Wharf cable releasing and mooring operation safety monitoring method based on video AI technology

Abstract

The invention relates to the technical field of safety monitoring methods, and discloses a wharf cable untying and mooring operation safety monitoring method based on a video AI technology, comprising the following steps: relying on an existing video monitoring platform, through wharf cable untying and mooring operation historical video data and on-site artificial simulation; and in combination with the latest artificial intelligence deep learning technology, various hidden danger characteristics in image videos and a disaster situation database are compared and analyzed, potential safety hazards generated in a video monitoring area are detected in time, and intelligent recognition of cable solving and tying operation in a petrochemical wharf area is achieved. According to the wharf cable untying operation safety monitoring method based on the video AI technology, the station detection function of wharf cable untying operation personnel is achieved, early discovery of safety risks, early warning of potential safety hazards, early prevention of field violation and early processing of problem processes are achieved, the phenomenon that safety precautionary measures are not implemented in place is eliminated in time, and the safety of wharf cable untying operation personnel is guaranteed. The problem that safety measures cannot be executed easily can be rapidly solved, and therefore the safety management capacity of a company is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of operation safety monitoring methods, and particularly to a safety monitoring method for mooring and unmooring operations at a wharf based on video AI technology. Background Art

[0002] Video AI technology is the application of artificial intelligence in the field of video processing and analysis. Combining technologies such as computer vision, deep learning, and natural language processing, it realizes the understanding, generation, editing, and optimization of video content;

[0003] Its core technologies include:

[0004] Computer vision (CV), which analyzes objects, scenes, and actions in a video through technologies such as object detection (YOLO, FasterR-CNN), image segmentation (MaskR-CNN), and action recognition (3DCNN).

[0005] Temporal modeling, using RNN, LSTM, or Transformer to handle the temporal continuity of a video, such as predicting action trajectories or analyzing the context logic of a long video;

[0006] Generative models, GAN and Diffusion models are used for video generation (such as Sora), style transfer, or restoration of old videos to achieve high-fidelity content creation;

[0007] Multimodal fusion, combining speech recognition (ASR), text (NLP), and visual data for cross-modal search or generating videos with subtitles;

[0008] In the safety accidents of chemical enterprises, accidents caused by untimely discovery of environmental safety hazards account for a relatively high proportion. Especially in the mooring and unmooring operations at wharf berths, accidents caused by personnel's illegal standing positions resulting in personal injuries occur from time to time, posing a serious threat to the safe production of chemical enterprises. Timely detection of violations in mooring and unmooring operations is an urgent problem for chemical enterprises. Therefore, there is an urgent need for a safety monitoring method for mooring and unmooring operations at a wharf based on video AI technology to solve the above technical problems. Summary of the Invention

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] A safety monitoring method for mooring and unmooring operations at a wharf based on video AI technology, the safety monitoring method for mooring and unmooring operations at the wharf includes the following steps:

[0011] Step 1: Rely on the existing video monitoring platform, through the historical video data of mooring and unmooring operations at the wharf and on-site artificial simulation;

[0012] Step 2: Combine the latest artificial intelligence deep learning technology to compare and analyze the various hidden danger features in the image video and the disaster database, timely detect the safety hazards in the video surveillance area, and realize the intelligent identification of mooring operations in the petrochemical terminal area;

[0013] Step 3: The algorithm identifies whether the operator is in a safe position when the cable is in a straight state in the mooring pier area, calculates the cross entropy loss and sums it up. The algorithm for the mooring pier operator position is as follows:

[0014]

[0015] Step 4: If any personnel are found to be operating in violation of regulations, an alarm will be sounded through the platform, and the on-site footage will be captured to remind the dispatch personnel to handle the situation on site in a timely manner.

[0016] Preferably, in the mooring pier personnel positioning algorithm, Y represents its output vector, and (t{ij}) is the value of the (j)th element in the true label vector of the (i)th sample (the value is (0) or (1)).

[0017] Preferably, in the mooring pier personnel positioning algorithm, p(y{ij} represents the predicted probability that the (i)th sample output is the (j)th class under given neural network parameters (bold symbol{theta}).

[0018] Preferably, in the mooring pier personnel positioning algorithm, L represents the total number of neuron layers.

[0019] Preferably, in the mooring pier personnel positioning algorithm, N represents the number of training samples, and a one-hot encoding form is used, that is, only one element is (1), and the rest are (0), and M represents the category to which it belongs.

[0020] Preferably, in the mooring pier personnel positioning algorithm, K represents the number of neurons in the output layer of the neural network.

[0021] Preferably, the logic of the terminal unpacking operation safety monitoring method based on video AI technology includes the following:

[0022] S1, update roi in real time from the target field;

[0023] S2. If the cable is continuously detected for a period of time, it is considered that the cable has been stretched straight, and human detection begins: first, human targets with a width or height less than 15 pixels are filtered out, and the midpoint below the human is determined to be in the ROI (if there is no ROI, all people are in the ROI). If the human is continuously in the ROI for a period of time, an alarm is triggered;

[0024] S3. If the cable is not detected for a period of time, it is considered that the cable has been retracted and the detection history is reset.

[0025] Preferably, if the target field transmits an ROI and the cable is continuously detected within a certain period of time, it is considered that the cable is straightened and the detection of people starts. First, filter out the person targets with a width or height less than 15 pixels. If the midpoint of the person is within the ROI for a continuous period of time, an alarm is issued. If the cable is not detected within a certain period of time, it is considered that the cable has been retracted and the historical record of person detection is reset.

[0026] Preferably, if the ROI is not transmitted in the target field, expand the target box A to the left, up, right, and down respectively, and expand the target box B to the left, up, right, and down respectively. If the cable is continuously detected within a certain period of time, it is considered that the cable is straightened and the detection of people starts. If the cable is not detected within a certain period of time, it is considered that the cable has been retracted and the historical record of person detection is reset.

[0027] Compared with the prior art, the present invention has the following beneficial effects: The present invention is a method for safety monitoring of dock mooring and unmooring operations based on video AI technology, which develops functions such as video reading, risk scenario modeling, and real-time intelligent recognition, analysis, and early warning, realizes the function of detecting the standing position of dock mooring and unmooring operators, realizes early discovery of safety risks, early alarm of potential safety hazards, early prevention of on-site violations, and early handling of problem processes, timely eliminates the phenomenon of ineffective implementation of safety prevention measures, can quickly solve the problem of ineffective implementation of safety measures, and thus enhances the company's safety management ability. This not only contributes to the long-term stable development of the company, but also is a key step in the transformation of Tianjin Port's dock safety management from traditional informatization to advanced intelligentization. Specific Embodiments

[0028] The following further elaborates on the present application in conjunction with embodiments. It can be understood that the specific embodiments described herein are only used to explain the related invention and do not limit the invention. Additionally, it should be noted that for the sake of description, only the parts related to the invention are shown. The different types of hatching lines in the embodiments of the present invention are not marked according to national standards, nor are there requirements for the materials of the components.

[0029] A method for safety monitoring of dock mooring and unmooring operations based on video AI technology, the method for safety monitoring of dock mooring and unmooring operations includes the following steps:

[0030] Step 1: Rely on the existing video monitoring platform through the historical video data of dock mooring and unmooring operations and on-site human simulation;

[0031] Step 2: Combine the latest artificial intelligence deep learning technology to compare and analyze various potential hazard features in the image video and the disaster database, timely detect the potential safety hazards generated in the video monitoring area, and realize the intelligent recognition of mooring and unmooring operations in the petrochemical terminal area;

[0032] Step 3: The algorithm calculates the cross-entropy loss and sums it up by identifying whether the operator is in a safe position when the cable is in a taut state in the mooring pier area. The mooring pier personnel standing position algorithm is as follows:

[0033]

[0034] Calculate its cross-entropy loss and sum it up:

[0035] Y represents its output vector, and (t{ij}) is the value of the j-th element in the true label vector of the i-th sample (taking values of 0 or 1);

[0036] p(y{ij} represents the predicted probability that the i-th sample outputs as the j-th class given the neural network parameters (boldsymbol{theta}).

[0037] L represents the total number of neuron layers;

[0038] N represents the number of training samples, using one-hot encoding, that is, only one element is 1 and the rest are 0, and M represents the category to which it belongs).

[0039] K represents the number of neurons in the output layer of the neural network;

[0040] Step 4: If it is found that there is a situation of personnel violating the operation regulations, the platform will give an alarm, capture the on-site video, and remind the dispatcher to conduct on-site disposal in a timely manner;

[0041] Among them, in the mooring pier personnel standing position algorithm, Y represents its output vector, and (t{ij}) is the value of the j-th element in the true label vector of the i-th sample (taking values of 0 or 1).

[0042] Among them, in the mooring pier personnel standing position algorithm, p(y{ij} represents the predicted probability that the i-th sample outputs as the j-th class given the neural network parameters (boldsymbol{theta}).

[0043] Among them, in the mooring pier personnel standing position algorithm, L represents the total number of neuron layers.

[0044] Among them, in the mooring pier personnel standing position algorithm, N represents the number of training samples, using one-hot encoding, that is, only one element is 1 and the rest are 0, and M represents the category to which it belongs.

[0045] Among them, in the mooring pier personnel standing position algorithm, K represents the number of neurons in the output layer of the neural network.

[0046] Among them, the logic of the safety monitoring method for the dock mooring and unmooring operation based on video AI technology includes the following:

[0047] S1. Update the ROI in real time from the target field;

[0048] S2. If the cable is continuously detected for a period of time, it is considered that the cable is straightened, and start to detect people: First, filter out the person targets with a width or height less than 15 pixels, and judge whether the midpoint below the person is in the ROI (if there is no ROI, all people are in the ROI). If the person is continuously in the ROI for a period of time, then alarm;

[0049] S3. If the cable is not detected for a period of time, it is considered that the cable has been retracted, and reset the historical record of detecting people.

[0050] Among them, if the target field passes the ROI, if the cable is continuously detected for a period of time, it is considered that the cable is straightened, and start to detect people. First, filter out the person targets with a width or height less than 15 pixels. If the midpoint of the person is in the ROI for a period of time, then alarm. If the cable is not detected for a period of time, it is considered that the cable has been retracted, and reset the historical record of detecting people.

[0051] Among them, if the ROI is not passed in the target field, expand the target box A to the left, up, right, and down respectively, and expand the target box B to the left, up, right, and down respectively. If the cable is continuously detected for a period of time, it is considered that the cable is straightened, and start to detect people. If the cable is not detected for a period of time, it is considered that the cable has been retracted, and reset the historical record of detecting people.

[0052] The working principle and usage process of the present invention: The algorithm for the standing position of the cable pier personnel

[0053] Algorithm support formula:

[0054]

[0055] Calculate its cross-entropy loss and sum:

[0056] Y represents its output vector, and (t{ij}) is the value of the j-th element in the true label vector of the i-th sample (taking values of (0) or (1));

[0057] p(y{ij} represents the predicted probability that the output of the i-th sample is the j-th class given the neural network parameters (boldsymbol{theta}).

[0058] Y represents its output vector, and (t{ij}) is the value of the j-th element in the true label vector of the i-th sample (taking values of (0) or (1));

[0059] p(y{ij} represents the predicted probability that the (i)th sample output is the (j)th class given the neural network parameters (boldsymbol{theta}).

[0060] L represents the total number of neuronal layers;

[0061] N represents the number of training samples, which are encoded in one-hot format, that is, only one element is (1) and the rest are (0), and M represents the category).

[0062] K represents the number of neurons in the output layer of the neural network;

[0063] The terminal unwinding and hauling operation safety monitoring method comprises the following steps:

[0064] Step 1: Relying on the existing video surveillance platform, using historical video data of the dock mooring operation and on-site manual simulation;

[0065] Step 2: Combine the latest artificial intelligence deep learning technology to compare and analyze the various hidden danger features in the image video and the disaster database, timely detect the safety hazards in the video surveillance area, and realize the intelligent identification of mooring operations in the petrochemical terminal area;

[0066] Step 3: The algorithm identifies whether the operator is in a safe position when the cable is in a straight state in the mooring pier area, calculates the cross entropy loss and sums it;

[0067] Step 4: If any personnel are found to be operating in violation of regulations, an alarm will be raised through the platform, and the on-site images will be captured to remind the dispatch personnel to handle the situation on site in a timely manner;

[0068] The present invention is based on the video AI technology of the terminal untethering operation safety monitoring method, develops video reading, risk scenario modeling, and real-time intelligent identification, analysis, early warning and other functions, realizes the terminal untethering operation personnel position detection function, realizes early detection of safety risks, early alarm of safety hazards, early prevention of on-site violations, early processing of problem processes, and timely eliminates the phenomenon of inadequate implementation of safety precautions, and can quickly solve the problem of poor implementation of safety measures, thereby enhancing the company's safety management capabilities. This will not only contribute to the long-term and stable development of the company, but also is a key step in the transformation of Tianjin Port terminal safety management from traditional information technology to advanced intelligence.

[0069] Example 1

[0070] The logic of the terminal unpacking and handling operation safety monitoring method based on video AI technology includes the following:

[0071] S1, update roi in real time from the target field;

[0072] S2. If the cable is continuously detected for a period of time, it is considered that the cable is straightened, and the detection of people begins: First, filter out the person targets with a width or height less than 15 pixels, and determine whether the midpoint below the person is in the ROI (if there is no ROI, all people are in the ROI). If the person is continuously in the ROI for a period of time, an alarm is issued;

[0073] S3. If the cable is not detected for a period of time, it is considered that the cable has been retracted, and the historical record of person detection is reset.

[0074] Among them, if the ROI is passed in the target field, if the cable is continuously detected for a period of time, it is considered that the cable is straightened, and the detection of people begins. First, filter out the person targets with a width or height less than 15 pixels. If the midpoint of the person is in the ROI for a period of time, an alarm is issued. If the cable is not detected for a period of time, it is considered that the cable has been retracted, and the historical record of person detection is reset.

[0075] Among them, if the ROI is not passed in the target field, expand the target box A outward to the left, up, right, and down respectively, and expand the target box B outward to the left, up, right, and down respectively. If the cable is continuously detected for a period of time, it is considered that the cable is straightened, and the detection of people begins. If the cable is not detected for a period of time, it is considered that the cable has been retracted, and the historical record of person detection is reset.

[0076] The content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0077] The above description is only the preferred embodiment of the present application and the explanation of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by the specific combination of the above technical features, but also covers other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the present application.

Claims

1. A safety monitoring method for the mooring and unmooring operations at the dock based on video AI technology, characterized in that: The safety monitoring method for mooring line releasing and tying operations at the wharf includes the following steps: Step 1: Rely on the existing video monitoring platform and conduct simulations based on the historical video data of mooring line releasing and tying operations at the wharf and on-site human simulations; Step 2: Combine the latest artificial intelligence deep learning technology to compare and analyze various potential hazard features in the image videos and the disaster situation database, detect the safety hazards generated in the video monitoring area in a timely manner, and achieve intelligent identification of mooring line releasing and tying operations in the petrochemical wharf area; Step 3: The algorithm calculates the cross-entropy loss and sums it by identifying whether the operator is in a safe position when the mooring line is in a taut state in the mooring dolphin area. The algorithm for the personnel standing position in the mooring dolphin is as follows: Step 4: If it is found that there is a situation of personnel violating the operation regulations, an alarm will be issued through the platform, and the on-site picture will be captured to remind the dispatcher to conduct on-site disposal in a timely manner.

2. The method for safety monitoring of mooring and unmooring operations at a dock based on video AI technology according to claim 1, wherein: In the algorithm for the personnel standing position in the mooring dolphin, Y represents its output vector, and (t{ij}) is the value of the j-th element in the true label vector of the i-th sample (taking values of (0) or (1)).

3. The safety monitoring method for dock mooring and unmooring operations based on video AI technology according to claim 1, characterized in that: In the algorithm for the personnel standing position in the mooring dolphin, p(y{ij} represents the predicted probability that the output of the i-th sample is the j-th class given the neural network parameters (boldsymbol{theta}).

4. The safety monitoring method for dock mooring and unmooring operations based on video AI technology according to claim 1, characterized in that: In the algorithm for the personnel standing position in the mooring dolphin, L represents the total number of neuron layers.

5. The safety monitoring method for dock mooring and unmooring operations based on video AI technology according to claim 1, characterized in that: In the algorithm for the personnel standing position in the mooring dolphin, N represents the number of training samples, which adopts the one-hot encoding form, that is, only one element is (1) and the rest are (0), and M represents the category to which it belongs.

6. The safety monitoring method for mooring and unmooring operations at a dock based on video AI technology according to claim 1, characterized in that: In the algorithm for the personnel standing position in the mooring dolphin, K represents the number of neurons in the output layer of the neural network.

7. The safety monitoring method for dock mooring and unmooring operations based on video AI technology according to claim 1, characterized in that: The logic of the safety monitoring method for mooring line releasing and tying operations based on video AI technology includes the following: S1: Update the roi in real time from the target field; S2: If the mooring line is continuously detected for a period of time, it is considered that the mooring line is taut, and the detection of people begins: First, filter out the human targets with a width or height less than 15 pixels, and judge whether the midpoint below the person is in the roi (if there is no roi, then all people are in the roi). If the person continues to be in the roi for a period of time, an alarm will be issued; S3: If the mooring line is not detected for a period of time, it is considered that the mooring line has been retracted, and the historical record of detecting people is reset.

8. The safety monitoring method for mooring and unmooring operations at a dock based on video AI technology according to claim 1, characterized in that: If the roi is passed in the target field, if the mooring line is continuously detected for a period of time, it is considered that the mooring line is taut, and the detection of people begins. First, filter out the human targets with a width or height less than 15 pixels. If the midpoint of the person is in the roi for a period of time, an alarm will be issued. If the mooring line is not detected for a period of time, it is considered that the mooring line has been retracted, and the historical record of detecting people is reset.

9. The method for monitoring the safety of mooring and unmooring operations at a dock based on video AI technology according to claim 1, wherein: If the roi is not passed in the target field, expand the target box A to the left, up, right, and down respectively, and expand the target box B to the left, up, right, and down respectively. If the mooring line is continuously detected for a period of time, it is considered that the mooring line is taut, and the detection of people begins. If the mooring line is not detected for a period of time, it is considered that the mooring line has been retracted, and the historical record of detecting people is reset.

Images