A multi-faceted appearance inspection apparatus and inspection calculation method

Abstract

This invention provides a multi-faceted appearance inspection device and inspection calculation method, including a transfer platform mounted on a machine base, a gantry frame reciprocating on the transfer platform, a hinged rotating beam on the gantry frame, a swing drive and a rotating fixture mounted on the rotating beam, and a shooting inspection mechanism mounted above the transfer platform. The shooting inspection mechanism includes a transverse module, a translation-driven lifting module on the transverse module, and a lifting plate on the lifting module. A camera is rotated on the lifting plate via a rotator. The lifting plate is also equipped with a bright field light source, a dark field light source, and a light-dark transition light source. The bright field light source is fixed directly below the camera, the dark field light source is tilted to one side of the camera, and the light-dark transition light source is tilted to the other side of the camera. The tilt angle between the dark field light source and the camera is smaller than the tilt angle between the light-dark transition light source and the camera. Combined with a multi-angle light source time-division strobe shooting method, the images are synthesized to obtain accurate and realistic appearance defect data, improving inspection accuracy.

Description

Technical Field

[0001] This invention belongs to the field of appearance inspection technology, and in particular to a multi-faceted appearance inspection device and inspection calculation method. Background Technology

[0002] With the increasing demands for stringent appearance standards and high production volumes in the 3C industry, defect detection of product casings is necessary. Traditional methods rely entirely on manual visual inspection, but prolonged repetitive work can lead to visual fatigue and susceptibility to external interference, making it difficult to guarantee inspection efficiency. Furthermore, individual judgments of defects vary, and relying primarily on visual inspection makes it difficult to establish quantifiable quality standards. Consequently, manual visual inspection cannot meet reliability requirements, exhibiting poor stability and accuracy, and its highly subjective nature makes the inspection results unreliable.

[0003] To address the aforementioned technical problems, for example, a Chinese patent document has disclosed an appearance inspection device [Chinese Patent No.: 202223307778.6]. This utility model relates to the field of inspection device technology, and particularly to an appearance inspection device. The appearance inspection device includes a transfer module, a first inspection mechanism, and a second inspection mechanism. The transfer module can sequentially transfer the product to the first inspection station and the second inspection station, and can rotate the product circumferentially at the second inspection station. The first inspection mechanism is located above the first inspection station and can inspect the top surface of the product passing through the first inspection station. The second inspection mechanism is located above the second inspection station and can inspect the side surfaces, rounded corners, and the rounded edge between the top surface and the side surfaces of the product. Through the first and second inspection mechanisms, automatic inspection of various positions of the product can be achieved, unifying the inspection standards for product appearance and improving inspection efficiency and accuracy.

[0004] The above technical solution only sets a coaxial light source below the line scan camera of the first inspection agency. The light source in a single direction is prone to causing shadows around the product under illumination, which affects the clarity and authenticity of the captured image, thereby reducing the authenticity and objectivity of the appearance defect detection, resulting in large error in the detection results and increasing the defect rate of finished products. Summary of the Invention

[0005] The purpose of this invention is to address the aforementioned problems in existing technologies by proposing a multi-faceted appearance inspection device and inspection calculation method that uses a time-division stroboscopic imaging method with multiple light sources to capture images from different angles, and then comprehensively analyzes and judges these images to obtain accurate and high-precision defect data.

[0006] The objective of this invention can be achieved through the following technical solution: A multi-faceted appearance inspection device includes a machine base, a transfer platform is provided on the machine base, the transfer platform reciprocates to drive a gantry frame, a rotating beam is hinged on the gantry frame, a swing drive and a rotating fixture are installed on the rotating beam, a shooting inspection mechanism is erected above the transfer platform, the shooting inspection mechanism includes a transverse module, a translation drive lifting module is provided on the transverse module, a lifting drive lifting plate is provided on the lifting module, a camera is rotated on the lifting plate by a rotator, and a bright field light source, a dark field light source and a light-dark transition light source are also installed on the lifting plate, the bright field light source is fixed directly below the camera, the dark field light source is inclined on one side of the camera, the light-dark transition light source is inclined on the other side of the camera, and the angle of inclination between the dark field light source and the camera is smaller than the angle of inclination between the light-dark transition light source and the camera.

[0007] In the aforementioned multi-faceted appearance inspection equipment, the gantry frame includes two opposing vertical plates, and the rotating beam is hinged between the two vertical plates by a bearing. The swing drive is fixed to one end of the rotating beam, and the swing shaft of the swing drive is inserted into the hinge point of one of the vertical plates.

[0008] In the aforementioned multi-faceted appearance inspection equipment, the transfer platform includes a feeding servo motor, which is rotatably connected to a lead screw. A guide rail is arranged parallel to the side of the lead screw, and a nut block is threaded onto the outer circumference of the lead screw. The nut block is fixedly connected to a guide block, which is engaged with the guide rail to form a sliding connection. The guide block is fixedly connected to the vertical plate through a connecting plate.

[0009] In the aforementioned multi-faceted appearance inspection equipment, the rotary fixture table includes a rotator fixed on the rotating beam, the rotator is connected to a commutator, the commutator is connected to a fixture disk, and the fixture disk has a plurality of suction holes evenly distributed on it.

[0010] In the aforementioned multi-faceted appearance inspection equipment, a contour block is protruding from the center of the top surface of the fixture disk, and several suction cups are arranged around the periphery of the fixture disk.

[0011] In the aforementioned multi-faceted appearance inspection equipment, a truss is provided above the transfer platform, and the transverse module includes a horizontal electric cylinder fixedly mounted on the truss, with the translation block of the horizontal electric cylinder fixedly connected to the translation plate.

[0012] In the aforementioned multi-faceted appearance inspection equipment, the lifting module includes a lifting servo motor fixedly mounted on the translation plate. The lifting servo motor is connected to a lead screw via a synchronous belt drive. A vertical rail is arranged parallel to the side of the lead screw. A nut sleeve is threaded onto the outer circumference of the lead screw. The nut sleeve is fixedly connected to the lifting plate. A vertical block is correspondingly arranged on the lifting plate. The vertical block engages with the vertical rail to form a guide sliding connection.

[0013] A detection calculation method for a multi-faceted appearance inspection device includes the following steps:

[0014] 1) Place the product on the rotating fixture table to fix it. Start the transfer platform to move the gantry from the front end to the bottom of the shooting and inspection mechanism. Start the horizontal movement module to drive the camera, bright field light source, dark field light source and light-dark transition light source to move back and forth. Start the lifting module to drive the camera, bright field light source, dark field light source and light-dark transition light source to move up and down, so that the camera is located at the set height above the product.

[0015] 2) Start the swing drive to rotate the swing shaft, which will cause the rotating beam to rotate around the hinge point to 90°. Start the rotator to drive the jig disk to rotate. The product is in a vertical position with one side facing up. Turn on the camera to take pictures of several set points on one side of the product in sequence.

[0016] 3) Control the bright field light source, dark field light source, and light-dark transition light source to switch on and off at high speed in sequence. When the bright field light source, dark field light source, and light-dark transition light source illuminate the product, the camera captures bright field images, dark field images, and light-dark transition images at each set point. Collect all side images for target detection classification and localization. Then, fine-grained classification is used to classify minor defects and serious defects. At the same time, semantic segmentation is used to calculate the area. Then, the PS and HS measurement algorithms are used to obtain the area, quantity, and distance of defects. When the defect area, quantity, and distance are all less than or equal to the standard value, the product is judged to be qualified. When any of the area, quantity, and distance is greater than the standard value, the product is judged to be unqualified.

[0017] 4) Start the rotator to drive the jig plate to rotate, and sequentially switch the other side of the product to face upwards. Then turn on the camera again to take pictures of several set points on the other side of the product. Repeat step 3) to calculate the area, quantity and distance of the defects. Repeat the operation until all sides of the product have been photographed and judged.

[0018] 5) Start the swing drive to rotate the swing shaft, which will drive the rotating beam to rotate around the hinge point to 0°. The product will be in a horizontal position with the front facing up. Turn on the camera to take pictures of the front of the product. The camera will move along the set route in the X and Y directions of the product to take pictures.

[0019] 6) Control the bright field light source, dark field light source, and light-dark transition light source to switch on and off at high speed in sequence. When the bright field light source, dark field light source, and light-dark transition light source illuminate the product, the camera takes bright field images, dark field images, and light-dark transition images for each set route. Collect all front-facing images to perform multi-light source high-resolution image detection of minimal targets. Based on defects, randomly crop the defect area for training. Perform area calculation through semantic segmentation. Then, use a simple threshold to judge whether the product is qualified or unqualified.

[0020] 7) If the results of steps 5) and 6) are both qualified, the product is a good product; if at least one of them is unqualified, the product is a defective product.

[0021] In the above-mentioned detection calculation method of the multi-faceted appearance inspection equipment, in step 3), when defects overlap, the area, quantity and distance are judged by the bounding rectangle of the overlapping area; when the PS and HS measurement algorithms are shielded by the automatic optical inspection, the single-point bitmap registration algorithm is used to match the actual captured image with the standard image, and then the transformation matrix is ​​estimated to obtain the calibration image and the automatic optical inspection area.

[0022] In the detection calculation method of the multi-faceted appearance inspection equipment mentioned above, in step 2) or 5), during the camera shooting process, the rotator is started to drive the camera to rotate and adjust the lens angle.

[0023] Compared with existing technologies, this multi-faceted appearance inspection equipment and inspection calculation method have the following advantages:

[0024] 1. Based on three-dimensional translation and adjustment, two to three types of rotation and adjustment have been added to form five-axis linkage or six-axis linkage, which increases the shooting angle of the product and allows the product to be rotated and adjusted in multiple directions at the same inspection station. This enables multi-faceted inspection and shooting at a fixed point, reduces the inspection and transportation path, improves inspection efficiency, and saves equipment space and cost. At the same time, the use of single-point multi-faceted inspection avoids the accumulation of material transfer errors and improves the accuracy of inspection.

[0025] 2. In the six-axis linkage, the lens rotation adjustment itself has been added to prevent blind spots in shooting and improve the integrity and realism of the shooting.

[0026] 3. Five-axis linkage and six-axis linkage interpolate with each other to realize the conversion of light fields at different angles, enabling the defect detection equipment that acquires images from the side of the product and the line scan color image detection equipment to be connected in parallel, forming an integrated defect detection equipment that can detect the entire side and front of the product at one time.

[0027] 4. Detection is performed using optical principles, weighted imaging difference data analysis principles, color extraction methods, similarity principles, and binarization principles. Defects are identified through image comparison, and then displayed / marked on a monitor or automatically. A multi-angle light source time-division stroboscopic imaging method is used to capture images from different angles. These images are then combined through effective calculations to avoid errors caused by unilateral lighting shadows and blind spots, thereby obtaining accurate and realistic data on appearance defects and improving detection accuracy. Attached Figure Description

[0028] Figure 1 This is a three-dimensional structural diagram of a multi-faceted appearance inspection device.

[0029] Figure 2 This is the main view structure diagram of a multi-faceted appearance inspection device.

[0030] Figure 3 This is a structural diagram of the gantry in a multi-faceted appearance inspection device.

[0031] Figure 4 This is a schematic diagram of the principle of light source imaging in a multi-faceted appearance inspection device.

[0032] In the diagram, 1 is the transfer platform; 2 is the gantry; 3 is the rotating beam; 4 is the swing drive; 5 is the rotator; 6 is the commutator; 7 is the jig plate; 8 is the horizontal electric cylinder; 9 is the lifting servo motor; 10 is the rotary device; 11 is the camera; 12 is the bright field light source; 13 is the dark field light source; and 14 is the light source that transitions between light and dark. Detailed Implementation

[0033] The following are specific embodiments of the present invention, which are described in conjunction with the accompanying drawings to further illustrate the technical solutions of the present invention. However, the present invention is not limited to these embodiments.

[0034] Example 1

[0035] like Figure 1 and Figure 2 As shown, the multi-faceted appearance inspection equipment includes a machine base, on which a transfer platform 1 is installed. The transfer platform 1 reciprocates and drives a gantry frame 2. A rotating beam 3 is hinged on the gantry frame 2. A swing drive 4 and a rotating fixture table are installed on the rotating beam 3. A shooting inspection mechanism is installed above the transfer platform 1. The shooting inspection mechanism includes a transverse module, on which a translation drive lifting module is installed. On the lifting module, a lifting drive lifting plate is installed. A camera 11 is rotated and driven by a rotator 10 on the lifting plate. A bright field light source 12, a dark field light source 13, and a light-dark transition light source 14 are also installed on the lifting plate. The bright field light source 12 is fixed directly below the camera 11. The dark field light source 13 is tilted to one side of the camera 11. The light-dark transition light source 14 is tilted to the other side of the camera 11. The tilt angle between the dark field light source 13 and the camera 11 is smaller than the tilt angle between the light-dark transition light source 14 and the camera 11.

[0036] Bright field light source 12 is a uniform line light source, while dark field light source 13 and light-dark transition light source 14 are ultra-bright line light sources. Bright field light source 12 illuminates the main surface of the object perpendicularly, dark field light source 13 illuminates the side surface of the object obliquely, and light-dark transition light source 14 illuminates the junction edge between the main surface and the side surface of the object obliquely.

[0037] like Figure 3As shown, the gantry frame 2 includes two opposing vertical plates, with a rotating beam 3 hinged between them via bearings. A swing actuator 4 is fixed to one end of the rotating beam 3, and the swing shaft of the swing actuator 4 is inserted into the hinge point of one of the vertical plates. When the swing actuator 4 is started, the swing shaft rotates, and the reaction force drives the rotating beam 3 to rotate around the hinge point to a suitable tilt angle.

[0038] The transfer platform 1 includes a feeding servo motor, which is rotatably connected to a lead screw. A guide rail is parallel to the side of the lead screw, and a nut block is threaded onto the outer circumference of the lead screw. The nut block is fixed to a guide block, which engages with the guide rail to form a sliding connection. The guide block is fixed to an upright plate via a connecting plate. Starting the feeding servo motor drives the lead screw to rotate forward and backward, causing the guide block and upright plate to reciprocate along the guide rail, thus realizing the feeding and discharging conveying operations.

[0039] like Figure 3 As shown, the rotary fixture table includes a rotator 5 fixed on a rotating beam 3, a commutator 6 connected to the rotator 5, and a fixture disk 7 connected to the commutator 6. The fixture disk 7 has several suction holes evenly distributed on it. Starting the rotator 5 causes the fixture disk 7 to rotate, thus adjusting the circumferential angle of the fixture disk 7.

[0040] A contour block is protruding from the center of the top surface of the jig plate 7, and several suction cups are arranged around the perimeter of the jig plate 7. A feature groove is provided in the center of the product. The product is placed on the jig plate 7, and the contour block fits into the feature groove to position the product. The product is firmly adhered by the suction cups.

[0041] like Figure 2 As shown, a truss is installed above the transfer platform 1. The lateral movement module includes a horizontal electric cylinder 8 fixed on the truss, and the translation block of the horizontal electric cylinder 8 is fixed to a translation plate. Activating the horizontal electric cylinder 8 drives the translation plate to move back and forth, thereby synchronously driving the camera 11 and several light sources to move back and forth.

[0042] The lifting module includes a lifting servo motor 9 fixedly mounted on a translation plate. The lifting servo motor 9 is connected to a lead screw via a synchronous belt drive. A vertical rail is arranged parallel to the side of the lead screw. A nut sleeve is threaded onto the outer circumference of the lead screw, and the nut sleeve is fixedly connected to the lifting plate. A vertical block is correspondingly arranged on the lifting plate, and the vertical block engages with the vertical rail to form a guide sliding connection. Starting the lifting servo motor 9 drives the camera 11 and the light source to move up and down, and starting the rotator 10 drives the camera 11 to rotate.

[0043] Example 2

[0044] Based on Embodiment 1, the difference in this embodiment is:

[0045] A detection calculation method for a multi-faceted appearance inspection device includes the following steps:

[0046] 1) Place the product on the rotating fixture table to fix it. Start the transfer platform 1 to move the gantry 2 from the front end to below the shooting and testing mechanism. Start the horizontal movement module to drive the camera 11, bright field light source 12, dark field light source 13 and light-dark transition light source 14 to move back and forth. Start the lifting module to drive the camera 11, bright field light source 12, dark field light source 13 and light-dark transition light source 14 to move up and down, so that the camera 11 is located at the set height above the product.

[0047] 2) Start the swing drive 4 to rotate the swing shaft, which drives the rotating beam 3 to rotate around the hinge point to 90°. Start the rotator 5 to drive the jig disk 7 to rotate. The product is in a vertical state with one side facing up. Turn on the camera 11 to take pictures of several set points on one side of the product in sequence.

[0048] 3) For example Figure 4 As shown, the bright field light source 12, dark field light source 13, and light-dark transition light source 14 are controlled to switch on and off at high speed in sequence. When the bright field light source 12 / dark field light source 13 / light-dark transition light source 14 illuminates the product, the camera 11 captures bright field images, dark field images, and light-dark transition images at each set point. All side images are collected for target detection classification and localization. Then, fine-grained classification is used to classify minor defects and serious defects. At the same time, semantic segmentation is used to calculate the area. Then, the area, quantity, and distance of defects are obtained through PS and HS measurement algorithms. When the area, quantity, and distance of defects are all less than or equal to the standard value, the product is judged to be qualified. When any of the area, quantity, and distance is greater than the standard value, the product is judged to be unqualified.

[0049] 4) Start the rotator 5 to drive the jig disk 7 to rotate, and sequentially switch the other side of the product to face up. Then turn on the camera 11 again to take pictures of several set points on the other side of the product in sequence. Repeat step 3) to calculate the area, quantity and distance of the defects. Repeat the operation until all sides of the product have been photographed and judged.

[0050] 5) Start the swing drive 4 to rotate the swing shaft, which drives the rotating beam 3 to rotate around the hinge point to 0°. The product is in a horizontal state with the front facing up. Turn on the camera 11 to take pictures of the front of the product. The camera 11 moves along the set route of the product in the X and Y directions to take pictures.

[0051] 6) For example Figure 4As shown, the bright field light source 12, dark field light source 13, and light-dark transition light source 14 are controlled to switch on and off at high speed in sequence. When the bright field light source 12 / dark field light source 13 / light-dark transition light source 14 illuminates the product, the camera 11 captures bright field images, dark field images, and light-dark transition images for each set route. All front-facing images are collected for multi-source high-resolution image detection of minimal targets. Based on defects, the defect area is randomly cropped for training. The area is calculated through semantic segmentation, and then the product is judged to be qualified or unqualified through a simple threshold judgment.

[0052] 7) If the results of steps 5) and 6) are both qualified, the product is a good product; if at least one of them is unqualified, the product is a defective product.

[0053] In step 3), when defects overlap, the area, quantity, and distance are determined using the bounding rectangle of the overlapping area. When the PS and HS measurement algorithms are shielded by the automatic optical inspection, a single-point bitmap registration algorithm is used to match the actual captured image with the standard image, and then the transformation matrix is ​​estimated to obtain the calibration image and the automatic optical inspection area.

[0054] In step 2) or 5), during the shooting process of camera 11, the rotator 10 is activated to drive camera 11 to rotate and adjust the lens angle.

[0055] The product is driven to move along the Y-axis by the transfer platform 1, the camera 11 is driven to move along the X-axis by the transverse module, and the camera 11 is driven to move along the Z-axis by the lifting module; the product is driven to circumferentially rotate by the swing driver 4, which is the axis for flipping the arc angle of the product's side; the product is driven to rotate axially by the rotator 5, which is the axis for the product's own 360° rotation; the above five motion adjustment methods form a five-axis linkage structure; by adding a rotator 10 to drive the camera 11 to rotate, a six-axis linkage structure is formed. The rotation of the camera 11 adjusts the coaxial light to provide supplementary lighting when there are blind spots in the four-zone ring light source, which can capture blind spots at different angles and improve the product defect detection capability.

[0056] By utilizing five-axis and six-axis linkage interpolation, light field conversion at different angles is achieved. The interpolation motion is controlled by software, which performs digital calculations based on given information, continuously calculating the feed commands for each axis involved in the motion. It then drives the corresponding actuators to produce coordinated motion, enabling the controlled mechanical components to move along the desired route and speed. Therefore, trajectory interpolation and axis position servo control are two main components of the motion control system; simply put, it involves densifying data points between the start and end points of a line segment. Interpolation calculation involves the device calculating the required travel path contour based on the input basic data. During calculation, feed pulses are sent to each axis based on the calculation results. For each pulse, each axis moves a distance equivalent to one pulse in the corresponding coordinate direction, thus achieving the desired travel path between the axes.

[0057] Compared with existing technologies, this multi-faceted appearance inspection equipment and inspection calculation method have the following advantages:

[0058] 1. Based on three-dimensional translation and adjustment, two to three types of rotation and adjustment have been added to form five-axis linkage or six-axis linkage, which increases the shooting angle of the product and allows the product to be rotated and adjusted in multiple directions at the same inspection station. This enables multi-faceted inspection and shooting at a fixed point, reduces the inspection and transportation path, improves inspection efficiency, and saves equipment space and cost. At the same time, the use of single-point multi-faceted inspection avoids the accumulation of material transfer errors and improves the accuracy of inspection.

[0059] 2. In the six-axis linkage, the lens rotation adjustment itself has been added to prevent blind spots in shooting and improve the integrity and realism of the shooting.

[0060] 3. Five-axis linkage and six-axis linkage interpolate with each other to realize the conversion of light fields at different angles, enabling the defect detection equipment that acquires images from the side of the product and the line scan color image detection equipment to be connected in parallel, forming an integrated defect detection equipment that can detect the entire side and front of the product at one time.

[0061] 4. Detection is performed using optical principles, weighted imaging difference data analysis principles, color extraction methods, similarity principles, and binarization principles. Defects are identified through image comparison, and then displayed / marked on a monitor or automatically. A multi-angle light source time-division stroboscopic imaging method is used to capture images from different angles. These images are then combined through effective calculations to avoid errors caused by unilateral lighting shadows and blind spots, thereby obtaining accurate and realistic data on appearance defects and improving detection accuracy.

[0062] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

[0063] Although this paper frequently uses terms such as transfer platform 1; gantry frame 2; rotating beam 3; swing drive 4; rotator 5; commutator 6; jig plate 7; horizontal electric cylinder 8; lifting servo motor 9; rotary device 10; camera 11; bright field light source 12; dark field light source 13; light-dark transition light source 14, the possibility of using other terms is not excluded. The use of these terms is merely for the convenience of describing and explaining the essence of the invention; interpreting them as any additional limitation would contradict the spirit of the invention.

[0064] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

Claims

1. A detection calculation method for a multi-faceted appearance inspection device, the multi-faceted appearance inspection device comprising a machine base, a transfer platform (1) provided on the machine base, the transfer platform (1) reciprocatingly driving a gantry frame (2), a rotating beam (3) hinged on the gantry frame (2), a swing drive (4) and a rotating fixture table mounted on the rotating beam (3), a shooting inspection mechanism erected above the transfer platform (1), the shooting inspection mechanism comprising a transverse module, a translation drive lifting module on the transverse module, a lifting drive lifting plate on the lifting module, and a rotary actuator (4) on the lifting plate. 10) Rotate the driving camera (11). The lifting plate is also equipped with a bright field light source (12), a dark field light source (13) and a light-dark transition light source (14). The bright field light source (12) is fixed directly below the camera (11). The dark field light source (13) is inclined on one side of the camera (11). The light-dark transition light source (14) is inclined on the other side of the camera (11). The angle of inclination between the dark field light source (13) and the camera (11) is smaller than the angle of inclination between the light-dark transition light source (14) and the camera (11). The rotating fixture table includes a rotator (5) fixed on the rotating beam (3), the rotator (5) is connected to a commutator (6), the commutator (6) is connected to a fixture plate (7), and a plurality of suction holes are evenly distributed on the fixture plate (7); a contour block is protruding from the center of the top surface of the fixture plate (7), and a plurality of suction cups are arranged around the periphery of the fixture plate (7); The detection calculation method of the multi-faceted appearance inspection equipment is characterized in that... Includes the following steps: 1) Place the product on the rotating fixture table to fix it. Start the transfer platform (1) to move the gantry (2) from the front end to the bottom of the shooting and testing mechanism. Start the horizontal movement module to drive the camera (11), bright field light source (12), dark field light source (13) and light and dark transition light source (14) to move back and forth. Start the lifting module to drive the camera (11), bright field light source (12), dark field light source (13) and light and dark transition light source (14) to move up and down, so that the camera (11) is located at the set height above the product. 2) Start the swing drive (4) to rotate the swing shaft, drive the rotating beam (3) to rotate around the hinge point to 90°, start the rotator (5) to drive the jig disk (7) to rotate, the product is in a vertical state with one side facing up, turn on the camera (11) to take pictures of several set points on one side of the product in sequence. 3) Control the bright field light source (12), dark field light source (13), and light-dark transition light source (14) to switch on and off at high speed in sequence. When the bright field light source (12) / dark field light source (13) / light-dark transition light source (14) illuminate the product, the camera (11) takes bright field images, dark field images, and light-dark transition images for each set point. Collect all side images for target detection classification and positioning. Then, the product is classified into minor defects and serious defects through fine-grained classification. At the same time, the area is calculated through semantic segmentation. Then, the area, quantity, and distance of the defects are obtained through PS and HS measurement algorithms. When the area, quantity, and distance of the defects are all less than or equal to the standard value, the product is judged to be qualified. When any of the area, quantity, and distance is greater than the standard value, the product is judged to be unqualified. In step 3), when defects overlap, the area, quantity, and distance are determined using the bounding rectangle of the overlapping area; when the PS and HS measurement algorithms are shielded by the automatic optical inspection, a single-point bitmap registration algorithm is used to match the actual captured image with the standard image, and then the transformation matrix is ​​estimated to obtain the calibration image and the automatic optical inspection area. 4) Start the rotator (5) to drive the jig disk (7) to rotate, and sequentially switch the other side of the product to face up. Then turn on the camera (11) again to take pictures of several set points on the other side of the product. Repeat step 3) to calculate the area, quantity and distance of the defects. Repeat the operation until all sides of the product are photographed and judged. 5) Start the swing drive (4) to rotate the swing shaft, drive the rotating beam (3) to rotate around the hinge point to 0°, the product is in a horizontal state with the front facing up, turn on the camera (11) to take pictures of the front of the product, the camera (11) moves along the set route of the product in the X and Y directions to take pictures. 6) Control the bright field light source (12), dark field light source (13), and light-dark transition light source (14) to switch on and off at high speed in sequence. When the bright field light source (12) / dark field light source (13) / light-dark transition light source (14) illuminates the product, the camera (11) takes bright field image, dark field image and light-dark transition image for each set route, collects all front-facing images to perform multi-source high-resolution image miniature target detection, and trains the defect area by randomly cropping the defect based on the defect. The area is calculated by semantic segmentation, and then the product is judged to be qualified or unqualified by simple threshold judgment. 7) If the results of steps 5) and 6) are both qualified, the product is a good product; if at least one of them is unqualified, the product is a defective product.

2. The detection calculation method of the multi-faceted appearance inspection equipment as described in claim 1, characterized in that, The gantry (2) includes two opposing vertical plates, and the rotating beam (3) is hinged between the two vertical plates by a bearing. The swing drive (4) is fixed to one end of the rotating beam (3), and the swing shaft of the swing drive (4) is inserted into the hinge point of one of the vertical plates.

3. The detection calculation method of the multi-faceted appearance inspection equipment as described in claim 2, characterized in that, The transfer platform (1) includes a feeding servo motor, which is rotatably connected to a lead screw. A guide rail is arranged parallel to the side of the lead screw. A nut block is threaded onto the outer circumference of the lead screw. The nut block is fixedly connected to a guide block. The guide block is engaged with the guide rail to form a sliding connection. The guide block is fixedly connected to the vertical plate through a connecting plate.

4. The detection calculation method of the multi-faceted appearance inspection equipment as described in claim 1, characterized in that, A truss is provided above the transfer platform (1), and the transverse module includes a horizontal electric cylinder (8) fixed on the truss, and the translation block of the horizontal electric cylinder (8) is fixed to the translation plate.

5. The detection calculation method of the multi-faceted appearance inspection equipment as described in claim 4, characterized in that, The lifting module includes a lifting servo motor (9) fixedly mounted on the translation plate. The lifting servo motor (9) is connected to a lead screw via a synchronous belt drive. A vertical rail is arranged parallel to the side of the lead screw. A nut sleeve is threaded onto the outer circumference of the lead screw. The nut sleeve is fixedly connected to the lifting plate. A vertical block is correspondingly arranged on the lifting plate. The vertical block is engaged with the vertical rail to form a guide sliding connection.

6. The detection calculation method of the multi-faceted appearance inspection equipment as described in claim 1, characterized in that, In step 2) or 5), during the shooting process of the camera (11), the rotator (10) is activated to drive the camera (11) to rotate and adjust the lens angle.

Images