C of the laser light incident on the transparent member to be detected satisfies

C≥sin ^-1 (n ₂ / n ₁ ),

Where n ₂ is the refractive index of the total reflection upper auxiliary member and total reflection lower auxiliary member, and n ₁ is the refractive index of the transparent member to be measured.

In addition, the positioning drive device includes a microcontroller, an X-direction motor 6, an X-direction positioning screw 7, a Y-direction motor 11, a Y-direction positioning screw 12, an X-direction load platform 8, a Y-direction load platform 9, and angular positioning Stepper motor 2, X-direction screw 7 is connected to X-direction motor 6 and mounted on X-direction load platform 8, Y-direction load platform 9 is screwed on X-direction screw 7 and Y-direction screw 12 Connected to the Y-direction motor 11 and mounted on the Y-direction load platform 9, the angular positioning stepper motor 2 is mounted on the Y-direction screw 12 through a connecting block, and the laser emitter 1 is connected to the drive shaft of the angular positioning stepper motor 2 , X-direction motor, Y-direction motor, angle positioning stepper motor are respectively connected to the microcontroller. The microcontroller is connected to the ARM embedded controller via CAN bus. The microcontroller is connected to the grating scale, the X-direction load platform, One end of the Y-direction load platform is also provided with zero sensors in communication connection with the microcontroller. The X-direction motor drives the X-direction positioning screw to rotate, which in turn can adjust the movement stroke of the Y-direction load platform in the X direction, thereby adjusting the position of the laser emitter in the X direction. Then adjust the position of the laser emitter in the Y direction by the Y-direction motor, so as to adjust the position of the XY direction, and then rotate the laser emitter through the angular positioning stepper motor to adjust the angle, so that the laser emitter is adjusted to the predetermined position. The grating ruler can accurately control the travel of X-direction motor and Y-direction motor. The X-direction motor, Y-direction motor and grating ruler form a closed-loop control loop. The stepper motor uses an angle zero sensor to automatically return to zero correction after each transparent component to be tested is completed, thereby providing an environment for high-precision positioning.

The X-direction motor pauses after a fixed step size at each interval (1/5 to 1/10 of the width of the 3C transparent member to be tested). During the movement of the X-direction motor, the laser emitter continuously emits laser light to detect the time of the camera during this detection process The camera is in the exposure state. After the fixed step in the X direction moves into place, the camera is detected to take pictures. The angular positioning stepper motor automatically returns to the original position after each piece to be detected is detected by the angle zero sensor, which provides a reference for the angular positioning stepper motor recalibration; the microcontroller communicates with the ARM embedded controller through the CAN bus Connect, sense the information of the detected component, and provide parameters for its servo and positioning motion.

The angular positioning stepper motor drives the laser emitter to rotate, and the angle of rotation is adjusted by the ARM embedded controller according to the size information of the member to be inspected and its position to be inspected, so that the inspected laser can meet the total reflection condition for the region to be inspected.

The laser emitter is provided with three laser heads of different specifications side by side for measuring transparent members to be measured with different thicknesses. In this embodiment, three laser heads with different specifications of 1.5mm × 1.5mm, 3.5mm × 3.5mm and 6mm × 6mm respectively detect 3C transparent members with different thickness specifications less than 1mm, 1-3mm and 3-5mm .

Based on the size information of the transparent component to be tested, the ARM embedded controller samples according to the length and width direction of the component, and completes the detection of the entire component by taking multiple photos (length sampling times × width sampling times).

In addition, the whole device is sealed by sheet metal parts to reduce the interference of the external light source on the detection results; further, the detection camera is placed under the light-transmitting workbench, and a set of sheet metal shells are set on the periphery to seal, again reducing interference.

The deep learning arithmetic unit is implemented with FPGA hardware. Before being passed in, the ARM embedded controller performs graying and segmentation. According to the specifications of the laser head, each image after segmentation is compressed to 64 × 64 × 1. In the three-dimensional grayscale image space of 128 × 128 × 1 or 256 × 256 × 1, after performing three more convolution and pooling operations, perform two fully connected neural networks and output to a 256-dimensional vector. The 256-dimensional vector is output as a vector (normal, defective) through soft regression, and the detection result is transmitted to the ARM embedded controller through the CAN bus communication module, and the detection personnel is informed by an audible and visual alarm.

The sample inventory that the deep learning arithmetic unit depends on is in the ARM embedded controller (internal NAND Flash), and the deep convolutional neural network parameters can be updated in the background by the ARM embedded controller and sent to the deep learning arithmetic unit through the CAN bus Store and use in operations.

The offline training sample library of the deep convolutional neural network can increase the number of samples. Therefore, the detection of 3C transparent components can be increased or decreased according to the actual situation of the sample, and the detection accuracy of the components to be tested with specific specifications can be improved.

The deep convolutional neural network can be trained and updated by the user during use, or can be periodically updated by the device manufacturer; the device of the present invention supports multiple versions of the deep convolutional neural network, which can be performed by the end user according to the actual application scenario Independent choice.

The servo motor in the XY direction drives the positioning screw to move in the X and Y directions, drives the load platform to link in the "XY" direction, the angular positioning stepper motor is installed on the load platform, and then drives the laser transmitter to rotate through the three axes The linkage scheme can position the laser transmitter to the state to be detected to meet the detection requirements. During the detection process, the servo motor in the X direction is paused after a fixed step length (1/5 to 1/10 of the width of the 3C transparent member to be detected). During this process, the laser transmitter continuously emits laser light and the detection camera is exposed In the state, after the fixed step in the X direction moves into place, the detection camera takes a picture; by selecting the material of the auxiliary component on the total reflection, the refractive index n ₂ is 1/2 or less of the refractive index n ₁ of the transparent member to be detected, so as to achieve The laser incident angle C is less than 45 degrees, and then select the laser head of the corresponding specification according to the thickness of the transparent member to be detected, and then it can be covered by one scan in the Y direction. The specific rules are as follows:

a) For the case where the thickness of the transparent member to be tested is less than 1 mm, the positioning drive device microcontroller selects a 1.5 mm × 1.5 mm laser head to work;

b) For the case where the thickness of the transparent member to be tested is between 1 and 3 mm, a 3.5 mm × 3.5 mm laser head is selected for operation by the microcontroller of the positioning drive device;

c) For the case where the thickness of the transparent member to be detected is 3 to 5 mm, a 6 mm × 6 mm laser head is selected for operation by the microcontroller of the positioning drive device.

d) The microcontroller of the above positioning drive device is controlled by the ARM embedded controller through the CAN bus, and the user can choose according to the actual situation.

After the above-mentioned area is sampled and moved by one unit of laser head width in the Y direction, the detection is repeated again, and the two detections can completely cover the complete area within a fixed step size. In the same way, repeat the above process for the remaining area to be tested, and continue to scan and test in the X direction to cover the entire area of the transparent member to be tested to complete the detection. The detection principle is shown in Figure 8, the first detection in Figure 8 And the second detection is the two detections along the Y direction in the sampling area. Therefore, the total number of photographs is 2 * (5-10) times.

As shown in Figure 4, the detection laser beam is controlled by the ARM embedded controller, and the position of the detection laser beam is adjusted through the positioning drive device so that the principle of total reflection is satisfied; the auxiliary part on the total reflection and the 3C transparent member to be detected There is a gap, usually 1.2 to 1.5 times the width of the laser beam, so that the laser beam can enter the inside of the transparent member to be detected; under the total reflection, there is a detection camera under the auxiliary part, which may exist (defective detection sees light leakage Phenomenon) to take photos of light leakage.

As shown in Figure 5-1, for the 3C transparent member to be inspected, if there are no defects inside or on the surface, the detection laser beam satisfies the working principle of total reflection. And the auxiliary component under total reflection, so that the detection camera disposed under the auxiliary component under total reflection cannot acquire the detection laser signal. On the contrary, as shown in Figure 5-2, there is a defect on the surface of the transparent member to be tested, as shown in Figure 5-3, there is a defect inside the transparent member to be detected, then a part of the detection laser beam reflects, refracts, and scatters, which cannot fully satisfy the full Reflection conditions, so that the detection camera placed under the auxiliary part under total reflection can capture the detection laser signal.

Figure 6 only shows a schematic diagram of the placement of planar transparent components. For curved transparent components, two sets of total reflection upper / lower auxiliary parts need to be configured to complete the detection of curved surfaces and flat surfaces respectively according to the detection object. Since the principles are the same, here No longer.

As shown in Figures 6-1 and 6-2, they are schematic diagrams of the optical paths of the planar transparent member and the curved transparent member. For a flat 3C transparent component, the laser head only needs to emit the detection laser beam into the transparent component to be detected at a fixed angle, and the entire component can be detected by moving the load table in the X direction; however, for The curved 3C transparent member needs to detect the plane part and the curved part separately. The ARM embedded controller calculates the incident angle of the laser beam on the plane and the curved part separately according to the geometric size information to be detected, and then the driving device microcontroller drives the servo The movement of the motor and stepper motor makes the laser head reach the calculated angle, especially the curved surface part, and the laser head with a small size is selected for detection.

The ARM embedded controller not only receives the information of the microcontroller, but also sends control instructions to the microcontroller to realize the positioning of the laser head angle, and during the detection process, the load platform drives the laser head to sample at intervals in the X direction, so that The detection camera can take pictures of the detection imaging.

In addition, a laser total reflection type 3C transparent member defect detection method includes the following steps:

The transparent member to be measured is placed between the total reflection upper auxiliary part and the total reflection lower auxiliary part and placed on the light-transmitting workbench.

Move the laser emitter to a predetermined position, rotate and adjust the angle of the laser emitter, position the laser emitter to the state to be detected, and then the laser emitter continuously emits laser light toward the transparent member to be tested, and the laser light is incident into the transparent member to be tested.

Detect the camera to take a picture to obtain an image, transfer the image to the ARM embedded controller for image preprocessing, and then transfer the image to the deep learning arithmetic unit to automatically identify the strength and position of the light leakage, thus automatically detecting the type of defect of the component to be detected And its location.

As shown in Figure 7-1, Figure 7-2, and Figure 7-3, the deep learning arithmetic unit is implemented with FPGA hardware. Prior to the introduction, the ARM embedded controller performs graying and segmentation processing. The specifications of the header, each divided image is compressed into a three-dimensional gray image space of 64 × 64 × 1, 128 × 128 × 1, or 256 × 256 × 1, and after three convolution and pooling operations, The neural network that is fully connected twice is output to a 256-dimensional vector, and then the 256-dimensional vector is output as a vector (normal, defective) through soft regression. The network data processing flow is as follows:

a) In order to improve the recognition accuracy, the three networks are separately trained offline, that is, corresponding to three different specifications of the laser head, the input images of the three networks are 64 × 64 × 1, 128 × 128 × 1, or 256 × 256, respectively. × 1 three-dimensional gray scale size.

b) The deep learning arithmetic unit can process multiple images at a time. Therefore, a batch of divided images can be set to input 32 images into the FPGA-implemented network at the same time, and the identification time can be reduced through concurrent processing.

c) For 64 × 64 × 1, the A1 convolutional layer sent to the convolutional network, after using the 3 × 3 window convolution operation, 12 images of 62 × 62 pixels are generated, and then the A2 pooling layer in the module Perform compression processing to generate 12 images of 31 × 31 pixels, and then perform the second convolution operation and send it to the A3 convolution layer in the convolution network. After using the 3 × 3 window convolution operation again, 29 × is generated. The 36 images of 29 pixels are compressed by the A4 pooling layer in the convolution network to generate 36 images of 14 × 14 pixels. After that, the third convolution operation is performed and sent to A5 in the convolution network. The convolutional layer uses the 3 × 3 window convolution operation again to generate 72 images of 12 × 12 pixels, and then performs compression processing by the A6 pooling layer in the convolution network to generate 72 images of 6 × 6 pixels. Further, the A7 fully connected layer of the convolutional network is processed to output 1024-dimensional vectors. Further, the A8 fully connected layer of the convolutional network is processed to output 256-dimensional vectors. Finally, the convolutional network A9 The soft regression layer outputs a 2-dimensional vector, indicating that the transparent component to be detected belongs to the category 2 (normal, defective). Density distribution, so that the transparent member identification to be examined for defects 3C.

d) For 128 × 128 × 1, the B1 convolutional layer sent to the convolutional network, using a 5 × 5 window convolution operation, generates 12 images of 124 × 124 pixels, and then the B2 pooling layer in the module Perform compression processing to generate 12 images of 62 × 62 pixels, and then perform the second convolution operation and send it to the B3 convolution layer in the convolution network. After using the 3 × 3 window convolution operation again, 60 × is generated. 36 images of 60 pixels, then compressed by the B4 pooling layer in the convolution network to generate 36 images of 30 × 30 pixels, after which, the third convolution operation is performed and sent to B5 in the convolution network The convolution layer uses the 3 × 3 window convolution operation again to generate 72 images of 28 × 28 pixels, which are then compressed by the B6 pooling layer in the convolutional network to generate 72 images of 14 × 14 pixels. Further, the B7 fully connected layer of the convolutional network is processed to output a 2048-dimensional vector. Further, the B8 fully connected layer of the convolutional network is processed to output a 256-dimensional vector. Finally, the B9 in the convolutional network The soft regression layer outputs a 2-dimensional vector, indicating that the transparent component to be detected belongs to category 2 (normal, defective) Probability density distribution, so that the transparent member identification to be examined for defects 3C.

e) For 256 × 256 × 1, the C1 convolutional layer sent to the convolutional network, after using the 5 × 5 window convolution operation, 12 images of 252 × 252 pixels are generated, and then the C2 pooling layer in the module Perform compression processing to generate 12 images of 126 × 126 pixels, and then perform the second convolution operation and send them to the C3 convolution layer in the convolution network. After using the 3 × 3 window convolution operation again, 124 × is generated. The 36 images of 124 pixels are compressed by the C4 pooling layer in the convolution network to generate 36 images of 62 × 62 pixels. After that, the third convolution operation is performed and sent to C5 in the convolution network. The convolution layer uses the 3 × 3 window convolution operation again to generate 72 images of 60 × 60 pixels, which are then compressed by the C6 pooling layer in the convolution network to generate 72 images of 30 × 30 pixels. Further, the C7 fully connected layer of the convolutional network is processed to output a 4096-dimensional vector. Further, the C8 fully connected layer of the convolutional network is processed to output a 256-dimensional vector. Finally, the C9 in the convolutional network The soft regression layer outputs a 2-dimensional vector, indicating that the transparent component to be detected belongs to category 2 (normal, missing) ) Is the probability density distribution, so that the transparent member identification to be examined for defects 3C.

Further, the offline training sample library of the convolutional network can increase the number of samples. Therefore, the accuracy of detection can be further improved with the increase of the number of samples; the convolutional network can also be trained and updated by the user during use, or can be periodically updated by the device manufacturer.

It should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art can still implement the foregoing The technical solutions described in the examples are modified, or some of the technical features are equivalently replaced, but any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the protection of the present invention Within range.

Claims (10)

1. A laser total reflection type 3C transparent member defect detection device, characterized in that the device includes a positioning drive device, a laser emitter, a light-transmitting workbench, a detection camera, a deep learning arithmetic unit, an ARM embedded controller and a sound The optical alarm and laser transmitter are connected to the positioning drive device. The positioning drive device, the detection camera, the deep learning arithmetic unit, and the sound and light alarm device are connected to the ARM embedded controller through the CAN bus, and the detection camera is shot against the light-transmitting work machine Get the image.
2. The laser total reflection type 3C transparent member defect detection device according to claim 1, characterized in that a total reflection upper auxiliary part and a total reflection lower part for clamping the transparent member to be tested are further provided on the light-transmitting worktable Auxiliary parts, the total reflection lower auxiliary part is set on the light-transmitting workbench, the transparent member to be measured is arranged between the total reflection upper auxiliary part and the total reflection lower auxiliary part, and there is bareness between the total reflection upper auxiliary part and the transparent member to be tested Gap section.
3. The laser total reflection type 3C transparent member defect detection device according to claim 2, wherein the positioning driving device comprises a microcontroller, an X-direction motor, an X-direction positioning screw, a Y-direction motor, and a Y-direction positioning Screw rod, X-direction load platform, Y-direction load platform and angle positioning stepper motor. The X-direction screw is connected to the X-direction motor and mounted on the X-direction load platform. The Y-direction load platform is screwed on the X-direction positioning wire On the rod, the Y-positioning screw is connected to the Y-direction motor and mounted on the Y-direction load platform. The angle positioning stepper motor is mounted on the Y-positioning screw through the connection block. The laser emitter and the angle positioning stepper motor are driven Axis connection, X-direction motor, Y-direction motor, and angular positioning stepper motor are connected to the microcontroller respectively. The microcontroller communicates with the ARM embedded controller via CAN bus, and the microcontroller is connected with a grating ruler.
4. The laser total reflection type 3C transparent member defect detection device according to claim 3, characterized in that one end of the X-direction load platform and the Y-direction load platform are respectively provided with a zero sensor in communication connection with a microcontroller .
5. The laser total reflection type 3C transparent member defect detection device according to claim 4, characterized in that the total reflection upper auxiliary member, the transparent member to be detected, and the total reflection lower auxiliary member satisfy laser total reflection between the three The requirement, that is, the incident angle C of the laser light incident on the transparent member to be detected satisfies

   C≥sin ^-1 (n ₂ / n ₁ ),

   Where n ₂ is the refractive index of the total reflection upper auxiliary member and total reflection lower auxiliary member, and n ₁ is the refractive index of the transparent member to be measured.
6. The laser total reflection type 3C transparent member defect detection device according to claim 5, wherein the laser emitter is provided with at least three laser heads of different specifications side by side for measuring transparent members to be measured with different thicknesses .
7. The laser total reflection type 3C transparent member defect detection device according to claim 6, characterized in that the detection camera is provided below the light-transmitting worktable, and a sheet metal shell is provided on the periphery of the detection camera.
8. A laser total reflection type 3C transparent component defect detection method, including the following steps:

   Place the transparent member to be tested between the total reflection upper auxiliary part and the total reflection lower auxiliary part and place it on the light-transmitting workbench;

   Move the laser emitter to a predetermined position, rotate and adjust the angle of the laser emitter, position the laser emitter to the state to be detected, then the laser emitter continuously emits laser light toward the transparent member to be tested, and the laser light is incident into the transparent member to be tested.

   Detect the camera to take a picture to obtain an image, transfer the image to the ARM embedded controller for image preprocessing, and then transfer the image to the deep learning arithmetic unit to automatically identify the strength and position of the light leakage, thus automatically detecting the type of defect of the component to be detected And its location.
9. The laser total reflection type 3C transparent member defect detection method according to claim 8, wherein the refractive index n ₂ of the total reflection upper auxiliary member and the total reflection lower auxiliary member is the refractive index n ₁ of the transparent member to be measured 1/2 or less than 1/2, so that the laser incident angle C of the transparent member to be measured is less than 45 degrees.
10. The laser total reflection type 3C transparent member defect detection method according to claim 9, wherein the transparent member to be tested is a flat 3C transparent member or a curved 3C transparent member.

Images