Hydropower station construction safety management method and system based on artificial intelligence

Abstract

The invention relates to the crossing field of artificial intelligence and civil engineering, and particularly discloses a hydropower station construction safety management method and system based on artificial intelligence. The target image to be detected is analyzed through the peccancy detection model, whether the peccancy behavior exists in the target image or not is judged, if the peccancy behavior exists, the peccancy report is output, safety management personnel are prompted to solve safety risks in time, and a user can adjust detection items of the peccancy detection model according to requirements; therefore, detection and identification of multiple security risks are realized, the detection range of the violation behaviors is expanded, the missing detection range of the violation behaviors is narrowed, and the security risks of construction projects are reduced.

Description

Technical Field

[0001] The present invention belongs to the cross - technical field of artificial intelligence and civil engineering, and particularly relates to a safety management method and system for hydropower station construction based on artificial intelligence. Background Art

[0002] During the construction process of a hydropower station, there are various safety risks, including the safety of construction workers, equipment safety, environmental safety, etc. These safety risks are all related to the progress of the hydropower station project. In order to ensure that the hydropower station can be completed on schedule and with high quality, it is necessary to strictly prevent safety risks that occur during the project progress.

[0003] Traditional engineering safety management methods mainly rely on manual monitoring and manual inspections. Safety management personnel need to conduct safety patrols on the construction site of the project, and identify safety risks through the knowledge and experience of safety management personnel. It can be seen that such a safety management method is too dependent on safety management personnel and has strong subjectivity, making it difficult to achieve real - time, accurate, and efficient safety warnings. With the progress of artificial intelligence technology, AI - based safety management systems have gradually developed. AI safety management systems can identify engineering safety risks through data analysis and machine learning. However, the current AI safety management systems are mainly used to identify relatively simple safety risk items such as safety helmets. There are various safety risks at the construction site of water conservancy projects, and it is not only necessary to identify whether a safety helmet is worn. Therefore, it is necessary to design a safety management method and system that can identify various safety risks according to the needs of users. Summary of the Invention

[0004] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a safety management method and system for hydropower station construction based on artificial intelligence, so as to solve the problem that the traditional engineering safety management system can only handle single safety risk problems and cannot effectively identify various complex safety risks.

[0005] To achieve the above - mentioned purpose, the present invention provides the following technical solutions:

[0006] A safety management method for hydropower station construction based on artificial intelligence, comprising:

[0007] Obtain a target image to be detected;

[0008] Input the obtained target image into a violation detection model;

[0009] The violation detection model analyzes the target image, detects whether there is a violation phenomenon in the target image. If there is a violation phenomenon, the violation detection model determines that there is a violation behavior and outputs a violation report. If there is no violation phenomenon, the violation detection model determines that the behavior is compliant.

[0010] Preferably, the method for establishing the violation detection model includes:

[0011] Obtain pictures of violation behaviors, make a dataset of violation behaviors, and divide the dataset of violation behaviors into a training set and a test set according to the quantity ratio of 8:2;

[0012] Use a data annotation tool to annotate the violation behaviors shown in the pictures in the dataset. For each type of violation behavior, draw a bounding box and assign a label to each box;

[0013] Select the YOLO model, convert the annotated dataset file into the corresponding YOLO format, and use the training set file in the YOLO format to train the YOLO model to obtain a violation detection model for monitoring violation behaviors;

[0014] Use the test set file to evaluate the violation detection model, evaluate the accuracy rate, recall rate, F2-score, and false detection and missed detection indicators of the violation detection model, and adjust the violation detection model according to the test results.

[0015] Preferably, the violation detection model is separately installed at the front end and the back end, and respectively identify the violation behaviors in the target image to be detected at the construction site where the front-end device is located and the management center where the back-end device is located, mark the target through the bounding box, issue an alarm for the detected violation behaviors and notify the safety management personnel, and at the same time generate a violation report including the violation content, violation classification, violation level, and violation regulations corresponding to the reported violation phenomenon.

[0016] Preferably, the violation detection model collects the pictures of the construction project site collected during the use process as new sample data of the dataset, confirms and annotates the new violation behaviors in the new sample data as incremental data for online learning, combines the incremental data with the data in the original training set, and continues to train the model until the violation detection model adapts to the new violation behaviors.

[0017] Preferably, set an update period, summarize the pictures of the violation behaviors misjudged and missed in an update period, establish an additional training dataset, and use the additional training dataset to retrain the model at the end of the update period.

[0018] Preferably, the acquisition method of the target image to be detected includes automatic shooting by the camera equipment at the construction project site and active shooting by the safety management personnel.

[0019] A hydropower station construction safety management system based on artificial intelligence includes:

[0020] An acquisition module, used to acquire the target image to be detected at the construction project site;

[0021] The illegal behavior detection module analyzes and determines whether there is an illegal behavior in the target image to be detected through an illegal detection model;

[0022] The alarm module is used to alarm and prompt the illegal behavior detected by the illegal behavior detection module;

[0023] The storage module is used to store the target image to be detected and the illegal detection model.

[0024] Preferably, the illegal behavior detection module includes:

[0025] An illegal detection model for detecting whether there is an illegal behavior in the target image to be detected;

[0026] The model optimization module is used to introduce an online learning mechanism to continuously optimize the illegal detection model during the usage period.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] The present invention analyzes the target image to be detected through an illegal detection model, determines whether there is an illegal behavior in the target image, and if there is an illegal behavior, an illegal report will be output to prompt the safety management personnel to solve the safety risk in time. The user can adjust the detection items of the illegal detection model according to needs to realize the detection and identification of multiple safety risks, thereby expanding the detection range of illegal behaviors, reducing the undetected range of illegal behaviors, and reducing the safety risks of construction projects. Description of the Drawings

[0029] Figure 1 It is a block diagram of the safety management method of the present invention;

[0030] Figure 2 It is a block diagram of the safety management system structure of the present invention. Specific Embodiments

[0031] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0032] Embodiment 1:

[0033] Please refer to Figure 1 shown, a safety management method for hydropower station construction based on artificial intelligence, including:

[0034] Obtain the target image to be detected. The acquisition methods of the target image to be detected include automatic recording by the camera equipment at the construction site and active recording by the safety management personnel. Through automatic collection by equipment and manual supplementary collection, it is expected to obtain as many target images as possible at the construction site to facilitate the subsequent identification of as many safety risks as possible;

[0035] Input the obtained target image into the violation detection model;

[0036] The violation detection model analyzes the target image to detect whether there is a violation in the target image. If there is a violation, the violation detection model determines that there is a violation behavior and outputs a violation report. If there is no violation, the violation detection model determines that the behavior is compliant.

[0037] As can be seen from the above, the violation detection model analyzes the target image to be detected to determine whether there is a violation behavior in the target image. If there is a violation behavior, a violation report will be output to prompt the safety management personnel to solve the safety risk in time. The user can adjust the detection items of the violation detection model according to the needs to achieve the detection and identification of multiple safety risks, thereby expanding the detection range of violation behaviors, reducing the missed detection range of violation behaviors, and reducing the safety risks of construction projects.

[0038] The method for establishing the violation detection model includes:

[0039] Obtain violation behavior pictures and make a violation behavior data set. As many violation behavior pictures as possible are set in the data set. The more the data set, the more accurate the subsequent trained recognition model will be. Divide the violation behavior data set into a training set and a test set according to the quantity ratio of 8:2, which are used to train the model and test the model respectively;

[0040] Use a data annotation tool to annotate the violation behaviors shown in the pictures in the data set. For each violation behavior, draw a bounding box and assign a label to each box;

[0041] Select the YOLO model. The YOLO model can be selected as YOLO5 or YOLO8. Convert the annotated data set file into the corresponding YOLO format and use the training set file in YOLO format to train the YOLO model to obtain a violation detection model for monitoring violation behaviors;

[0042] Evaluate the violation detection model using the test set file, and evaluate the accuracy rate, recall rate, F2-score, false detection and missed detection indicators of the violation detection model, which are used to evaluate the accuracy of the model in identifying construction site violations, the proportion of all violations that the model can correctly identify, the indicator that comprehensively considers the accuracy rate and recall rate, and analyze the possible false detections and missed detections of the model. Adjust the violation detection model according to the test results, and the performance of the model can be further improved by adjusting the network architecture, training data set, and data augmentation technology, etc.

[0043] The violation detection model is separately set at the front end and the back end, and identifies the violation behaviors in the target image to be detected at the construction site where the front-end device is located and the management center where the back-end device is located respectively. The violation detection model set in the front end is used to identify some relatively simple and highly important safety risk items, such as safety helmets, safety reflective vests, open flames and other safety risks. For some violation detection models with high difficulty in identifying safety risk items, because of the large processing difficulty, they are set at the back end, and the superior processing devices at the back end are used to identify the complex safety risks in the target image, such as dismantling the scaffolding in reverse order, not setting safety protection facilities in special areas, etc. Mark the target through the bounding box, and send out an alarm for the detected violation behaviors and notify the safety management personnel. At the same time, generate a violation report including the violation content, violation classification, violation level, and violation regulations corresponding to the reported violation phenomenon.

[0044] In order to improve the recognition ability of the violation detection model, the violation detection model collects the construction project site pictures collected during the use process as the new sample data of the data set, confirms and marks the new violation behaviors in the new sample data as the incremental data for online learning, combines the incremental data with the data in the original training set, and continues to train the model until the violation detection model adapts to the new violation behaviors. Exercise and improve the recognition ability of the violation detection model during the use process, continuously enrich the recognition items of the violation detection model, and then improve the violation detection model.

[0045] Set an update cycle, summarize the pictures of the violation behaviors misjudged and missed in an update cycle, establish an additional training data set, and use the additional training data set to retrain the model at the end of the update cycle. Regularly detect and check the update of the violation detection model, repair the deficiencies of the violation detection model, and use the samples misidentified by the violation detection model to conduct targeted training on the violation detection model to fill the deficiencies of the violation detection model.

[0046] An artificial intelligence-based safety management system for hydropower station construction, including:

[0047] An acquisition module, which is used to acquire the target image to be detected at the construction project site;

[0048] The illegal behavior detection module analyzes and determines whether there is an illegal behavior in the target image to be detected through an illegal detection model;

[0049] The alarm module is used to alarm and prompt the illegal behavior detected by the illegal behavior detection module;

[0050] The storage module is used to store the target image to be detected and the illegal detection model.

[0051] The engineering site image is collected through the collection module, and the engineering site image is detected by the illegal behavior detection module to identify whether there is an illegal behavior in the image. If there is an illegal behavior, the alarm module issues an alarm to prompt the safety management personnel to deal with the illegal behavior in time.

[0052] The illegal behavior detection module includes:

[0053] The illegal detection model is used to detect whether there is an illegal behavior in the target image to be detected;

[0054] The model optimization module is used to introduce an online learning mechanism to continuously optimize the illegal detection model during the usage period.

[0055] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not have to be directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without conflict, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0056] In the drawings of the disclosed embodiments of the present invention, only the structures related to the disclosed embodiments are involved, and other structures can refer to the general design. Without conflict, the same embodiment and different embodiments of the present invention can be combined with each other.

Claims

1. A safety management method for hydropower station construction based on artificial intelligence, characterized in that Including: Obtain a target image to be detected; Input the obtained target image into a violation detection model; The violation detection model analyzes the target image to detect whether there is a violation in the target image. If there is a violation, the violation detection model determines that there is a violation behavior and outputs a violation report. If there is no violation, the violation detection model determines that the behavior is compliant.

2. The safety management method for hydropower station construction based on artificial intelligence according to claim 1, characterized in that: The method for establishing the violation detection model includes: Obtain violation behavior pictures, create a violation behavior dataset, and divide the violation behavior dataset into a training set and a test set according to a quantity ratio of 8:2; Use a data annotation tool to annotate the violation behaviors shown in the pictures in the dataset. For each violation behavior, draw a bounding box and assign a label to each box; Select the YOLO model, convert the annotated dataset file into the corresponding YOLO format, and use the training set file in the YOLO format to train the YOLO model to obtain a violation detection model for monitoring violation behaviors; Use the test set file to evaluate the violation detection model, evaluate the accuracy rate, recall rate, F2-score, and false detection and missed detection indicators of the violation detection model, and adjust the violation detection model according to the test results.

3. The safety management method for hydropower station construction based on artificial intelligence according to claim 2, characterized in that: The violation detection model is separately set at the front end and the back end, and identifies the violation behaviors in the target image to be detected at the construction site where the front-end device is located and the management center where the back-end device is located respectively. Mark the target through a bounding box, issue an alarm for the detected violation behaviors and notify the safety management personnel, and at the same time generate a violation report including the violation content corresponding to the reported violation phenomenon, violation classification, violation level, and violated regulations.

4. The safety management method for hydropower station construction based on artificial intelligence according to claim 2, characterized in that: The violation detection model collects the construction project site pictures collected during the use process as new sample data for the dataset, confirms and annotates the new violation behaviors in the new sample data as incremental data for online learning, combines the incremental data with the data in the original training set, and continues to train the model until the violation detection model adapts to the new violation behaviors.

5. The safety management method for hydropower station construction based on artificial intelligence according to claim 2, wherein: Set an update period, summarize the violation behavior pictures misjudged and missed during an update period, establish an additional training dataset, and use the additional training dataset to retrain the model at the end of the update period.

6. The safety management method for hydropower station construction based on artificial intelligence according to claim 1, characterized in that: The acquisition method of the target image to be detected includes automatic recording by a camera device at the construction project site and active recording by a safety management personnel.

7. An artificial intelligence-based safety management system for hydropower station construction, characterized in that, Including: An acquisition module for acquiring a target image to be detected at the construction project site; A violation behavior detection module that analyzes and determines whether there is a violation behavior in the target image to be detected through a violation detection model; An alarm module for alarmingly prompting the violation behaviors detected by the violation behavior detection module; A storage module for storing the target image to be detected and the violation detection model.

8. An artificial intelligence-based safety management system for hydropower station construction according to claim 7, characterized in that: The violation behavior detection module includes: A violation detection model for detecting whether there is a violation behavior in the target image to be detected; A model optimization module for continuously optimizing the violation detection model during the usage period by introducing an online learning mechanism.

Images