Construction video transmission method based on electric power engineering safety monitoring

Abstract

The invention relates to the technical field of image processing, and particularly discloses a construction video transmission method based on electric power engineering safety monitoring, and the method comprises the following steps: S1, when the real-time network signal intensity is lower than a network signal intensity threshold value, marking the real-time network signal intensity as an adjustment state; s2, taking the construction video in the monitoring time period as a target video, and splitting the target video into target video frames; marking a target area in the target video frame based on an area identification model, and determining a minimum enclosing rectangle of a union set of the target area in the target video frame; s3, calculating a rectangle correction coefficient, and calculating a standard area; and clipping the real-time video frames, re-synthesizing the clipped real-time video frames into a construction video, and transmitting the construction video. According to the invention, the problem of construction video jamming caused by network fluctuation can be avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a construction video transmission method based on power engineering safety monitoring. Background Art

[0002] In actual construction, power engineering projects involve high-voltage electrical operations, which are very dangerous and difficult to construct. For example, power engineering construction usually involves high-voltage electrical operations. When workers come into contact with or handle high-voltage equipment, improper operation may cause accidents such as electric shock. The construction sites of many power projects may be located at high altitudes (such as erecting high-voltage wires), underground (such as laying cables), or in mountainous areas far away from cities. These environments increase the difficulty and risk of construction.

[0003] Through video surveillance, managers can remotely view construction progress and promptly detect and correct violations. Especially in high-risk operations such as high-altitude operations or high-voltage electrical equipment installation, video records can provide valuable safety guarantees. At the same time, construction videos can also be used for post-analysis and training, summarizing safety experience and preventing similar accidents.

[0004] In actual situations, there may be a situation where the network status is poor. In this case, the transmitted construction video will be stuck. The commonly used solution in the prior art is the adaptive bit rate mechanism. Although this method can solve this problem, it may lead to a decrease in image quality and is not suitable for construction site videos that need to focus on certain areas. Based on this, a construction video transmission method based on power engineering safety monitoring is proposed to solve the problem of stuck construction videos when the network status is poor. Summary of the invention

[0005] The purpose of the present invention is to provide a construction video transmission method based on power engineering safety monitoring to solve the following technical problems:

[0006] When the network status is poor, the transmitted construction video will be stuck. The commonly used solution in the existing technology is the adaptive bit rate mechanism. Although this method can solve this problem, it may lead to a decrease in image quality and is not suitable for construction site videos that need to focus on certain areas.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] A construction video transmission method based on power engineering safety monitoring comprises the following steps:

[0009] S1: obtaining the network signal strength C in real time, and marking it as an adjustment state when the network signal strength is lower than a preset network signal strength threshold Cyz;

[0010] S2: Setting a monitoring period [Ttz-T, Ttz], where Ttz represents a time point marked as an adjustment state, and T represents a preset duration, taking the construction video within the monitoring period as a target video, and splitting the target video into target video frames;

[0011] Mark the target area in the target video frame based on the pre-trained area recognition model, the target area is the area where the construction personnel are located in the target video frame, determine the union A of the target areas in the target video frame, and determine the minimum circumscribed rectangle WJA of the union A;

[0012] S3: Calculate the rectangular correction coefficient K = C / Cyz and calculate the standard area S WJA '=S WJA +(Sys-S WJA )*K, Sys represents the total area of ​​the original video frame, S WJA represents the area of ​​the minimum circumscribed rectangle WJA;

[0013] Acquire a real-time construction video, split it into a real-time video frame sequence, crop the real-time video frames in the real-time video frame sequence, the area of ​​the real-time video frame after cropping is the standard area, and the minimum circumscribed rectangle WJA is included in the area where the standard area is located, re-synthesize the cropped real-time video frames into a construction video, and transmit it.

[0014] As a further solution of the present invention: in the step S2, the process of pre-training the region recognition model specifically includes:

[0015] Establishing a database, wherein the database stores images with marked target areas;

[0016] A region recognition model is established based on the deep learning model, and is trained and verified through the database to obtain a pre-trained region recognition model.

[0017] As a further solution of the present invention: the images with the target areas marked and stored in the database are acquired manually.

[0018] As a further solution of the present invention: after a preset time interval, m target video frames are randomly extracted, where m is a preset number, and the target region is determined based on a pre-trained region recognition model;

[0019] The accuracy of the target area is determined manually. If it is inaccurate, the target video frame and the manually determined target area are input into the area recognition model together, and deep learning training is performed again.

[0020] As a further solution of the present invention: the step S3 further includes the following steps:

[0021] The border of each video frame in the re-synthesized real-time construction video is filled to adapt to the screen size.

[0022] As a further solution of the present invention: in the step S2, the construction video is acquired based on a drone, and the drone is at a preset position.

[0023] As a further solution of the present invention: in the step S2, when the area S of the minimum circumscribed rectangle WJA is WJA When ≥0.9Sys, a warning message will be sent to prompt.

[0024] Beneficial effects of the present invention: In the present invention, by monitoring the network signal strength, when it is found that the signal is weak, the system can react in time, enter the "adjustment state" and execute the subsequent corresponding steps to avoid video packet loss or low transmission quality due to poor signal; through the pre-trained area recognition model, the system can automatically identify the key area in the video, that is, the area where the construction personnel are located, and the important part in the construction video is also the area where the construction personnel are located. The specific size of this area can be set according to experience during the initial training model; when the network signal is weak, by calculating the standard area of ​​cutting, the system can compress the data volume of the video transmission, but ensure that the key construction area (i.e., the construction personnel) is still within the monitoring range to avoid losing important pictures; through the above scheme, the transmission efficiency of the construction video can be optimized when the network signal is weak, while ensuring that the safety monitoring of the construction personnel is not affected, by cutting irrelevant areas in the video, focusing on the area where the construction personnel are located, reducing the data volume of the video transmission, thereby improving the fluency and reliability of the video transmission. The present invention can avoid the problem of construction video jamming caused by network fluctuations. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The present invention will be further described below in conjunction with the accompanying drawings.

[0026] Figure 1 The present invention is a flow chart of a construction video transmission method based on power engineering safety monitoring. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] See also Figure 1As shown, the present invention is a construction video transmission method based on power engineering safety monitoring, comprising the following steps:

[0029] S1: obtaining the network signal strength C in real time, and marking it as an adjustment state when the network signal strength is lower than a preset network signal strength threshold Cyz;

[0030] S2: Setting a monitoring period [Ttz-T, Ttz], where Ttz represents a time point marked as an adjustment state, and T represents a preset duration, taking the construction video within the monitoring period as a target video, and splitting the target video into target video frames;

[0031] Mark the target area in the target video frame based on the pre-trained area recognition model, the target area is the area where the construction personnel are located in the target video frame, determine the union A of the target areas in the target video frame, and determine the minimum circumscribed rectangle WJA of the union A;

[0032] S3: Calculate the rectangular correction coefficient K = C / Cyz and calculate the standard area S WJA '=S WJA +(Sys-S WJA )*K, Sys represents the total area of ​​the original video frame, S WJA represents the area of ​​the minimum circumscribed rectangle WJA;

[0033] Acquire a real-time construction video, split it into a real-time video frame sequence, crop the real-time video frames in the real-time video frame sequence, the area of ​​the real-time video frame after cropping is the standard area, and the minimum circumscribed rectangle WJA is included in the area where the standard area is located, re-synthesize the cropped real-time video frames into a construction video, and transmit it.

[0034] It should be noted that by monitoring the network signal strength, when the signal is found to be weak, the system can respond in time, enter the "adjustment state" and execute the subsequent corresponding steps to avoid video packet loss or low transmission quality due to poor signal; through the pre-trained area recognition model, the system can automatically identify the key area in the video, that is, the area where the construction workers are located, and the important part of the construction video is also the area where the construction workers are located. The specific size of this area can be set according to experience during the initial training model; when the network signal is weak, by calculating the standard cropping area, the system can compress the amount of data transmitted by the video, but ensure that the key construction area (that is, the construction workers) is still within the monitoring range to avoid the loss of important pictures; through the above scheme, the transmission efficiency of the construction video can be optimized when the network signal is weak, while ensuring that the safety monitoring of the construction workers is not affected. By cropping irrelevant areas in the video and focusing on transmitting the area where the construction workers are located, the amount of data transmitted by the video is reduced, thereby improving the smoothness and reliability of the video transmission.

[0035] In another preferred embodiment of the present invention, in step S2, the process of pre-training the region recognition model specifically includes:

[0036] Establishing a database, wherein the database stores images with marked target areas;

[0037] A region recognition model is established based on the deep learning model, and is trained and verified through the database to obtain a pre-trained region recognition model.

[0038] It is worth noting that in order to train an efficient area recognition model, it is necessary to first establish an image dataset containing known correct labels as the basis for model training. This dataset can cover various different situations in construction scenes to ensure that the model has strong generalization capabilities; establishing a labeled database can provide the model with a large amount of real and verified data to ensure that the model can be fully trained during the learning process, so as to accurately identify the target area; deep learning models, especially convolutional neural networks, have powerful image feature extraction and recognition capabilities, and can automatically learn different features from images to accurately identify the area where construction workers are located. Through repeated training, the model can gradually master the characteristics of identifying the area where construction workers are located.

[0039] In another preferred embodiment of the present invention, the images with the target areas marked and stored in the database are manually acquired.

[0040] In another preferred embodiment of the present invention, after a preset time interval, m target video frames are randomly extracted, where m is a preset number, and the target region is determined based on a pre-trained region recognition model;

[0041] The accuracy of the target area is determined manually. If it is inaccurate, the target video frame and the manually determined target area are input into the area recognition model together, and deep learning training is performed again.

[0042] It should be noted that even for pre-trained deep learning models, in actual applications, recognition accuracy may decrease due to factors such as environmental changes, insufficient data sets, and complex human movements, and it is impossible to completely avoid misrecognition. Therefore, regularly randomly extracting target video frames and performing manual verification can discover and correct potential errors in these model recognitions; randomly extracting video frames means that the model will be exposed to diverse data at different time points and in different scenarios, ensuring that the model will not only perform well in specific scenarios, but can effectively identify target areas in a wide range of construction environments.

[0043] In another preferred embodiment of the present invention, the step S3 further includes the following steps:

[0044] The border of each video frame in the re-synthesized real-time construction video is filled to adapt to the screen size.

[0045] It should be noted that in the previous steps, the real-time construction video has been cropped. The frame size of the cropped video is usually smaller and no longer the original full-size video. When such a video is transmitted to display screens of different sizes, it may not fill the entire screen; different screen devices may have different resolutions or display ratios. During the video transmission process, if appropriate adjustments are not made, the cropped video may show black edges, stretching, or incompleteness on some devices. By filling the borders, these problems can be avoided and the video display effect can be guaranteed to be consistent.

[0046] In another preferred embodiment of the present invention, in the step S2, the construction video is acquired based on a drone, and the drone is at a preset position.

[0047] In another preferred embodiment of the present invention, in step S2, when the area S of the minimum circumscribed rectangle WJA is WJA When ≥0.9Sys, a warning message will be sent to prompt.

[0048] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A construction video transmission method based on power engineering safety monitoring, characterized in that: The following steps are involved: S1: obtaining the network signal strength C in real time, and marking it as an adjustment state when the network signal strength is lower than a preset network signal strength threshold Cyz; S2: Setting a monitoring period [Ttz-T, Ttz], where Ttz represents a time point marked as an adjustment state, and T represents a preset duration, taking the construction video within the monitoring period as a target video, and splitting the target video into target video frames; Mark the target area in the target video frame based on the pre-trained area recognition model, the target area is the area where the construction personnel are located in the target video frame, determine the union A of the target areas in the target video frame, and determine the minimum circumscribed rectangle WJA of the union A; S3: Calculate the rectangular correction coefficient K = C / Cyz and calculate the standard area S WJA '=S WJA +(Sys-S WJA )*K, Sys represents the total area of ​​the original video frame, S WJA represents the area of ​​the minimum circumscribed rectangle WJA; Acquire a real-time construction video, split it into a real-time video frame sequence, crop the real-time video frames in the real-time video frame sequence, the area of ​​the real-time video frame after cropping is the standard area, and the minimum circumscribed rectangle WJA is included in the area where the standard area is located, re-synthesize the cropped real-time video frames into a construction video, and transmit it.

2. A construction video transmission method based on power engineering safety monitoring according to claim 1, characterized in that: In step S2, the process of pre-training the region recognition model specifically includes: Establishing a database, wherein the database stores images with marked target areas; A region recognition model is established based on the deep learning model, and is trained and verified through the database to obtain a pre-trained region recognition model.

3. A construction video transmission method based on power engineering safety monitoring according to claim 2, characterized in that: The images with the target areas marked and stored in the database are acquired manually.

4. A construction video transmission method based on power engineering safety monitoring according to claim 2, characterized in that: After a preset time interval, m target video frames are randomly extracted, where m is a preset number, and the target region is determined based on a pre-trained region recognition model; The accuracy of the target area is determined manually. If it is inaccurate, the target video frame and the manually determined target area are input into the area recognition model together, and deep learning training is performed again.

5. A construction video transmission method based on power engineering safety monitoring according to claim 1, characterized in that: The step S3 further includes the following steps: The border of each video frame in the re-synthesized real-time construction video is filled to adapt to the screen size.

6. A construction video transmission method based on power engineering safety monitoring according to claim 1, characterized in that: In the step S2, the construction video is acquired based on a drone, and the drone is at a preset position.

7. A construction video transmission method based on power engineering safety monitoring according to claim 1, characterized in that: In the step S2, when the area S of the minimum bounding rectangle WJA is WJA When ≥0.9Sys, a warning message will be sent to prompt.

Images