A method for detecting the wearing of air respirators based on TinyML technology
Through TinyML technology, the method of detecting staff wearing air ventilators has been solved, and the problem of staff in petrochemical enterprises not wearing or incorrectly wearing safety equipment is realized, and the intelligent safety equipment detection and reminder functions are realized, ensuring the safety of staff.
Patent Information
- Application Number
- CN202211118375.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-15
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2042-09-15
AI Technical Summary
In petrochemical enterprises, staff are prone to poisoning or suffocation due to failure or wrong wearing of safety equipment when conducting large oil storage tank inspection and maintenance, and it is difficult for the existing technology to effectively detect and ensure wear.
Using the detection method based on TinyML technology, the image information of staff wearing air ventilators is collected and processed through the image acquisition module and the image processing unit MCU. The trained TinyML model is used to determine whether the safety equipment is correctly worn. The output is not worn or not pressed. Information for wearing.
Intelligent detection of safety equipment worn by staff has been realized, and staff are reminded to wear it in accordance with rules and regulations to avoid safety accidents caused by mistake or not wearing it.
Smart Images

Figure CN115471789B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of visual recognition, and in particular to a method for detecting wearing of an air respirator based on TinyML technology. Background Art
[0002] Oil storage tanks are containers for storing oil products. They are one of the most critical equipment for petroleum and petrochemical enterprises, and they also have high-risk characteristics. The volume of a large oil storage tank can reach 100 cubic meters. Long-term storage of oil products will lead to corrosion and deformation of the outer wall, inner wall and bottom of the tank. If they are not inspected and repaired in time, they will cause fires or even explosions, endangering national property and personal safety.
[0003] When workers inspect and repair large oil storage tanks, they need to use fire and electricity, and the internal environment of the oil storage tanks is complex, and the oil and gas vapors are volatile. If workers do not wear the required safety equipment according to the rules and regulations and do not take comprehensive protective measures, it is easy to cause poisoning or suffocation. For example, a poisoning incident occurred in a chemical company in 201X. Afterwards, it was analyzed that the employees did not wear safety equipment as required. Employees forgot to wear a certain protective equipment after working for a long time or did not think that the danger happened to them, which led to the occurrence of danger. Summary of the invention
[0004] In order to overcome the shortcomings of the above technologies, the present invention provides a method for detecting the wearing of air respirators based on TinyML technology. Its purpose is to collect images of the wearing of air respirator safety equipment before the staff performs inspection and maintenance operations. The MCU processes the collected data and informs the staff of the information that the air respirator equipment is not worn or is not worn according to regulations.
[0005] The technical solution adopted by the present invention to overcome the technical problems is:
[0006] A method for detecting wearing an air respirator based on TinyML technology, comprising:
[0007] The hardware system consists of an image acquisition module, an image processing unit MCU, and a display module;
[0008] The software system consists of a host computer, a Web terminal, and an APP terminal;
[0009] The detection method is:
[0010] a) obtaining a data set, the data set is image information of the mask and gas cylinder of the air respirator at various angles, using a labeling tool to manually label the image information of the mask and gas cylinder of the air respirator at various angles, and using the labeled images to train a TinyML model to obtain a trained TinyML model;
[0011] b) Convert the trained TinyML model into a binary file;
[0012] c) Deploy the binary converted TinyML model to the image processing unit MCU;
[0013] d) After the staff arrives at the image acquisition area, the image acquisition module acquires image information of the front, back, left and right sides of the human body, and the image processing unit MCU acquires one image as a processing source from the acquired image information of the front, back, left and right sides of the human body including the mask and gas cylinder of the air respirator, thereby obtaining four images;
[0014] e) Save the four images and run the trained TinyML model. If the image contains a mask and a gas cylinder of an air respirator, execute step f). If the image does not contain a mask and a gas cylinder of an air respirator, determine that the staff member did not carry an air respirator as required.
[0015] f) MCU intercepts the mask and gas cylinder of the staff and the air respirator in the four images, and obtains the key point position information of the staff, the mask, and the gas cylinder. The horizontal coordinate of the center point of the mask is x1, and its vertical coordinate is y1. The horizontal coordinate of the center position between the two eyes of the staff is x2, and its vertical coordinate is y2. The horizontal coordinate of the reflective strip on the left side of the staff's shoulder strap is x3, and its vertical coordinate is y3. The horizontal coordinate of the reflective strip on the right side of the staff's shoulder strap is x4, and its vertical coordinate is y4. The belt of the staff forms a continuous and uninterrupted rectangular frame at the navel;
[0016] g) If the following conditions are met simultaneously: |x1-x2|≤5cm, 0≤(y2-y1)≤15cm, the gas cylinder is detected in the rear image and not in the front image, |(x3+x4) / 2-x2|≤10cm, (y3+y4) / 2≤y2, |y3-y4|≤8cm, and the length of the rectangular frame is ≥20cm, then it is determined that the staff member is wearing the air respirator correctly as required. If these conditions are not met simultaneously, then the corresponding unmet item is output.
[0017] Preferably, the image acquisition module is a camera.
[0018] Preferably, the display module is a PC.
[0019] Furthermore, the method also includes saving the image and video after step g), and the host computer pushes the image and video to the client's Web or APP terminal to display the detection results.
[0020] Furthermore, in step g), when the staff fails to wear the air respirator correctly as required, the host computer saves the failure to wear the air respirator correctly as an Excel table.
[0021] The beneficial effects of the present invention are as follows: the image processing unit MCU is based on TinyML technology and target detection algorithm, training models, identifying device information collected by the camera, and judging whether the staff is carrying safety equipment; obtaining key location information of the equipment and people, performing location information comparison, determining whether the rectangular frame and gap requirements are met, and judging whether the staff is wearing safety equipment correctly. It realizes intelligent detection of whether the staff is wearing safety equipment correctly according to regulations, reminding the staff to follow the rules and regulations, and avoiding safety accidents caused by incorrect wearing or not wearing safety equipment. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 It is a system detection flow chart of the present invention;
[0023] Figure 2 is an image processing flow chart of the present invention;
[0024] Figure 3 A flowchart for model training of the present invention;
[0025] Figure 4 It is a flow chart of the judgment process of the present invention. DETAILED DESCRIPTION
[0026] The following is combined with Figure 1 To Attachment Figure 4 The present invention is further described.
[0027] As attached Figure 1 As shown, a method for detecting wearing an air respirator based on TinyML technology includes:
[0028] The hardware system consists of an image acquisition module, an image processing unit MCU, and a display module;
[0029] The software system consists of a host computer, a Web terminal, and an APP terminal;
[0030] The detection method is:
[0031] a) Obtain a data set, which is image information of masks and gas cylinders of air respirators at various angles. Use annotation tools to manually annotate the image information of masks and gas cylinders of air respirators at various angles. Use the annotated images to train a TinyML model to obtain a trained TinyML model.
[0032] b) Convert the trained TinyML model into a binary file.
[0033] c) Deploy the binary converted TinyML model to the image processing unit MCU.
[0034] d) After the staff arrives at the image acquisition area, the image acquisition module collects image information of the front, back, left and right sides of the human body. Among the collected image information of the front, back, left and right sides of the human body including the mask and gas cylinder of the air respirator, the image processing unit MCU collects one image as the processing source respectively, and obtains four images.
[0035] e) Save the four images and run the trained TinyML model. If the image contains an air respirator mask and a gas cylinder, execute step f). If the image does not contain an air respirator mask and a gas cylinder, determine that the staff member did not carry an air respirator as required.
[0036] f) MCU captures the mask and gas cylinder of the staff and the air respirator in the 4 images, and obtains the key point position information of the staff, mask and gas cylinder (the position coordinates are based on the actual situation, and the proportion is scaled by inputting the staff height in advance). The mask needs to be placed on the face, the horizontal coordinate of the center point of the mask is x1, and its vertical coordinate is y1. The horizontal coordinate of the center between the two eyes of the staff is x2, and its vertical coordinate is y2. The shoulder straps of the staff must be worn correctly and ensure left-right symmetry to ensure the balance of the gas cylinder behind. The horizontal coordinate of the reflective strip on the left side of the shoulder strap is x3, and its vertical coordinate is y3. The horizontal coordinate of the reflective strip on the right side of the shoulder strap of the staff is x4, and its vertical coordinate is y4. Ensure that the male buckle of the medicine bag is inserted into the female buckle, and the staff's belt forms a continuous and uninterrupted rectangular frame at the navel.
[0037] g) If the following conditions are met simultaneously: |x1-x2|≤5cm, 0≤(y2-y1)≤15cm, the gas cylinder is to be carried behind the body, so the gas cylinder is detected in the rear image and not in the front image, |(x3+x4) / 2-x2|≤10cm, (y3+y4) / 2≤y2, |y3-y4|≤8cm, and the length of the rectangular frame is ≥20cm, then it is determined that the staff member is wearing the air respirator correctly in accordance with regulations. If these conditions are not met simultaneously, the corresponding unsatisfied items are output.
[0038] The image processing unit MCU is based on TinyML technology and target detection algorithms, training models, identifying device information collected by the camera, and judging whether the staff is carrying safety equipment; obtaining key location information of equipment and people, comparing location information, determining whether the location conditions are met, and judging whether the staff is wearing safety equipment correctly. It can be applied to petrochemical enterprises to realize intelligent detection of whether the staff is wearing safety equipment correctly in accordance with regulations, reminding the staff to follow the rules and regulations, and avoiding safety accidents caused by wearing or not wearing safety equipment incorrectly.
[0039] Preferably, the image acquisition module is a camera, and the display module is a PC.
[0040] Furthermore, the method also includes saving the image and video after step g), and the host computer pushes the image and video to the client's Web or APP terminal to display the detection results.
[0041] Furthermore, in step g), when the staff fails to wear the air respirator correctly as required, the host computer saves the failure to wear the air respirator correctly as an Excel table.
[0042] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A method for detecting wearing an air respirator based on TinyML technology, characterized in that: include: The hardware system consists of an image acquisition module, an image processing unit MCU, and a display module; The software system consists of a host computer, a Web terminal, and an APP terminal; The detection method is: a) obtaining a data set, the data set is image information of the mask and gas cylinder of the air respirator at various angles, using a labeling tool to manually label the image information of the mask and gas cylinder of the air respirator at various angles, and using the labeled images to train a TinyML model to obtain a trained TinyML model; b) Convert the trained TinyML model into a binary file; c) Deploy the binary converted TinyML model to the image processing unit MCU; d) After the staff arrives at the image acquisition area, the image acquisition module acquires image information of the front, back, left and right sides of the human body, and the image processing unit MCU acquires one image as a processing source from the acquired image information of the front, back, left and right sides of the human body including the mask and gas cylinder of the air respirator, thereby obtaining four images; e) Save the four images and run the trained TinyML model. If the image contains a mask and a gas cylinder of an air respirator, execute step f). If the image does not contain a mask and a gas cylinder of an air respirator, determine that the staff member did not carry an air respirator as required. f) MCU intercepts the mask and gas cylinder of the staff and the air respirator in the four images, and obtains the key point position information of the staff, the mask, and the gas cylinder. The horizontal coordinate of the center point of the mask is x1, and its vertical coordinate is y1. The horizontal coordinate of the center position between the two eyes of the staff is x2, and its vertical coordinate is y2. The horizontal coordinate of the reflective strip on the left side of the staff's shoulder strap is x3, and its vertical coordinate is y3. The horizontal coordinate of the reflective strip on the right side of the staff's shoulder strap is x4, and its vertical coordinate is y4. The belt of the staff forms a continuous and uninterrupted rectangular frame at the navel; g) If the following conditions are met simultaneously: |x1-x2|≤5cm, 0≤(y2-y1)≤15cm, the gas cylinder is detected in the rear image and not in the front image, |(x3+x4) / 2-x2|≤10cm, (y3+y4) / 2≤y2, |y3-y4|≤8cm, and the length of the rectangular frame is ≥20cm, then it is determined that the staff member is wearing the air respirator correctly as required. If these conditions are not met simultaneously, then the corresponding unmet item is output.
2. The method for detecting wearing an air respirator based on TinyML technology according to claim 1, characterized in that: The image acquisition module is a camera.
3. The method for detecting wearing an air respirator based on TinyML technology according to claim 1, characterized in that: The display module is a PC.
4. The method for detecting wearing an air respirator based on TinyML technology according to claim 1, characterized in that: The method also includes saving the image and video after step g), and the host computer pushes the image and video to the client's Web or APP terminal to display the detection results.
5. The method for detecting wearing an air respirator based on TinyML technology according to claim 1, characterized in that: It also includes that in step g), when the staff fails to wear the air respirator correctly as required, the host computer saves the failure to wear the air respirator as an Excel table.
Citation Information
Patent Citations
Transformer substation respirator state intelligent identification method and system
CN112069886A
Breathing machine wearing detection method, device and equipment and storage medium
CN113780442A
Protective mask wearing compliance detection method and system for electric power welding operation
CN114581810A
