CN115496992A - Visual identification and detection method for melting point of overheating welding - Google Patents
Visual identification and detection method for melting point of overheating welding Download PDFInfo
- Publication number
- CN115496992A CN115496992A CN202110673749.2A CN202110673749A CN115496992A CN 115496992 A CN115496992 A CN 115496992A CN 202110673749 A CN202110673749 A CN 202110673749A CN 115496992 A CN115496992 A CN 115496992A
- Authority
- CN
- China
- Prior art keywords
- melting point
- deep learning
- learning model
- welding
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Library & Information Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a visual identification and detection method for a melting point of overheating welding, which comprises the following steps: s1, establishing a deep learning model; s2, capturing an image of the welding process at a fixed angle through a vision system; s3, transmitting the welding process image to the deep learning model; s4: and performing seeding characteristic detection and logic judgment on the welding process image through the deep learning model, and outputting a recognition result. The invention has the advantages that the size of the melting point is identified by developing a visual system, the pixel value of the melting point is identified, the seeding signal is directly output, the processor executes the seeding process through the visual signal to achieve the consistency of seeding, the working efficiency of workers is improved, the production efficiency is further improved, and the enterprise competitiveness is improved; the state of the fusion melting point in the overheating fusion welding process of the seed crystal does not need to be judged manually, automatic industrial production is realized, working hour waste and abnormal accidents caused by insufficient manual experience are avoided, and the safety in the operation process is higher.
Description
Technical Field
The invention belongs to the technical field of photovoltaic single crystal drawing production, and particularly relates to a visual identification and detection method for a melting point of overheating fusion.
Background
Whether need detect the appearance that has the seeding characteristic before getting into seeding process at present, in overheated butt fusion in-process, when the liquid level temperature of silicon solution reaches the butt fusion temperature, fall the seed crystal automatically, fall the liquid level department of silicon solution with the seed crystal to with seed crystal location to primary seed crystal department, the liquid level contact of seed crystal and silicon solution.
In the prior temperature stabilizing process, a fusion melting point is not formed in the early stage of seeding, the temperature is high, seeding breakage is easily caused, seeding consistency cannot be unified, an operator cannot visually observe whether seeding characteristics appear in the early stage of entering a seeding process, the size of the fusion melting point cannot be visually identified, a seeding signal cannot be directly sent out to carry out operation prompting, and therefore working efficiency is low; moreover, due to the influence of manual errors, the consistency of the seeding process cannot be ensured, and the waste of resources is possibly caused.
Disclosure of Invention
The invention aims to provide a visual identification and detection method for an overheating fusion melting point, which is suitable for visual judgment of seeding characteristics before a seeding process.
In order to solve the technical problems, the invention adopts the technical scheme that: a visual identification and detection method for a melting point of overheating welding comprises the following steps:
s1, establishing a deep learning model;
s2, capturing an image of the welding process at a fixed angle through a vision system;
s3, transmitting the welding process image to the deep learning model;
s4: and performing seeding characteristic detection and logic judgment on the welding process image through the deep learning model, and outputting a recognition result.
Further, in S1: acquiring basic source data of the quantity of fusion melting points on each seed crystal in each single crystal furnace from the beginning of the overheating fusion welding process to the end of seeding;
and processing the acquired source data, constructing the deep learning model by using the source data as a material, and training the deep learning model to realize iterative optimization.
Further, before S2, a melting point image needs to be captured by the vision system through four-point tracking;
transmitting the melting point image to the deep learning model;
and carrying out fusion melting point detection and logic judgment on the melting point image through the deep learning model, and outputting a recognition result.
Further, the vision system is arranged opposite to the seed crystal, and the vision system captures one melting point image when the seed crystal rotates by 90 degrees counterclockwise; the vision system captures the melting point image of one rotation of the seed crystal.
Further, the melting point image is sent to the deep learning model through a CCD program.
Further, the deep learning model outputs a result according to whether the clearly visible fusion melting point exists in each melting point image or not;
when the welding melting points exist and are one, entering a seeding characteristic judgment process;
and triggering a crystal change alarm prompt to process the seed crystal when the fusion melting point does not exist or the number of the melting points is more than one.
Further, in S2: capturing the welding process image from the beginning to the end of entering a seeding state after overheating is completed;
the vision system is arranged opposite to the seed crystal, captures one welding process image every second, and captures the number of small welding white dots in the area of the sector field opposite to the vision system.
Further, in S3: and the fusion process image is sent to the deep learning model through a CCD program.
Further, in S4:
the deep learning model converts the number of the small fusion white points into the number of pixel points of the small fusion white points in the area of the region in the image in the fusion process;
and the deep learning model outputs a result according to whether the number of the welding small white dots in the area of the detected welding process image meets a threshold value in the deep learning model.
Further, the threshold value is that the area of the region contains 40-80 pixel points, and when the number of the pixel points of the small welded white points meets the threshold value, a seeding process is performed;
and when the number of the pixel points of the small fused white points does not meet the threshold value, the seeding process is not carried out.
Due to the adoption of the technical scheme, the method has the following advantages:
1. the size of the melting point is identified by developing a visual system, the pixel value of the melting point is identified, seeding signals are directly output, the seeding process is executed by the processor through the visual signals to achieve seeding consistency, the working efficiency of workers is improved, the production efficiency is further improved, and the enterprise competitiveness is improved.
2. The state of the fusion melting point in the overheating fusion welding process of the seed crystal does not need to be judged manually, automatic industrial production is realized, working hour waste and abnormal accidents caused by insufficient manual experience are avoided, and the safety in the operation process is higher.
3. The machine visual angle replaces the manual visual angle for judgment, so that the consistency of seed crystals before seeding is improved, and the labor and the working hours are saved;
4. the seed crystal state can be monitored in real time, abnormal seed crystals can be identified in time, and the single crystal manufacturing efficiency is improved.
Drawings
FIG. 1 is a block diagram of a process according to an embodiment of the present invention;
FIG. 2 is a schematic view of a visual system capturing a field of view in accordance with an embodiment of the present invention.
In the figure:
1. seed crystal 2, rectangular region
Detailed Description
The invention is further illustrated by the following examples and figures:
in an embodiment of the present invention, as shown in fig. 1, a method for visually identifying and detecting a melting point of an overheated welding includes:
s1, establishing a deep learning model, and establishing the deep learning model for the quantity of fusion melting points of each seed crystal from the overheating fusion welding process to the seeding finishing process through deep learning, wherein the deep learning model specifically comprises the following steps:
s11: acquiring basic source data of the quantity of fusion melting points on each seed crystal in each single crystal furnace from the beginning of the overheating fusion welding process to the end of seeding;
s12: processing the obtained source data of the fusion melting point, constructing a deep learning model by using the source data as a material, and training the deep learning model to realize iterative optimization;
the deep learning model is established on the processor, the basic source data are acquired and captured through the visual system on the basis of the visual system on the processor, a large number of fusion melting point images on the seed crystals 1 are acquired and captured through the visual system, a fusion melting point database is constructed, the deep learning model is constructed, before the overheating fusion melting point is subjected to visual identification and detection, the deep learning model is required to be subjected to simulation training and trial operation, and iterative optimization is continuously performed on the deep learning model.
S2: melting point images were captured by a vision system with four-point tracking: specifically, the vision system is arranged over against the seed crystal, and the vision system captures a melting point image when the seed crystal rotates 90 degrees anticlockwise; the vision system captures the melting point image of one rotation of the seed crystal.
The vision system is disposed on the server, and specifically, the vision system may be, but is not limited to, an industrial camera. In the embodiment, the visual system is arranged right ahead, namely the visual system is arranged right ahead of the arrow in the middle of the figure, the melting point image right opposite to one point of the visual system is captured firstly, when the seed crystal rotates by 90 degrees, the visual system captures one melting point image, and the seed crystal rotates anticlockwise, namely the visual system captures 4 melting point images of four points in total in the process that the seed crystal 1 rotates anticlockwise for one circle.
S3: transmitting the melting point image to a deep learning model, specifically: the melting point image is sent to the deep learning model through the CCD program, and the CCD program sends one melting point image to the deep learning model every second, wherein the CCD program is used for transmitting the image in real time.
S4: carry out welding melting point detection and logic judgement to the fusing point image through the degree of deep learning model, output recognition result specifically is: the deep learning model outputs a result according to whether a clearly visible fusion melting point exists in each melting point image or not;
when the fusion melting points exist and are one, entering a seeding characteristic judgment process;
and when the fusion melting point does not exist or the number of the melting points is more than one, triggering a crystal change alarm prompt to process the seed crystal.
The crystal change alarm device is also arranged on the processor, and is used for judging whether the crystal change condition of the seed crystal exists or not through four-point tracking, wherein the crystal change is that the crystal structure is abnormal, and the phenomenon is that a plurality of melting points or melting points do not exist, wherein the four-point tracking is a prerequisite for judging the seed crystal entering, and when the seed crystal 1 enters a visual field range of a visual system in the rotating process, the directional tracking is carried out until the seed crystal rotates to be over against the visual system to start capturing; the seed crystal has four clearly visible fusion melting points in the process of rotating for a circle, the sizes and the forms of the four fusion melting points are consistent, four points are adopted to track whether the prior filtration meets the seeding condition, seeding operation is not carried out abnormally, and the seed crystal is processed by outputting alarm prompt.
When the fusion melting point exists and is one, the seeding characteristic judgment is continued, specifically:
s5, capturing a welding process image at a fixed angle through a vision system, and specifically comprising the following steps: the vision system is arranged opposite to the seed crystal, captures a welding process image every second, and captures the number of small welding white dots in the area of the sector vision opposite to the vision system; the process of capturing the welding process image by the vision system starts after the overheating welding is finished, and the process enters a seeding state to be finished.
Similarly, the set position of the vision system is unchanged, the vision system is opposite to the seed crystal 1 and is arranged at one side of the seed crystal 1, in the embodiment, the vision system is arranged right ahead, namely the vision system is arranged right ahead of the middle arrow shown in fig. 2, wherein the fixed angle is the capture of the selected area, the traceable visual field range of the vision system is in a fan shape, as shown in fig. 2, the fan-shaped visual field range limited by the arrows at two sides is the traceable visual field range of the vision system, the vision system begins tracing when the melting point is seen in the visible fan-shaped area until the number of the detected small welding white points in the area right ahead reaches the set threshold value, wherein the small welding white points are the welding melting points in a clear and stable state, and the number of the small welding white points in the area is a condition for judging whether the seeding process can be entered; in the embodiment, the area of the region is a rectangular region opposite to the vision system, the rectangular region has a size capable of accommodating 40-80 welding small white dots, wherein the seed crystal 6s completes one rotation, and the vision system continuously captures images of the welding process during the rotation of the seed crystal until the number of the detected welding small white dots in the area of the region right in front reaches a set threshold value.
S6, transmitting the image of the welding process to a deep learning model, specifically: the welding process image is sent to the deep learning model through the CCD program, the CCD program sends one welding process image to the deep learning model every second, and the CCD program is used for transmitting the images in real time.
S7: seeding characteristic detection and logic judgment are carried out on the welding process image through the deep learning model, and an identification result is output, specifically:
s71: the deep learning model identifies and converts the number of the small fused white dots into the number of pixel points of the small fused white dots in the area of the image in the fusion process;
s72: and the deep learning model outputs a result according to whether the number of the small fusion white dots in the area of the detected fusion process image meets the threshold value in the deep learning model.
The threshold is set to be 40-80 pixel points in the area, and when the number of the pixel points for welding the small white points meets the threshold, a seeding process is carried out;
and when the number of the pixels for welding the small white points does not meet the threshold value, the seeding process is not carried out.
In the embodiment, the area is set to be a rectangular frame, the number of the welded small white dots in the rectangular area 2 is set to be 40-80 pixel points, images are continuously captured in the capturing process through a visual system, the number of the pixel points of the welded small white dots in the rectangular frame is captured in each image in the welding process, a set threshold value is met, a seeding signal is directly sent out to enter a seeding process, the size of a melting point is identified through developing the visual system, the output of the seeding signal is directly carried out by identifying the pixel value of the melting point, a processor executes a seeding process through the visual signal to achieve the consistency of seeding, and the working efficiency and the production efficiency are improved.
The embodiments of the present invention have been described in detail, but the description is only for the preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. All equivalent changes and modifications made within the scope of the present invention shall fall within the scope of the present invention.
Claims (10)
1. A method for visually identifying and detecting a melting point of overheating welding is characterized by comprising the following steps:
s1, establishing a deep learning model;
s2, capturing an image of the welding process at a fixed angle through a vision system;
s3, transmitting the welding process image to the deep learning model;
s4: and performing seeding characteristic detection and logic judgment on the welding process image through the deep learning model, and outputting a recognition result.
2. The method for visually recognizing and detecting the melting point of the overheated welding according to claim 1, wherein the method comprises the following steps: in the step S1: acquiring basic source data of the quantity of fusion melting points on each seed crystal in each single crystal furnace from the beginning of the overheating fusion welding process to the end of seeding;
and processing the acquired source data, constructing the deep learning model by using the source data as a material, and training the deep learning model to realize iterative optimization.
3. The method for visually recognizing and detecting the melting point of the overheated welding according to claim 2, wherein the method comprises the following steps: before S2, a melting point image needs to be captured through four-point tracking of the vision system;
transmitting the melting point image to the deep learning model;
and carrying out fusion melting point detection and logic judgment on the melting point image through the deep learning model, and outputting a recognition result.
4. The method for visually recognizing and detecting the melting point of the overheated welding according to claim 3, wherein the method comprises the following steps: the visual system is arranged opposite to the seed crystal, and the visual system captures one melting point image when the seed crystal rotates by 90 degrees anticlockwise; the vision system captures the melting point image of one rotation of the seed crystal.
5. The method for visually recognizing and detecting the melting point of the overheated molten metal according to claim 3 or 4, wherein the method comprises the following steps: the melting point image is sent to the deep learning model through a CCD program.
6. The method for visually recognizing and detecting the melting point of the overheated welding according to claim 5, wherein the method comprises the following steps: the deep learning model outputs a result according to whether the clearly visible fusion melting point exists in each melting point image or not;
when the welding melting points exist and are one, entering a seeding characteristic judgment process;
and when the fusion melting point does not exist or the number of the melting points is more than one, triggering a crystal change alarm prompt to process the seed crystal.
7. The method for visually recognizing and detecting the melting point of the overheated molten metal according to claim 2 or 6, wherein the method comprises the following steps: in the S2: capturing the welding process image from the beginning to the end of entering a seeding state after overheating is completed;
the vision system is arranged right opposite to the seed crystal, captures one welding process image every second, and captures the number of small welding white dots in the area of the sector field of view right opposite to the vision system.
8. The method for visually recognizing and detecting the melting point of the overheated molten metal according to claim 7, wherein the method comprises the following steps: in the S3: the fusion process image is sent to the deep learning model via a CCD program.
9. The method for visually recognizing and detecting the melting point of the overheated molten metal according to claim 8, wherein the method comprises the following steps: in the step S4:
the deep learning model converts the number of the small fusion white points into the number of pixel points of the small fusion white points in the area of the region in the image in the fusion process;
and the deep learning model outputs a result according to whether the number of the welding small white dots in the area of the detected welding process image meets a threshold value in the deep learning model.
10. The method for visually recognizing and detecting the melting point of the overheated molten metal according to claim 9, wherein the method comprises the following steps: the threshold value is that the area of the region contains 40-80 pixel points, and when the number of the pixel points of the small welded white points meets the threshold value, a seeding process is carried out;
and when the number of the pixel points of the small fused white points does not meet the threshold value, the seeding process is not carried out.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110673749.2A CN115496992A (en) | 2021-06-17 | 2021-06-17 | Visual identification and detection method for melting point of overheating welding |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110673749.2A CN115496992A (en) | 2021-06-17 | 2021-06-17 | Visual identification and detection method for melting point of overheating welding |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115496992A true CN115496992A (en) | 2022-12-20 |
Family
ID=84464806
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110673749.2A Pending CN115496992A (en) | 2021-06-17 | 2021-06-17 | Visual identification and detection method for melting point of overheating welding |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115496992A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116934727A (en) * | 2023-07-28 | 2023-10-24 | 保定景欣电气有限公司 | Seed crystal welding control method and device in crystal pulling process and electronic equipment |
-
2021
- 2021-06-17 CN CN202110673749.2A patent/CN115496992A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116934727A (en) * | 2023-07-28 | 2023-10-24 | 保定景欣电气有限公司 | Seed crystal welding control method and device in crystal pulling process and electronic equipment |
| CN116934727B (en) * | 2023-07-28 | 2024-03-08 | 保定景欣电气有限公司 | Seed crystal welding control method and device in crystal pulling process and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109064440B (en) | Loudspeaker voice coil bonding wire identification method based on machine vision | |
| CN112837302B (en) | Method and device for monitoring state of die, industrial personal computer, storage medium and system | |
| CN107402220A (en) | A kind of selective laser fusing shaping powdering quality vision online test method and system | |
| CN101907453B (en) | Online measurement method and device of dimensions of massive agricultural products based on machine vision | |
| CN106392304B (en) | A kind of laser assisted weld seam Intelligent tracing system and method | |
| CN103051872A (en) | Method for detecting conveyor belt deviation based on image edge extraction | |
| KR102322546B1 (en) | Sewing stitching defect detection method and system | |
| CN205538710U (en) | Inductance quality automatic check out system based on machine vision | |
| CN111368727B (en) | Dressing detection method, storage medium, system and device for inspection personnel in power distribution room | |
| CN110555838A (en) | Image-based part fault detection method and device | |
| CN115496992A (en) | Visual identification and detection method for melting point of overheating welding | |
| CN112577969A (en) | Defect detection method and defect detection system based on machine vision | |
| CN105260716A (en) | Fault indicator state identification method and fault indicator state identification device | |
| CN106897995A (en) | A kind of parts automatic identifying method towards Automatic manual transmission process | |
| CN105096305A (en) | Method and device for analyzing state of insulators | |
| CN107014818A (en) | A kind of laser carving defective vision detecting system and method | |
| CN102663781A (en) | Sub-pixel level welding center extraction method based on visual sense | |
| CN114022810A (en) | Method, system, medium and terminal for detecting working state of climbing frame protective net in construction site | |
| CN102401800A (en) | Online detection device for detecting welding quality of high voltage ceramic capacitor based on visual sense of device | |
| CN112525163A (en) | Punch piercing detection system, method, device, control device and storage medium | |
| CN114199127B (en) | Automobile part size detection system and method based on machine vision | |
| CN117169227A (en) | Plug production method, device, equipment and storage medium | |
| CN110977292A (en) | Automatic detection method for welding seam of ocean platform module structure | |
| CN102829891B (en) | Visual detection system for judging material temperature and posture and detection method | |
| CN107356232A (en) | A kind of vision detection system image processing method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |