Cyanide warehouse personnel detection method and detection system based on cavity convolution technology

Abstract

The invention discloses a cyanide warehouse personnel detection method and detection system based on a cavity convolution technology. The detection method comprises the following steps: S1, acquiring video images inside and outside a warehouse door; s2, training the historical video image to obtain a personnel detection model; s3, the opening and closing state of the warehouse door is detected; s4, identifying the video image by using a personnel detection model and carrying out statistics to obtain personnel information inside and outside the warehouse door; s5, whether an early warning condition is triggered or not is judged according to the personnel information inside and outside the warehouse door and the opening and closing state of the warehouse door; and S6, when the early warning condition is triggered, performing alarm operation through an audible and visual alarm. The detection system comprises a limit switch, a network controller, an audible and visual alarm, a camera, a switch and a video analysis server, the limit switch and the audible and visual alarm are both connected with the network controller, and the network controller and the camera are both connected with the video analysis server through the switch.

Description

Technical Field

[0001] The present invention relates to the field of security monitoring, and in particular to a method and system for detecting personnel in a cyanide warehouse based on dilated convolution technology. Background Art

[0002] Due to its special chemical properties, cyanide is widely used in the industrial production field. However, cyanide itself is a highly toxic substance, and there is a great danger in inappropriate use or storage. As a special place for storing such highly toxic chemicals, strict safety management of cyanide warehouses can ensure safe production to the greatest extent and reduce potential hazards to the environment and human health.

[0003] Currently, the management of cyanide warehouses mainly adopts strict access systems such as two-person double-lock, requisition approval, and special person in charge, which can ensure the safety of cyanide use to a great extent. However, the existing management methods for cyanide warehouses generally rely on manual labor. Behaviors such as manual inspection of surveillance cameras and manual registration are prone to errors, and there is no systematic and scientific management method for historical records and historical surveillance data, resulting in great resistance when it is necessary to consult and analyze, bringing certain management burdens and potential safety hazards to the management of cyanide warehouses.

[0004] Therefore, how to overcome the deficiencies of the existing cyanide warehouse management system, further strengthen the safety management of cyanide warehouses, improve management efficiency, and adopt intelligent technologies has become the focus of the next research. Summary of the Invention

[0005] The object of the present invention is to provide a method and system for detecting personnel in a cyanide warehouse based on dilated convolution technology, which can improve management efficiency and the safety of cyanide warehouses.

[0006] To achieve the above object, the technical solution of the present invention is as follows:

[0007] A method for detecting personnel in a cyanide warehouse based on dilated convolution technology, comprising the following steps:

[0008] S1. Obtain video images inside and outside the warehouse door;

[0009] S2. Train a personnel detection model with historical video images;

[0010] S3. Detect the opening and closing state of the warehouse door;

[0011] S4. Use the personnel detection model to identify the video images and statistically obtain the personnel information inside and outside the warehouse door;

[0012] S5. Determine whether a warning situation is triggered according to the personnel information inside and outside the warehouse door in combination with the opening and closing state of the warehouse door;

[0013] S6. When a warning situation is triggered, an alarm operation is carried out through an audible and visual alarm.

[0014] Furthermore, in step S1, cameras are respectively set directly opposite the warehouse door and inside the warehouse to perform video image acquisition operations.

[0015] Furthermore, in step S2, the historical video images are trained through the DC-YOLOv5 model; the DC-YOLOv5 model uses dilated convolution to calculate the size of the feature map, and its calculation formula is:

[0016]

[0017] where k is the size of the original convolution kernel, d is the dilation rate, i is the size of the input feature layer, s is the stride, p is the padding, and o is the size of the output feature map after dilated convolution;

[0018] The calculation formula for the actual receptive field size RF of the dilated convolution kernel is:

[0019] RF = (k - 1) × (d - 1) + 1.

[0020] Furthermore, in step S3, a limit switch is set on the warehouse door to detect the opening and closing state of the warehouse door.

[0021] Furthermore, in step S4, the personnel information inside and outside the warehouse door includes the number of personnel in the designated area outside the warehouse door, the number of personnel entering the warehouse, the identities of the personnel entering the warehouse, and the number of personnel closing the door.

[0022] Furthermore, in step S5, when judging whether a warning situation is triggered, it is judged whether the number of personnel inside and outside the warehouse door reaches the set standard according to the personnel information inside and outside the warehouse door. When the set standard is not reached, a warning of insufficient authorized personnel is triggered; when the set standard is reached, the opening and closing state of the warehouse door is detected. When the warehouse door is in the open state, the personnel entering and leaving the warehouse are counted and face recognition is performed. When the face recognition fails, a warning of illegal entry of personnel is triggered. At the same time, when the number of personnel inside the warehouse is less than 2, a warning of paying attention to locking the door is triggered; when the warehouse door is in the closed state, it ends.

[0023] Furthermore, in step S6, the audible and visual alarm has a variety of prompt sounds and can perform corresponding alarm prompt operations according to different warning situations.

[0024] A detection system, which includes a limit switch, a network controller, an audible and visual alarm, a camera, a switch, and a video analysis server. The limit switch is installed on the warehouse door to detect the opening and closing state of the warehouse door. The limit switch and the audible and visual alarm are both connected to the network controller. The network controller and the camera are both connected to the video analysis server through the switch to transmit the status signal of the limit switch and the captured image of the camera to the video analysis server.

[0025] Further, the detection system also includes a power supply module, which is respectively connected to the limit switch, the network controller, and the audible and visual alarm to provide power supply for them.

[0026] Further, a cyanide warehouse personnel access monitoring platform is set on the video analysis server to play the real-time video of the camera, receive alarm event information, display the latest access control records, display alarm screenshots and videos, and generate an alarm statistics ledger.

[0027] Compared with the prior art, the advantages and positive effects of the present invention are:

[0028] 1. Enhance warehouse safety management: Through real-time monitoring and intelligent analysis, it can effectively prevent unauthorized personnel from entering the cyanide warehouse and ensure that there is always a necessary number of safe people in the warehouse. Once an abnormal situation occurs, the detection system will immediately issue a warning and trigger the audible and visual alarm, greatly improving the safety of the cyanide warehouse.

[0029] 2. High recognition accuracy: By using the improved YOLOv5 model (DC-YOLOv5), the detection system can expand the receptive field through dilated convolution and obtain richer context information without increasing the network complexity, so as to accurately identify the number of people in the area, entry and exit behaviors, and facial features.

[0030] 3. Flexibility and scalability: This detection system is not only applicable to cyanide warehouses, but can also be adjusted according to needs and applied to other places that require strict safety management. In addition, with the development of technology, the performance of the detection system can be further improved by updating algorithms or adding new sensors.

[0031] 4. Historical data analysis: This detection system can automatically record all events and their related video materials, which is convenient for subsequent query and analysis, provides scientific data support for managers, and helps optimize management and formulate emergency plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0033] Figure 1 It is a logical flowchart of the detection method;

[0034] Figure 2 It is an installation effect diagram of the detection system;

[0035] Figure 3 It is a connection structure diagram of the detection system;

[0036] Figure 4 It is an interface display effect diagram of the detection system. Detailed implementation manners

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts, any modifications, equivalent replacements, improvements, etc., shall be included in the protection scope of the present invention.

[0038] As Figure 1 shown, the present invention discloses a cyanide warehouse personnel detection method based on the dilated convolution technology, including the following steps:

[0039] S1. Obtain video images: Install a camera at the warehouse door and inside the warehouse respectively to obtain the video images inside and outside the warehouse door;

[0040] One camera installed at the warehouse door needs to maintain a certain distance from the cyanide warehouse and face the warehouse door directly to obtain the best field of view and eliminate blind spots (as Figure 2 shown). To avoid flying wires, a bridge is used for communication; at the same time, the video images obtained by the two cameras are transmitted to the video access center for subsequent processing.

[0041] S2. Train the personnel detection model: Use the video analysis server to obtain the video images of the above two cameras and train the personnel detection model based on the obtained video images;

[0042] The personnel detection model is trained based on the improved YOLOv5 model (DC-YOLOv5), which can achieve functions such as accurate personnel counting within a region, detection of personnel crossing the line in and out, and personnel face recognition. The specific improvements in the improved YOLOv5 model (DC-YOLOv5) are as follows: Atrous convolution technology is introduced in the feature extraction part of the model, and conventional convolutions are replaced with atrous convolutions with a larger dilation rate in some layers. This can effectively expand the receptive field of the feature map and obtain richer context information without increasing the depth and width of the network. When introducing atrous convolution, the formula for calculating the size of the feature map is as follows:

[0043] Let the original convolution kernel size be k, the dilation rate be d, the size of the input feature layer be i, the stride be s, and the padding be p. Then the size o of the output feature map after atrous convolution can be calculated by the following formula:

[0044]

[0045] The actual receptive field size RF of the atrous convolution kernel is:

[0046] RF = (k - 1) × (d - 1) + 1

[0047] In practice, the appropriate dilation rate d needs to be selected according to the task requirements and experimental results.

[0048] S3. Detect the opening and closing state of the warehouse door: Install a limit switch on the warehouse door, and obtain the digital input signal of the limit switch through the network controller module, and then obtain the opening and closing state of the warehouse door;

[0049] The limit switch is connected in a normally closed manner. The network controller module reads the digital input signal of the limit switch through the Modbus TCP protocol, and represents the digital input signal of the obtained limit switch by high and low levels.

[0050] S4. Personnel detection: Use the personnel detection model to identify and count the number of personnel in the designated area outside the warehouse, the number of personnel entering the warehouse and their identities, and the number of personnel closing the door;

[0051] Draw a detection line at the warehouse door frame, and start the detection of personnel entering and leaving the warehouse with personnel crossing the line as the trigger condition. The identity of the personnel entering the warehouse is realized through personnel face recognition. The counting of the number of personnel closing the door is triggered by the change in the digital input state of the limit switch when the door is closed, and the frame image at this time is intercepted and the number of personnel in the designated area is counted; at the same time, to improve the detection efficiency and reduce power consumption, personnel detection is started with personnel entering the designated area as the trigger condition.

[0052] S5. Early warning information processing: When it is detected that the warehouse door is in the open state, the number of people in the designated area outside the warehouse does not meet the standard, the people entering the warehouse do not have relevant permissions, the number of people staying in the warehouse is less than two, or the number of people closing the door is less than two, relevant early warning information is triggered and processed;

[0053] When the early warning information is triggered, the video analysis system will intercept and save the pictures and videos related to the early warning information for subsequent processing and analysis.

[0054] S6. Triggering audible and visual alarms: An audible and visual alarm module is installed on one side outside the warehouse. This audible and visual alarm module will receive the processed and issued early warning information in real time and send out corresponding audible and visual alarm information according to the early warning information.

[0055] The processed early warning information will be sent to the audible and visual alarm module in the form of an instruction through the network controller module. At the same time, alarm voices that meet the requirements can be customized according to different early warning information and instructions can be generated, such as "Insufficient authorized personnel", "Please note to lock the door", etc.

[0056] The present invention also discloses a detection system for implementing the above detection method. This detection system can provide strong guarantee for the safety management of cyanide warehouses and has the characteristics of being easy to deploy, maintain, expand, etc.

[0057] The detection system includes a hardware part and a software part. The hardware part includes a power supply module, a limit switch, a network controller module, an audible and visual alarm module, a camera, a switch, and a video analysis server, as Figure 3 shown.

[0058] The power supply module is used to supply power to the limit switch, network controller module, and audible and visual alarm module; the network controller module is respectively connected to the limit switch, audible and visual alarm module, and switch, and is used to receive the digital quantity signal of the limit switch and send the alarm information fed back by the system to the audible and visual alarm module; the switch is respectively connected to the network controller module, camera, and video analysis server, and is used for network resource integration and data forwarding; the video analysis server is communicatively connected to the network controller module and camera through the switch, and is used for personnel detection model training and intelligent video analysis.

[0059] The software part, as a personnel access monitoring platform for cyanide warehouses, mainly includes functions such as playing the real-time images of the camera, receiving alarm event information, displaying the latest access control records, showing alarm screenshots and videos, and forming an alarm statistics ledger, etc., for better assisting management personnel in the safety control of cyanide warehouses. The interface effect of the monitoring platform is as Figure 4 shown.

[0060] The present invention has the following beneficial effects:

[0061] 1. Enhance warehouse safety management: Through real-time monitoring and intelligent analysis, it can effectively prevent unauthorized personnel from entering the cyanide warehouse and ensure that the necessary safety personnel are always maintained in the warehouse. Once an abnormal situation occurs, the detection system will immediately issue an alarm and trigger an audible and visual alarm, greatly improving the safety of the cyanide warehouse.

[0062] 2. High recognition accuracy: By using the improved YOLOv5 model (DC-YOLOv5), the detection system can expand the receptive field through dilated convolution to obtain richer context information without increasing the network complexity, thereby achieving accurate recognition of the number of personnel, entry and exit behaviors, and facial features in the area.

[0063] 3. Flexibility and scalability: This detection system is not only applicable to cyanide warehouses but can also be adjusted and configured for use in other places that require strict safety management as needed. In addition, with the development of technology, the performance of the detection system can be further improved by updating algorithms or adding new sensors.

[0064] 4. Historical data analysis: This detection system can automatically record all events and their related video materials, facilitating post-event query and analysis, providing scientific data support for managers, and helping to optimize management and formulate emergency plans.

Claims

1. A method for detecting personnel in a cyanide warehouse based on dilated convolution technology, characterized in that: It includes the following steps: S1. Obtain video images inside and outside the warehouse door; S2. Train a person detection model with historical video images; S3. Detect the opening and closing state of the warehouse door; S4. Use the person detection model to identify the video images and statistically obtain the personnel information inside and outside the warehouse door; S5. Determine whether to trigger an early warning situation based on the personnel information inside and outside the warehouse door in combination with the opening and closing state of the warehouse door; S6. When an early warning situation is triggered, perform an alarm operation through an audible and visual alarm.

2. The method for detecting personnel in a cyanide warehouse based on the dilated convolution technique according to claim 1, wherein: In step S1, cameras are respectively set directly opposite the warehouse door and inside the warehouse to obtain video images.

3. The method for detecting personnel in a cyanide warehouse based on the dilated convolution technology according to claim 2, wherein: In step S2, the historical video images are trained by the DC-YOLOv5 model; the DC-YOLOv5 model uses dilated convolution to calculate the size of the feature map, and its calculation formula is: where k is the original convolution kernel size, d is the dilation rate, i is the size of the input feature layer, s is the stride, p is the padding, and o is the size of the output feature map after dilated convolution; The calculation formula for the actual receptive field size RF of the dilated convolution kernel is: RF = (k - 1)×(d - 1)+1.

4. The method for detecting personnel in a cyanide warehouse based on the dilated convolution technique according to claim 3, wherein: In step S3, a limit switch is set on the warehouse door to detect the opening and closing state of the warehouse door.

5. The method for detecting personnel in a cyanide warehouse based on the dilated convolution technology according to claim 4, characterized in that: In step S4, the personnel information inside and outside the warehouse door includes the number of people in the designated area outside the warehouse door, the number of people entering the warehouse, the identities of the people entering the warehouse, and the number of people closing the door.

6. The method for detecting personnel in a cyanide warehouse based on the dilated convolution technology according to claim 5, wherein: In step S5, when determining whether to trigger an early warning situation, it is judged whether the number of people inside and outside the warehouse door reaches the set standard according to the personnel information inside and outside the warehouse door. When the set standard is not reached, a warning of insufficient authorized personnel is triggered; when the set standard is reached, the opening and closing state of the warehouse door is detected. When the warehouse door is in the open state, the people entering and leaving the warehouse are counted and face recognition is performed. When the face recognition fails, an early warning of illegal entry of personnel is triggered, and at the same time, when the number of people in the warehouse is less than 2, a warning of attention to lock the door is triggered; when the warehouse door is in the closed state, it ends.

7. The method for detecting personnel in a cyanide warehouse based on the dilated convolution technique according to claim 6, characterized in that: In step S6, the audible and visual alarm has a variety of prompt sounds and can perform corresponding alarm prompt operations according to different early warning situations.

8. A detection system for implementing the cyanide warehouse personnel detection method based on the dilated convolution technology according to claim 1, characterized in that: The detection system includes a limit switch, a network controller, an audible and visual alarm, a camera, a switch, and a video analysis server. The limit switch is installed on the warehouse door to detect the opening and closing state of the warehouse door. The limit switch and the audible and visual alarm are both connected to the network controller. The network controller and the camera are both connected to the video analysis server through the switch to transmit the state signal of the limit switch and the captured image of the camera to the video analysis server.

9. The detection system according to claim 8, characterized in that: The detection system also includes a power supply module, and the power supply module is respectively connected to the limit switch, the network controller, and the audible and visual alarm to provide power supply for them.

10. The detection system according to claim 8, wherein: A monitoring platform for the entry and exit of personnel in a cyanide warehouse is set on the video analysis server to play the real-time picture of the camera, receive alarm event information, display the latest access control records, display alarm screenshots and videos, and generate an alarm statistics ledger.

Images