CN115356259A - A square capacitor defect detection system and method based on deep learning - Google Patents

Abstract

The present invention provides a square capacitor defect detection system and method based on deep learning. The system includes a first transmission mechanism, a material dial mechanism, a flip mechanism, a push mechanism, a second transmission mechanism, a camera device, an AI processing unit, and a PLC controller. , a rejecting device and a plurality of sensors; the first conveying mechanism, the feeding mechanism, the turning mechanism, the pushing mechanism, the second conveying mechanism, the photographing device, the AI processing unit, the rejecting device and a plurality of sensors are respectively connected to the PLC controller, The PLC controller receives signals from various sensors, judges the position of the square capacitor according to the sensor signals, and triggers corresponding mechanisms or devices to perform corresponding actions. In the present invention, by setting shooting points on each station of the transmission mechanism, taking pictures of each surface that needs to be detected by the square capacitor, and using the AI algorithm to process the image, identify and output the detection result, it replaces the existing manual defect detection mechanism, and improves the overall performance. detection efficiency and detection accuracy.

Description

A square capacitor defect detection system and method based on deep learning

technical field

The present invention relates to the technical field of capacitance detection, in particular to a square capacitance defect detection system and method based on deep learning.

Background technique

With the increasing depletion of energy sources, the development of new energy sources is imminent. In this environment, China is vigorously developing new energy vehicles, which also increases China's demand for capacitors. However, at present, square capacitor production factories in various places mainly use manual inspection for newly produced square capacitors. They conduct defect detection on square capacitor products, and then remove the defective products to ensure that the quality of the final product meets the standard. However, such a detection method not only increases the cost, but also is inefficient, and it is easy to artificially miss and falsely detect.

Contents of the invention

The technical problem to be solved by the present invention is to provide a square capacitor defect detection system and method based on deep learning to solve the problems of low detection efficiency and insufficient accuracy of the existing square capacitor.

In the first aspect, the present invention provides a square capacitor defect detection system based on deep learning. The system includes a first transmission mechanism, a feeding mechanism, a turning mechanism, a pushing mechanism, a second transmission mechanism, a camera, and an AI processing unit. , PLC controller, rejecting device and multiple sensors;

The first conveying mechanism, material dialing mechanism, turning mechanism, pushing mechanism, second conveying mechanism, photographing device, AI processing unit, rejecting device and a plurality of sensors are respectively connected with the PLC controller, and the PLC controller receives each The signal of the sensor judges the position of the square capacitor according to the signal of the sensor, and triggers the corresponding mechanism or device to perform the corresponding action;

The material shifting mechanism is installed on the feeding end of the first conveying mechanism, and is used for evenly moving all the square capacitors entering the conveying mechanism, so that the distance between two adjacent square capacitors is equal;

The inverting mechanism is installed on the discharge end of the first conveying mechanism, and is used for inverting the square capacitor from the first conveying mechanism, so that the adhesive surface is upward;

The pushing mechanism is installed on one side of the turning mechanism, and is used for pushing the turned over capacitance so that it enters the feeding end of the second conveying mechanism;

The photographing device includes a first photographing assembly, a second photographing assembly, a third photographing assembly, a fourth photographing assembly and a fifth photographing assembly, and the first photographing assembly is located on one side of the first transmission mechanism for photographing a square capacitor The bottom surface of the second camera assembly and the third camera assembly are respectively located on both sides of the second transmission mechanism for photographing both sides of the square capacitor, and the fourth camera assembly and the fifth camera assembly are respectively located on the second transmission mechanism above, used to photograph the rubber surface, wherein the fifth photographing assembly includes a UV light source and an industrial camera, and the first photographing assembly, the second photographing assembly, the third photographing assembly and the fourth photographing assembly all include an industrial camera and a light source;

The photos captured by the photographing device are sent to the AI processing unit for recognition processing, and then the processing results are returned to the PLC controller, and the PLC controller controls the action of the rejecting device according to the returned results.

Further, the second transmission mechanism and the first transmission mechanism are belt transmission devices spliced end to end and 90 degrees to each other.

Further, the flipping mechanism includes four flipping blades, the angle between adjacent flipping blades is 90°, each flipping is 90° in the same direction.

Further, the light sources in the first camera assembly, the second camera assembly, the third camera assembly and the fourth camera assembly are all ring light sources, and are located between their corresponding industrial cameras and square capacitors, and are controlled by the PLC controller Controls the ring light to strobe.

Further, when the rejecting device receives the signal from the PLC controller, the cylinder in the rejecting device pushes the defective square capacitors to a predetermined station to separate them from the stack of normal square capacitors.

Further, the plurality of sensors include a sensor set at the material inlet of the first conveying mechanism, a sensor set at the corresponding station of the first shooting component, a sensor set at the corresponding station of the second shooting component, a sensor set at the corresponding station of the third shooting component, and a sensor set at the corresponding station of the third shooting component. The sensor provided at the station, the sensor provided at the corresponding station of the fourth photographing component, the sensor provided at the corresponding station of the fifth photographing component, the sensor provided at the turning mechanism, the sensor provided at the pushing mechanism and the sensor provided at the rejecting device sensor.

Further, the system also includes a display connected to the PLC controller.

Further, the AI processing unit includes pre-trained convolutional neural networks to obtain various types of defect recognition models.

In a second aspect, the present invention provides a square capacitor defect detection method based on deep learning, the method is based on the above-mentioned system, and the method includes the following steps:

Step S1, put the square capacitor into the material inlet of the first conveying mechanism, and the material feeding mechanism separates the square capacitors at the material inlet by a certain distance and then transports them through the first conveying mechanism;

Step S2, when the sensor senses that the square capacitor arrives at the first station, take a picture of the square capacitor through the first photographing component, and obtain two photos;

Step S3. When passing through the discharge port of the first transmission mechanism, the square capacitor is turned over by the turning mechanism so that the rubber side faces up, and then the turned square capacitor is pushed to the second transmission mechanism by the cylinder of the pushing mechanism ;

Step S4, when the sensor senses that the square capacitor arrives at the second station, take a picture of the square capacitor through the second photographing component, and obtain two photos;

Step S5, when the sensor senses that the square capacitor reaches the third station, take a picture of the square capacitor through the third photographing component, and obtain two photos;

Step S6, when the sensor senses that the square capacitor reaches the fourth station, take a photo of the square capacitor through the fourth photographing component, and obtain two photos;

Step S7, when the sensor senses that the square capacitor reaches the fifth station, take a photo of the square capacitor through the fifth photographing component, and obtain two photos;

Step S8, the AI processing unit performs defect recognition on the acquired photos of each station, returns the recognition result to the PLC controller, and controls the rejecting device through the PLC controller to remove the square capacitors that need to be removed.

Further, the detection process of the AI processing unit is as follows:

Obtain two images of each station, extract one image from each station for binarization processing, contour search, and particle analysis, and obtain the defect categories of square capacitors as oil on the shell, pores on the rubber surface, oil on the rubber surface and The recognition result on the lead wire bending, and another image is obtained from each station and input into the deep learning model, and the defect categories of the output square capacitor are shell dirt, shell wear, shell oil stains, shell micro-greasy stains, shell damaged edge Recognition results on oil stains, rubber surface air bubbles and rubber surface burrs.

One or more technical solutions provided by the present invention have at least the following technical effects or advantages:

1. Build a set of automatic detection device to accurately obtain the photos of each side of the square capacitor that needs to be inspected during the transmission of the square capacitor, and combine the corresponding software algorithm to detect the photos to achieve a fully automated process and uninterrupted detection. , improve detection efficiency, and avoid misjudgment caused by manual fatigue detection;

2. Quickly identify certain types of defects through processes such as global binarization, adjusting color thresholds, and brightness parameters. At the same time, for some defects that are applicable to deep learning methods, model the characteristics of the defects to be detected, and then conduct global inspection during detection. Retrieve and detect whether the defect exists, so as to judge the defect category of square capacitors, and eliminate products with large defects among them, which can identify and distinguish defective products from good products more quickly and accurately, reduce enterprise costs, and improve product quality. Accelerate plant efficiency.

Description of drawings

The present invention will be further described below in conjunction with the embodiments with reference to the accompanying drawings.

1 is a schematic diagram of the overall structure of a square capacitor defect detection system based on deep learning in Embodiment 1 of the present invention;

Fig. 2 is one of the internal structural diagrams of a square capacitance defect detection system based on deep learning in Embodiment 1 of the present invention;

3 is the second schematic diagram of the internal structure of a square capacitor defect detection system based on deep learning in Embodiment 1 of the present invention;

Fig. 4 is a schematic structural view of the feeding part in Embodiment 1 of the present invention;

FIG. 5 is an execution flowchart of a deep learning-based square capacitor defect detection method in Embodiment 2 of the present invention.

Explanation of reference numbers:

100-square capacitor defect detection system, 1-first transmission mechanism, 2-feeding mechanism, 3-turning mechanism, 31-turning blade, 4-push mechanism, 5-second transmission mechanism, 6-camera device, 61- The first camera component, 62-the second camera component, 63-the third camera component, 64-the fourth camera component, 65-the fifth camera component, 7-PLC controller, 8-rejecting device, 200-square capacitor.

Detailed ways

The embodiment of the present application provides a square capacitor defect detection system and method based on deep learning, which is used to solve the problems of time-consuming and labor-intensive manual detection of square capacitors, low efficiency, high missed detection rate, and high false detection rate. . The general idea of the technical solution in the embodiment of this application is as follows: By constructing an automatic defect detection system, the position of the square capacitor is reversed during the transportation process, and the parameters of the image acquisition device are set in advance, so as to obtain the square capacitor in real time. The image under the station obtains the side information of each square capacitor that needs to be detected, and transmits it to the algorithm processing module to perform the recognition operation in real time, and feeds back the recognition result to the PLC controller in real time, and removes the defective square capacitor through the rejecting device. Realize all-round shooting and automatic recognition of square capacitors, reduce the error rate of manual operation, and improve the overall defect detection efficiency.

In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

Embodiment one

This embodiment provides a square capacitor defect detection system based on deep learning. As shown in FIGS. Two transmission mechanism 5, photographing device 6, AI processing unit (not shown), PLC controller 7, rejecting device 8 and a plurality of sensors (not shown);

The first transmission mechanism 1, the feeding mechanism 2, the turning mechanism 3, the pushing mechanism 4, the second transmission mechanism 5, the photographing device 6, the AI processing unit, the rejecting device 8 and a plurality of sensors are respectively connected with the PLC controller, so The PLC controller judges the position of the square capacitor according to the sensor signal by receiving the signal of each sensor, and triggers the corresponding mechanism or device to perform the corresponding action;

Said material shifting mechanism 2 is installed on the feeding end of the first conveying mechanism 1, and is used to carry out the uniform operation of all the square capacitors entering the conveying mechanism, so that the distance between two adjacent square capacitors 200 is equal; Mechanism 2 is composed of a plurality of staggered L-shaped pressing pieces and cylinders, and performs clamping and peeling operations on a square capacitor by controlling every two L-shaped pressing pieces.

The turning mechanism 3 is installed on the discharge end of the first conveying mechanism 1, and is used for turning over the square capacitor on the first conveying mechanism, so that its adhesive surface is upward;

The pushing mechanism 4 is installed on one side of the flipping mechanism, and is used to push the flipped capacitor so that it enters the feeding end of the second conveying mechanism; the pushing mechanism 4 is composed of a cylinder, and is sensed to a corresponding position by a sensor. At this time, the cylinder of the pushing mechanism 4 of the PLC controller pushes the square capacitor in the turning mechanism to the second transmission mechanism.

The photographing device 6 includes a first photographing assembly 61, a second photographing assembly 62, a third photographing assembly 63, a fourth photographing assembly 64 and a fifth photographing assembly 65, and the first photographing assembly 61 is positioned at the first transport mechanism 2 One side is used to photograph the bottom surface of the square capacitor (that is, the opposite side of the capacitor glue surface, used for identifying defects and OCR character recognition), and the second photographing assembly 62 and the third photographing assembly 63 are respectively located on both sides of the second transmission mechanism , used to photograph both sides of the square capacitor, the fourth photographing assembly and the fifth photographing assembly are respectively located above the second conveying mechanism for photographing the rubber surface, wherein the fifth photographing assembly includes a UV light source and an industrial camera , the first photographing assembly, the second photographing assembly, the third photographing assembly and the fourth photographing assembly each include an industrial camera and a light source;

The photos captured by the photographing device are sent to the AI processing unit for recognition processing, and then the processing results are returned to the PLC controller, and the PLC controller controls the action of the rejecting device according to the returned results.

In this embodiment, the second transmission mechanism and the first transmission mechanism are belt transmission devices spliced end-to-end and 90 degrees to each other.

In this embodiment, the turning mechanism 3 includes four turning blades 31, the angle between adjacent turning blades is 90°, and each turning is turned 90° in the same direction, and the turning mechanism 3 passes through the corresponding motor control.

In this embodiment, the light sources in the first photographing assembly 61, the second photographing assembly 62, the third photographing assembly 63, and the fourth photographing assembly 64 are all ring light sources, and are located between their corresponding industrial cameras and square capacitors. time, and the strobe of the ring light source is controlled by the PLC controller, which can prolong the life of the light source in each camera component and reduce the frequency of replacement.

In this embodiment, when the rejecting device 8 receives the signal from the PLC controller, the cylinder in the rejecting device pushes the defective square capacitor to a predetermined station to separate it from the stack of normal square capacitors.

Preferably, the present invention also sets the peripheral barrier 9 and the positive pressure device 10 above the peripheral barrier 9. The external light can be blocked by the peripheral barrier to reduce the impact on the equipment. The positive pressure device prevents the external adhesive gas from adhering to the machine. Taiwan.

In this embodiment, the plurality of sensors include a sensor set at the material inlet of the first conveying mechanism, a sensor set at the corresponding station of the first photographing component, a sensor set at the corresponding station of the second photographing component, a third The sensor set at the station corresponding to the photographing component, the sensor set at the corresponding station of the fourth photographing component, the sensor set at the corresponding station of the fifth photographing component, the sensor set at the turning mechanism, the sensor set at the pushing mechanism, and the rejecting device sensor set up.

In this embodiment, the system further includes a display connected to the PLC controller, and the display is a resistive touch screen.

In this embodiment, the AI processing unit includes various types of defect recognition models obtained through pre-training using a convolutional neural network.

Embodiment two

In this embodiment, a square capacitor defect detection method based on deep learning is provided, as shown in Figure 5, the method is based on the above square capacitor defect detection system, and the method includes the following steps:

Step S1, put the square capacitor into the material inlet of the first transmission mechanism, and the material feeding mechanism separates the square capacitors at the material inlet according to a certain distance and then transports them through the first transmission mechanism; Each product is evenly separated, because when feeding, the products are connected together;

Step S2, when the sensor senses that the square capacitor arrives at the first station, take a picture of the square capacitor through the first photographing component, and obtain two photos;

Step S3. When passing through the discharge port of the first transmission mechanism, the square capacitor is turned over by the turning mechanism so that the rubber side faces up, and then the turned square capacitor is pushed to the second transmission mechanism by the cylinder of the pushing mechanism ;

Step S4, when the sensor senses that the square capacitor arrives at the second station, take a picture of the square capacitor through the second photographing component, and obtain two photos;

Step S5, when the sensor senses that the square capacitor reaches the third station, take a picture of the square capacitor through the third photographing component, and obtain two photos;

Step S6, when the sensor senses that the square capacitor reaches the fourth station, take a photo of the square capacitor through the fourth photographing component, and obtain two photos;

Step S7, when the sensor senses that the square capacitor reaches the fifth station, take a photo of the square capacitor through the fifth photographing component, and obtain two photos;

Step S8, the AI processing unit performs defect recognition on the acquired photos of each station, returns the recognition result to the PLC controller, and controls the rejecting device through the PLC controller to remove the square capacitors that need to be removed.

Preferably, the detection process of the AI processing unit is as follows:

Obtain two images of each station, extract one image from each station for binarization processing, contour search, and particle analysis, and obtain the defect categories of square capacitors as oil on the shell, pores on the rubber surface, oil on the rubber surface and The recognition result on the lead wire bending, and another image is obtained from each station and input into the deep learning model, and the defect categories of the output square capacitor are shell dirt, shell wear, shell oil stains, shell micro-greasy stains, shell damaged edge Recognition results on oil stains, rubber surface air bubbles and rubber surface burrs. Since different algorithms have differences in the detection accuracy of different types of defects, the present invention uses the binarization processing of traditional algorithms, contour search, particle analysis, and deep learning algorithms to classify and group all the defects of square capacitors, so as to achieve the best results. Good detection accuracy.

Before using deep learning for detection, the defective parts are manually marked and trained to obtain a defect detection algorithm for square capacitors, and different feature models are established for different defect types. When it is necessary to detect new square capacitor images, you can use The defect detection algorithm is trained to find out whether there is a defect. Through the established defect feature model for identification, it can identify whether there is a defect above the square capacitor and what type of defect exists. If the square capacitor cannot be detected, the image will not be processed. any processing.

Compared with the prior art, the technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: the square capacitors can be moved at the same distance through push rods, conveyor belts, flipping mechanisms and other action devices, and the four directions of the capacitors Install the shooting device to get the picture of the current product, and then use the corresponding AI algorithm to quickly identify the type and location of the defect, and eliminate the defective products, so as to achieve the purpose of rapid detection and sorting, and effectively avoid the leakage caused by the long manual detection time The accuracy of false detection is higher, and the labor cost is effectively reduced. The labor only needs to be loaded and received, and the limitation is smaller. The system of the present invention can be adapted to the transmission line with a maximum speed of 55 pieces/min, and the detection efficiency has been greatly improved. improve.

Although the specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments we have described are only illustrative, rather than used to limit the scope of the present invention. Equivalent modifications and changes made by skilled personnel in accordance with the spirit of the present invention shall fall within the protection scope of the claims of the present invention.

Claims (10)

1. A square capacitor defect detection system based on deep learning, characterized in that: the system includes a first transmission mechanism, a material dial mechanism, a turning mechanism, a pushing mechanism, a second transmission mechanism, a camera, an AI processing unit, and a PLC a controller, a reject device and a plurality of sensors; The first conveying mechanism, material dialing mechanism, turning mechanism, pushing mechanism, second conveying mechanism, photographing device, AI processing unit, rejecting device and a plurality of sensors are respectively connected with the PLC controller, and the PLC controller receives each The signal of the sensor judges the position of the square capacitor according to the signal of the sensor, and triggers the corresponding mechanism or device to perform the corresponding action; The material shifting mechanism is installed on the feeding end of the first conveying mechanism, and is used for evenly moving all the square capacitors entering the conveying mechanism, so that the distance between two adjacent square capacitors is equal; The inverting mechanism is installed on the discharge end of the first conveying mechanism, and is used for inverting the square capacitor from the first conveying mechanism, so that the adhesive surface is upward; The pushing mechanism is installed on one side of the turning mechanism, and is used for pushing the turned over capacitance so that it enters the feeding end of the second conveying mechanism; The photographing device includes a first photographing assembly, a second photographing assembly, a third photographing assembly, a fourth photographing assembly and a fifth photographing assembly, and the first photographing assembly is located on one side of the first transmission mechanism for photographing a square capacitor The bottom surface of the second camera assembly and the third camera assembly are respectively located on both sides of the second transmission mechanism for photographing both sides of the square capacitor, and the fourth camera assembly and the fifth camera assembly are respectively located on the second transmission mechanism above, used to photograph the rubber surface, wherein the fifth photographing assembly includes a UV light source and an industrial camera, and the first photographing assembly, the second photographing assembly, the third photographing assembly and the fourth photographing assembly all include an industrial camera and a light source; The photos captured by the photographing device are sent to the AI processing unit for recognition processing, and then the processing results are returned to the PLC controller, and the PLC controller controls the action of the rejecting device according to the returned results. 2. A square capacitance defect detection system based on deep learning according to claim 1, characterized in that: the second transmission mechanism and the first transmission mechanism are belt transmission devices spliced end to end and 90 degrees to each other. 3. A square capacitance defect detection system based on deep learning according to claim 1, characterized in that: the flipping mechanism includes four flipping blades, the angle between adjacent flipping blades is 90°, each time The flips are all flipped 90° in the same direction. 4. A kind of square capacitor defect detection system based on deep learning according to claim 1, characterized in that: the light sources in the first camera assembly, the second camera assembly, the third camera assembly and the fourth camera assembly are all It is a ring light source, and it is located between its corresponding industrial camera and the square capacitor, and the strobe of the ring light source is controlled by the PLC controller. 5. A kind of square capacitor defect detection system based on deep learning according to claim 1, characterized in that: when the rejecting device receives the PLC controller signal, it pushes the defective square capacitor to the The intended station separates it from the normal square capacitor stack. 6. A square capacitance defect detection system based on deep learning according to claim 1, characterized in that: the plurality of sensors include the sensor provided at the material inlet of the first conveying mechanism, the corresponding station of the first photographing component The sensor set at the corresponding station of the second photographing component, the sensor set at the corresponding station of the third photographing component, the sensor set at the corresponding station of the fourth photographing component, the sensor set at the corresponding station of the fifth photographing component The sensor set at the turning mechanism, the sensor set at the pushing mechanism and the sensor set at the rejecting device. 7. A square capacitance defect detection system based on deep learning according to claim 1, characterized in that: said system also includes a display connected to a PLC controller. 8. A square capacitor defect detection system based on deep learning according to claim 1, characterized in that: said AI processing unit includes various defect recognition models obtained through pre-training using a convolutional neural network. 9. A square capacitor defect detection method based on deep learning, characterized in that the method is based on the system according to any one of claims 1-7, and the method comprises the steps of: Step S1, put the square capacitor into the material inlet of the first conveying mechanism, and the material feeding mechanism separates the square capacitors at the material inlet by a certain distance and then transports them through the first conveying mechanism; Step S2, when the sensor senses that the square capacitor arrives at the first station, take a picture of the square capacitor through the first photographing component, and obtain two photos; Step S3, when passing through the discharge port of the first transmission mechanism, the square capacitor is turned over by the turning mechanism so that the rubber side faces upward, and then the turned square capacitor is pushed to the second transmission mechanism by the cylinder of the pushing mechanism ; Step S4, when the sensor senses that the square capacitor arrives at the second station, take a picture of the square capacitor through the second photographing component, and obtain two photos; Step S5, when the sensor senses that the square capacitor reaches the third station, take a picture of the square capacitor through the third photographing component, and obtain two photos; Step S6, when the sensor senses that the square capacitor reaches the fourth station, take a photo of the square capacitor through the fourth photographing component, and obtain two photos; Step S7, when the sensor senses that the square capacitor reaches the fifth station, take a photo of the square capacitor through the fifth photographing component, and obtain two photos; Step S8, the AI processing unit performs defect recognition on the acquired photos of each station, returns the recognition result to the PLC controller, and controls the rejecting device through the PLC controller to remove the square capacitors that need to be removed. 10. A kind of square capacitance defect detection method based on deep learning according to claim 1, is characterized in that: the detection process of described AI processing unit is as follows: Obtain two images of each station, extract one image from each station for binarization processing, contour search, and particle analysis, and obtain the defect categories of square capacitors as oil on the shell, pores on the rubber surface, oil on the rubber surface and The recognition result on the lead wire bending, and another image is obtained from each station and input into the deep learning model, and the defect categories of the output square capacitor are shell dirt, shell wear, shell oil stains, shell micro-greasy stains, shell damaged edge Recognition results on oil stains, rubber surface air bubbles and rubber surface burrs.

Images