A coated glass film surface detection method and detection device based on a YOLO model

Abstract

The application discloses a kind of based on YOLO model's coated glass film surface detection method and detection device, belong to coated glass film surface detection technical field, including the following steps: the reflection light image training set of the measured coated glass is constructed, based on the training set, the training and algorithm optimization of YOLO model are completed;The reflection light image of the measured glass is obtained, and the glass film surface class is detected using the trained YOLO model;Final output detection result.The detection device includes: image acquisition module is used to obtain glass surface reflection light image and store data file;Data processing module is used to process data and send signal instruction.The application analyzes glass reflection light image by machine learning, realizes non-contact detection, to maintain the integrity of coated glass film surface, it is suitable for different coated material quality coated glass film surface direction detection, with the advantages of high detection accuracy, good compatibility for production application scene.

Description

Technical Field

[0001] This invention relates to the field of coated glass film surface inspection technology, and in particular to a method and device for inspecting coated glass film surface based on the YOLO model. Background Technology

[0002] In the field of high-rise building curtain walls, coated insulated glass has gained widespread application due to its superior energy-saving performance. Currently, over 90% of high-performance energy-saving glass on the market is Low-E insulated glass. This type of glass uses a frame structure that combines coated glass with high-efficiency infrared light reflection capabilities with ordinary glass, and fills the hollow layer with inert gas. In this structure, given the chemical properties of the coating layer, its susceptibility to oxidation, poor abrasion resistance, and unsuitability for long-term exposure to air, the coated surface of the coated glass must face the insulated layer. Once the coated surface of the insulated glass is exposed to the external environment, it will not only cause color deviation but also lead to deterioration and peeling of the functional coating layer due to oxidation by oxygen in the air, thus severely impairing its energy-saving performance. The cost of subsequent repairs or replacement of malfunctioning glass is high, greatly increasing the economic burden on glass manufacturers and users. Therefore, on Low-E insulated glass production lines, implementing precise online detection of the coating surface orientation during operation is a key step in improving product yield and an important measure to ensure production efficiency.

[0003] Existing coated glass surface inspection devices mainly rely on the conductivity or optical reflectivity of the coating for detection. Invention patent CN102998335A and utility model patent CN215985831U both propose coated glass surface inspection devices based on coating conductivity detection. These devices determine whether the contacted glass surface is coated by checking whether a conductive circuit can be formed between a power transmitter probe, a power receiver probe, and a conductive coating layer. However, this type of device has two drawbacks for online inspection: first, there is physical contact between the coated glass and the power transmitter and receiver probes, which may cause wear and scratches to the functional film and reduce product durability during online inspection of moving glass on the production line; second, this detection device cannot be used to inspect coated glass with a non-conductive coating layer, and it is prone to misjudgment and reduced product yield when used on coated glass with poor conductivity.

[0004] Utility model patent CN201707039U proposes a coated glass film surface detection device based on the intensity of thermal infrared light reflection signals. It utilizes the difference in thermal infrared light reflection performance between the coated and uncoated surfaces of coated glass. By converting the captured reflected light signal into an electrical signal, obtaining the corresponding temperature value, and comparing it with a set reference value, the device determines the film surface orientation. Utility model patent CN223727673U proposes a coated glass film surface detection device based on the comparison of reflected light intensity. Since there is a significant difference in light reflection performance between the coated and uncoated surfaces of coated glass, while the light reflection ability of the two surfaces of ordinary glass is minimally different, this device simultaneously collects the light reflection intensity of both surfaces of the glass under test, converts them into electrical signals, and performs logical comparison to determine the film surface orientation of the coated glass. The following shortcomings exist in the application of this type of device: In actual production, the placement position and angle of glass of different sizes and weights on the glass conveyor belt will inevitably be offset. The offset of the glass will cause the incident angle and reflection angle of the signal light on the glass surface to change significantly, which will cause the direction of the reflected light to deviate and not be completely captured by the light receiving component of the detector. The resulting change in the light signal intensity will cause errors in the comparison results, leading to a decrease in the detection accuracy.

[0005] Therefore, a method and device for detecting the coating surface of coated glass based on the YOLO model are proposed to solve or alleviate the above problems. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method and device for detecting the surface coating of coated glass based on the YOLO model.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A method for detecting the surface coating of coated glass based on the YOLO model includes the following steps: A training set of reflected light images of coated glass to be detected is constructed; the training set includes reflected light images of ordinary glass and coated glass, and the target images are optimized, region annotation is performed, and category labeling is applied. Complete the YOLO model training; input the training set into the YOLO model to carry out machine learning, establish the association between the extracted reflected light image features and the glass surface category, and obtain the trained YOLO model; Complete the detection of the glass film type of the glass to be tested; acquire the reflected light image of the coated glass to be tested, input the image of the coated glass to be tested into the trained YOLO model, and let the YOLO model analyze and determine to obtain the glass film type and confidence level; The detection results are output in the form of glass film type and confidence level. When the detection result is background or the film surface of coated glass, the system starts to acquire and detect the next frame image. When the detection result is the film-free surface of coated glass, the signal processing module issues an alarm signal and the detection process is suspended.

[0008] Preferably, the specific steps for constructing the training set of the reflected light image of the coated glass to be detected are as follows: The image acquisition module is activated to take pictures of the background, the reflective areas of ordinary glass and coated glass under illumination, and to collect images of the reflected light from the uncoated and coated surfaces of the background, ordinary glass, and coated glass. To obtain reflected light image data of glass under different production scenarios, change the size of the glass, adjust the front and rear position and tilt angle of the glass on the conveyor belt, and collect reflected light images of ordinary glass and coated glass under different ambient light levels. The acquired glass reflection images were denoised and enhanced. The LabelImg software was used to label the coordinate data of the glass area in the reflection image. The effective glass reflection area was selected for training the YOLO model to accurately identify the location and features of the glass. Image category labels were specified, and the data files were organized into a training set according to the format required by the YOLO model.

[0009] Preferably, the specific steps for completing the YOLO model training are as follows: The prepared training set is input into the YOLO model to train the model to locate the area of ​​light reflected from the glass. The model performance is optimized by adjusting the training strategy and parameter settings to obtain key feature parameters of brightness, texture, spectrum and color information. Deep visual features are extracted and associated with the obtained feature parameters and glass image category labels to obtain the trained YOLO model.

[0010] Preferably, the specific steps for detecting the glass film type of the glass to be tested are as follows: When the glass under test enters the detection area, the reflected light image of the glass under test is acquired. The trained YOLO model automatically locates the glass area, extracts features, and compares them with the features of the training set. By calculating the feature similarity between the image under test and the images in the training set, the glass film category with the highest similarity is determined as the detection result, and the confidence level is calculated.

[0011] Preferably, the specific steps for outputting the detection results are as follows: The detection results are output in the form of glass film type and confidence level in the display and monitoring module, and different subsequent operation modes are executed according to different detection results. If the detection result is background or film surface of coated glass, the signal processing module does not send any instructions, and the detection device continues to collect and detect the next frame image in a loop. If the test result shows that the coated glass has no coating, and the placement direction of the coated glass to be tested is opposite to the placement direction of the coated glass required for production, the signal processing module will immediately send a "coating surface error" signal command to the alarm module. After receiving the signal command, the alarm module will issue an alarm and the testing process will be suspended.

[0012] The present invention also provides a YOLO model-based device for detecting the surface coating of coated glass, including... The image acquisition module includes a light source, a camera, and a memory. The light source is used to emit a light beam that shines on the glass on the glass conveyor belt. The camera is in the same direction as the light source and is used to acquire the reflected light image of the glass passing through the detection area. The memory is used to store the reflected light image of the glass surface acquired by the camera and the training set file. The data processing module is a microprocessor equipped with a YOLO model. It is used for training and optimizing the YOLO model, locating the glass reflection area and extracting key features from the collected reflection light image of the glass under test, and intelligently comparing the obtained parameters with the standard features of various glass film categories obtained in the training set through machine learning training to obtain the detection result of the glass film category and calculate the confidence level. When the detection result is "no film surface of coated glass", it sends a "film error" signal command to the alarm module. The display and monitoring module is used to display the test results of the glass to be tested obtained by the data processing module. Its output format is "glass film type and confidence level". At the same time, it displays the area captured by the camera in real time in the form of dynamic images. It is used to monitor the area of ​​data collection by the testing device and view the test results at any time. It also serves as an observation window to adjust the testing device to the normal position in a timely manner when the position and angle deviate due to unexpected events such as external force collisions. The alarm module, which includes a light and sound alarm and a relay, is used to receive the "film surface error" signal command sent by the data processing module when the detection result is that the coated glass has no film surface, and convert it into an alarm action. The light and sound alarm emits a flashing red light and an alarm sound, and the relay sends a signal to stop the glass conveyor belt in the production line.

[0013] Preferably, it also includes a glass detection module, which is connected to the image acquisition module via a signal line and has the function of logically determining the arrival of glass on the production line and activating the image acquisition module.

[0014] Preferably, the glass detection module includes a light emitting component, a lens, a light receiving component, a signal discriminator, and a fixed connector. It can use the relative magnitude of the intensity of the reflected light signal received by the light receiving component and a preset light signal intensity value to logically determine whether the glass has arrived.

[0015] The present invention has the following beneficial effects: This invention uses the reflected light image of the glass surface as the detection object. It analyzes the reflected light images of the coated and uncoated surfaces of the coated glass using the YOLO model. This not only achieves non-contact detection and effectively maintains the integrity of the coated glass, but also applies to coated glass with different coating materials and shows good compatibility with the placement position and angle of the glass under test. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the YOLO-based method for detecting the surface coating of coated glass in the first aspect; Figure 2 This is a schematic diagram of the structure of the YOLO-based coated glass film surface detection device in the second aspect; Figure 3 This is a schematic diagram of the image acquisition module of the YOLO-based coated glass film surface detection device in the second aspect; Figure 4 The output effect of the display and monitoring module of the YOLO model-based coated glass film surface detection device on the background detection results in the second aspect; Figure 5 The second aspect describes the output effect of the display and monitoring module of the YOLO model-based coated glass film surface detection device on the film surface detection results of the coated glass. Figure 6 The output effect of the display and monitoring module of the coated glass film surface detection device based on the YOLO model in the second aspect on the detection results of the uncoated surface of the coated glass; Figure 7 This is a schematic diagram of the coated glass film surface detection device based on the YOLO model, which includes a glass detection module, as described in the second aspect. Figure 8 This is a schematic diagram of the glass detection module of the coated glass surface detection device based on the YOLO model in the second aspect. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0019] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0021] In the description of this invention, it should be understood that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use, or the orientation or positional relationship commonly understood by those skilled in the art. They are only used to facilitate the description of this invention and to simplify the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0022] Furthermore, the terms "first," "second," and "third" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0023] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" 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.

[0024] A method for detecting the surface coating of coated glass based on the YOLO model includes the following steps: Step 102: Construct a training set of images of the reflected light from the coated glass to be detected, such as... Figure 2As shown, a YOLO-based coated glass surface inspection device is used as a physical tool, which is fixedly installed on the outer side of the glass conveyor belt 11, with the image acquisition module 2 of the inspection device facing the plane of the glass 7. The image acquisition module 2 includes a light source 21, a camera 22, and a memory 23, and its structure is as follows. Figure 3 As shown, the image acquisition module 2 is activated, and the light beam emitted by the light source 21 illuminates the glass 7 or support plate 12 on the glass conveyor belt 11 in front of it. The camera 22, which is in the same direction as the light source, first acquires the image of the support plate 12 under the illumination as the background, and then acquires the reflected light images of the uncoated and coated surfaces of ordinary glass and coated glass on the glass conveyor belt 11 respectively. In order to cover as many situations as possible in the actual production process, the size of the glass, its front and rear position on the conveyor belt and its tilt angle are further changed. The reflected light images of the uncoated and coated surfaces of ordinary glass and coated glass are acquired under different ambient light levels to obtain the reflected light images of the glass under different production scenarios. The above data is stored in the memory 23.

[0025] The glass reflection image in memory 23 is denoised and enhanced. The LabelImg software is used to label the coordinate data of the glass area in the reflection image. The effective glass reflection area is selected for training the YOLO model to accurately identify the location and features of the glass. Image category labels are specified, and the data files are organized into a training set according to the format required by the YOLO model.

[0026] Step 104: Complete YOLO model training. Input the prepared target image training set into the YOLO model. Train the model to locate the glass reflection area. Optimize model performance by adjusting training strategies and parameter settings. Obtain key feature parameters that reflect brightness, texture, spectrum, and color information. Extract deep visual features. Establish a correlation between the obtained feature parameters and the glass image category label to obtain the trained YOLO model.

[0027] Step 106: Complete the detection of the glass film type of the glass to be tested. Place the coated glass to be tested on the glass conveyor belt 11 in a random manner with the film orientation. When the glass 7 enters the detection area, the light beam emitted by the light source 21 illuminates the surface of the glass 7. The camera 22 acquires the reflected light image of the glass 7. The obtained reflected light image is automatically denoised and enhanced by the YOLO model, and the effective glass reflected light area is located and selected. The key features of the reflected light image of the glass to be tested are extracted, and the obtained parameters are intelligently compared with the standard features of various glass film types obtained in the training set through machine learning. The glass film type with the highest similarity is determined as the detection result, and the confidence level is calculated.

[0028] Steps 108-110 Output the test results. The test results are presented in two ways at the same time. In the first way, the test results are output on the display and monitoring module 4 in the form of glass film surface category and confidence level.

[0029] like Figure 4 As shown, at a certain moment when no glass passes through the detection area, the text "Background 0.99541" displayed in the upper area of ​​the LCD screen of the display and monitoring module 4 represents the detection result and confidence level of the YOLO model for the previously acquired image, indicating that the detection result is background and the confidence level is 0.99541. The square frame in the middle area of ​​the LCD screen is the border of the effective data area selected by the YOLO model for analyzing and extracting features from the reflected light image. The black image presented on the LCD screen is the real-time display effect of the support plate 12 as the background under the camera 21. If a marker is set at the position of the support plate 12 corresponding to the edge of the detection area, the marker image can be displayed at the corresponding specific position on the LCD screen. When an accidental event such as accidental collision occurs during production and causes a serious deviation in the position and angle of the detection device, the detection device can be easily and quickly adjusted to its original position and angle based on the correct position of the marker image in the display and monitoring module 4, ensuring detection accuracy and production efficiency.

[0030] like Figure 5 As shown, at a certain moment when the coated glass to be tested, with its film facing the image acquisition module 2, passes through the detection area, the text "Film 1 0.99257" displayed on the upper part of the LCD screen of the display and monitoring module 4 represents the detection result and confidence level of the YOLO model on the previously acquired reflected light image. This indicates that the detection result is the film surface of coated glass No. 1, with a confidence level of 0.99257. Here, coated glass No. 1 corresponds to the classification label of the first type of coated glass in the training set constructed in step 102. The bright white image in the middle area of ​​the LCD screen is the real-time display effect of the reflected light image of the film surface of the coated glass. The square lines in the middle area of ​​the LCD screen are the borders of the effective data area selected by the YOLO model for analyzing and extracting features from the reflected light image. The black image outside the reflected light image area is the real-time display effect of the coated glass outside the area illuminated by the beam emitted by the light source 21 under the camera 21.

[0031] like Figure 6As shown, at a certain moment when the uncoated side of the coated glass to be tested passes through the detection area of ​​the image acquisition module 2, the text "Uncoated 1 0.98973" displayed in the upper area of ​​the LCD screen of the display and monitoring module 4 represents the detection result and confidence level of the YOLO model on the previously acquired reflected light image. It indicates that the detection result is the uncoated side of coated glass No. 1, with a confidence level of 0.98973. Here, coated glass No. 1 corresponds to the classification label of the first type of coated glass in the training set constructed in step 102. The text "Put in reverse" appearing in the lower part of the LCD screen is the coating surface error prompt, used to indicate that the coating surface of the coated glass to be tested is placed in the wrong direction. The bright white image in the middle area of ​​the LCD screen is the real-time display effect of the reflected light image area of ​​the uncoated side of the coated glass. The square line in the middle area of ​​the LCD screen is the border of the effective data area selected by the YOLO model for analyzing and extracting features from the reflected light image. The black image outside the reflected light image area is the real-time display effect of the coated glass outside the area illuminated by the beam emitted by the light source 21 under the camera 21.

[0032] The second way to output the test results is to execute different subsequent operation modes based on different test results. If the test result is background or the coated glass has a film surface, the signal processing module 3 does not send any instructions, and the detection device continues to collect and detect the next frame of reflected light image in a loop. If the test result is the coated glass without a film surface, and the placement direction of the coated glass to be tested is opposite to the placement direction of the coated glass required for production, the signal processing module 3 immediately sends a "film surface error" signal instruction to the alarm module 5. The alarm module 5 includes a light and sound alarm 51 and a relay 52. ​​After receiving the "film surface error" signal instruction, it converts it into an alarm action. The light and sound alarm 51 emits a flashing red light and an alarm sound, and the relay 52 sends a signal to stop the glass conveyor belt 11, prompting the staff to correct the direction of the coated glass and prevent the misinstallation of Low-e insulated glass. At the same time, the detection process is suspended.

[0033] After an alarm for "coating surface error" occurs on the coated glass, the alarm status of the detection device can be manually deactivated and the operation of the glass conveyor belt 11 can be restored after the coated glass is correctly placed, so that the detection device can return to normal detection status and continue to detect the subsequent glass 7.

[0034] In another embodiment, a method for detecting the surface coating of coated glass based on the YOLO model is provided. The difference between this method and the method described in the previous embodiment lies only in step 106, where a detection device including a glass detection module 6 is used, as detailed below: like Figure 7As shown, the detection device includes a glass detection module 6, which is connected to the image acquisition module 2 via a signal line and is fixedly installed on the outer side of the glass conveyor belt 11 at the front end of the image acquisition module 2. The glass detection module 6 includes a light emitting component 61, a lens 62, a light receiving component 63, a signal discriminator 64, and a fixing connector 65, and its structure is as follows. Figure 8 As shown, it has the function of determining the arrival of glass on the production line and activating the image acquisition module 2 by using the relative magnitude of the light signal intensity obtained by the light receiving component 63 and the preset light intensity.

[0035] Step 106: Complete the detection of the glass coating type of the glass to be tested. Randomly place the coated glass to be tested on the glass conveyor belt 11. When the glass 7 has not yet moved into the detection area of ​​the glass detection module 6, the light emitted by the light emitting component 61 shines on the support plate 12 after passing through the lens 62. Since the reflectivity of the support plate 12 to the light beam is much weaker than that of the glass 7, the intensity of the reflected light signal received by the light receiving component 63 is weak and less than the preset light signal intensity value. The signal discriminator 64 determines through logic that the glass 7 has not yet reached the detection area and does not send a start signal to the image acquisition module 2. The image acquisition module 2 is in a sleep state, which can reduce the invalid operation time of the image acquisition module 2 and extend its service life. When the glass 7 moves into the detection area of ​​the glass detection module 6... When the light emitting component 61 emits light through the lens 62 and irradiates the surface of the glass under test, the glass 12 has a strong ability to reflect the light beam, and the intensity of the reflected light signal is greater than the preset light signal intensity value. The signal discriminator 64 determines that the glass 7 has reached the detection area through logical relationship and sends a start signal to the image acquisition module 2. The image acquisition module 2 starts, and the camera 22 acquires the reflected light image of the glass under test. The obtained reflected light image is automatically denoised and enhanced by the YOLO model, and the effective glass reflected light area is located and selected. The key features of the reflected light image of the glass under test are extracted, and the obtained parameters are intelligently compared with the standard features of various glass film categories obtained in the training set through machine learning training. The glass film category with the highest similarity is determined as the detection result, and the confidence level is calculated.

[0036] For specific limitations regarding the YOLO-based coated glass coating surface inspection device, please refer to the section above on the YOLO-based coated glass coating surface inspection method, which will not be repeated here. Each module in the aforementioned YOLO-based coated glass coating surface inspection device can be implemented entirely or partially through hardware, software, or a combination thereof.

[0037] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting the surface coating of coated glass based on the YOLO model, characterized in that, Includes the following steps: A training set of reflected light images of coated glass to be detected is constructed; the training set includes reflected light images of ordinary glass and coated glass, and the target images are optimized, region annotation is performed, and category labeling is applied. Complete the YOLO model training; input the training set into the YOLO model to carry out machine learning, establish the association between the extracted reflected light image features and the glass surface category, and obtain the trained YOLO model; Complete the detection of the glass film type of the glass to be tested; acquire the reflected light image of the coated glass to be tested, input the image of the coated glass to be tested into the trained YOLO model, and let the YOLO model analyze and determine to obtain the glass film type and confidence level; The detection results are output in the form of glass film type and confidence level. When the detection result is background or the film surface of coated glass, the system starts to acquire and detect the next frame image. When the detection result is the film-free surface of coated glass, the signal processing module issues an alarm signal and the detection process is suspended.

2. The method for detecting the surface of coated glass based on the YOLO model according to claim 1, characterized in that, The specific steps for constructing the training set of the reflected light image of the coated glass to be detected are as follows: The image acquisition module is activated to take pictures of the background, the reflective areas of ordinary glass and coated glass under illumination, and to collect images of the reflected light from the uncoated and coated surfaces of the background, ordinary glass, and coated glass. To obtain reflected light image data of glass under different production scenarios, change the size of the glass, adjust the front and rear position and tilt angle of the glass on the conveyor belt, and collect reflected light images of ordinary glass and coated glass under different ambient light levels. The acquired glass reflection images were denoised and enhanced. The LabelImg software was used to label the coordinate data of the glass area in the reflection image. The effective glass reflection area was selected for training the YOLO model to accurately identify the location and features of the glass. Image category labels were specified, and the data files were organized according to the format required by the YOLO model to form an image training set.

3. The method for detecting the surface of coated glass based on the YOLO model according to claim 1, characterized in that, The specific steps for completing the YOLO model training are as follows: The prepared training set is input into the YOLO model to train the model to locate the area of ​​light reflected from the glass. The model performance is optimized by adjusting the training strategy and parameter settings to obtain key feature parameters of brightness, texture, spectrum and color information. Deep visual features are extracted and associated with the obtained feature parameters and glass image category labels to obtain the trained YOLO model.

4. The method for detecting the surface of coated glass based on the YOLO model according to claim 1, characterized in that, The specific steps for completing the detection of the glass film type of the glass to be tested are as follows: When the glass under test enters the detection area, the reflected light image of the glass under test is acquired. The trained YOLO model automatically locates the glass area, extracts features, and compares them with the features of the training set. By calculating the feature similarity between the image under test and the images in the training set, the glass film category with the highest similarity is determined as the detection result, and the confidence level is calculated.

5. The method for detecting the surface of coated glass based on the YOLO model according to claim 1, characterized in that, The specific steps for outputting the detection results are as follows: The detection results are output in the form of glass film type and confidence level in the display and monitoring module, and different subsequent operation modes are executed according to different detection results. If the detection result is background or film surface of coated glass, the signal processing module does not send any instructions, and the detection device continues to collect and detect the next frame image in a loop. If the test result shows that the coated glass has no coating, and the placement direction of the coated glass to be tested is opposite to the placement direction of the coated glass required for production, the signal processing module will immediately send a "coating error" signal command to the alarm module. After receiving the signal command, the alarm module will issue an alarm and the testing process will be suspended.

6. A YOLO-based coated glass film surface detection device, used to perform the YOLO-based coated glass film surface detection method as described in any one of claims 1-5, characterized in that, The image acquisition module (2) includes a light source (21), a camera (22) and a memory (23). The light source (21) is used to emit a light beam that shines on the glass (7) on the glass conveyor belt (11). The camera (22) is in the same direction as the light source and is used to acquire the reflected light image of the glass (7) passing through the detection area. The memory (23) is used to store the reflected light image of the glass surface acquired by the camera (22) and the training set file. The data processing module (3) is a microprocessor equipped with a YOLO model. It is used for training and optimizing the YOLO model, locating the glass reflection area and extracting key features from the collected reflection light image of the glass to be tested, and intelligently comparing the obtained parameters with the standard features of various glass film categories obtained in the training set through machine learning training in advance, so as to obtain the detection result of the glass film category and calculate the confidence level. When the detection result is "no film surface of coated glass", it sends a "film surface error" signal instruction to the alarm module (5). The display and monitoring module (4) is used to display the test results of the glass to be tested obtained by the data processing module (3). Its output format is "glass film type and confidence level". At the same time, it displays the area captured by the camera (22) in real time in the form of dynamic images. It is used to monitor the area location of the data collected by the detection device and view the test results at any time. It is also used as an observation window to adjust the detection device to the normal position in time when the position and angle deviate due to unexpected events such as external force collisions. The alarm module (5) includes a light and sound alarm (51) and a relay (52) for receiving the "film surface error" signal command issued by the data processing module (3) when the detection result is that the coated glass has no film surface, and converting it into an alarm action. The light and sound alarm (51) emits a flashing red light and an alarm sound, and the relay (52) synchronously sends a signal to stop the glass conveyor belt (11) in the production line (1).

7. The YOLO-based coated glass film surface detection device according to claim 6, characterized in that, It also includes a glass detection module (6), which is connected to the image acquisition module (2) via a signal line and has the function of logically determining the arrival of glass (7) on the production line and activating the image acquisition module (2).

8. The YOLO-based coated glass film surface detection device according to claim 7, characterized in that, The glass detection module (6) includes a light emitting component (61), a lens (62), a light receiving component (63), a signal discriminator (64), and a fixed connector (65). It can use the relative magnitude of the intensity of the reflected light signal received by the light receiving component (63) and the preset light signal intensity value to determine whether the glass (7) has arrived.

Images