Image monitoring-based ferry in-cabin inspection system and abnormity positioning method

Abstract

The invention discloses a ferry cabin inspection system based on image monitoring and an abnormity positioning method, and relates to the technical field of inspection positioning, and the method comprises the steps: obtaining a historical inspection log in a ferry cabin; constructing a causal knowledge graph representing causal relevance among various events in the cabin; collecting an inspection image sequence in the cabin in real time; identifying a real-time event set currently occurring in the cabin; calculating the probability of occurrence of the other events in a specified time window in the future; taking abnormal events in future events of which the occurrence probability exceeds a set threshold value in a specified time window in the future and the occurrence probability as potential abnormal risk events to be output; and based on the output potential abnormal risk event and the corresponding causal association path, positioning a potential source position or an association region where the abnormality occurs. The system has the advantages that all-directional automatic coverage of a cabin core area is realized, the dependence on manual inspection is effectively reduced, and the active protection capability of ferry operation safety is comprehensively enhanced.

Description

Technical Field

[0001] The present invention relates to the field of patrol inspection and positioning technology, and in particular to an in-cabin patrol inspection system and an abnormality positioning method for a ferry based on image monitoring. Background Art

[0002] Current safety monitoring within ferry cabins relies primarily on fixed sensors, such as smoke alarms and temperature probes, and manual scheduled inspections. Sensors have a limited monitoring range and can only trigger alarms based on preset physical thresholds, making them incapable of identifying complex events. Manual inspections, on the other hand, suffer from low timeliness, high subjectivity, and insufficient coverage at night and in harsh environments. Both methods struggle to capture the dynamic state of the cabin in real time. This is especially true during emergencies like fires and mechanical failures. Existing systems can only respond passively after anomalies become apparent, failing to predict the evolution of the risk chain.

[0003] In recent years, computer vision-based cabin monitoring systems have begun to be used. However, these systems focus on detecting single-point anomalies, such as identifying flames or falls through image recognition. These systems fail to establish a causal relationship between historical and real-time events, nor do they proactively predict future risks. Due to the densely packed equipment and complex environment within ferry cabins, a single abnormal event can trigger secondary hazards through factors such as equipment linkage and spatial proximity, such as a localized leak leading to an electrical short circuit. Existing technologies, ignoring the causal propagation mechanism between events, fail to pinpoint the source of potential risks, making it even more difficult to upgrade from passive warnings to active protection. Summary of the Invention

[0004] To solve the above technical problems, a ferry cabin inspection system and anomaly positioning method based on image monitoring are provided. This technical solution solves the problem that the above-mentioned existing system can only respond passively after the anomaly becomes obvious and cannot predict the evolution process of the risk chain in advance.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0006] A method for locating abnormalities in a ferry cabin based on image monitoring, comprising:

[0007] Obtaining a historical inspection log in the ferry cabin, wherein the historical inspection log includes various historical events that occurred in the cabin and their corresponding occurrence timestamps;

[0008] Based on the historical inspection logs, the causal correlation strength and / or temporal probability between different historical events are calculated to construct a causal knowledge graph that characterizes the causal correlation between various events in the cabin;

[0009] Multiple image acquisition devices deployed in key areas of the ferry cabin collect inspection image sequences in real time;

[0010] Using an image recognition model to process the inspection image sequence collected in real time, identifying a set of real-time events currently occurring in the cabin, the real-time event set including possible event types and comprehensive probabilities of occurrence of event types;

[0011] Calculate the probability of occurrence of other events in the causal knowledge graph that are causally related to the current real-time event, as well as the corresponding causal association strength and / or temporal probability, within a specified future time window;

[0012] Output the abnormal time and probability of occurrence of future events with a probability exceeding the set threshold within a specified time window in the future as potential events;

[0013] Determine whether the potential event is of the defined abnormal event type. If so, mark the potential event as a potential abnormal risk event. If not, do not respond to the potential event output.

[0014] Based on the output potential abnormal risk events and their corresponding causal association paths, the potential source location or associated area where the abnormality occurs is located.

[0015] Preferably, the construction of a causal knowledge graph representing the causal relationships between various events in the cabin specifically includes:

[0016] Preprocess and define events in historical inspection logs;

[0017] Based on the timestamps of events, statistically analyze the time interval distribution and frequency of event A occurring before event B in the event sequence;

[0018] Calculate the causal weight of event A as the cause of event B based on the set time window and significance principle;

[0019] A directed weighted graph is constructed with events as nodes and causal relationships represented by causal weights that meet preset conditions as edges to form the causal knowledge graph.

[0020] Preferably, the image recognition model is used to process the inspection image sequence collected in real time to identify a set of real-time events currently occurring in the cabin. The real-time event set includes possible event types and comprehensive probabilities of occurrence of event types, specifically including:

[0021] Based on the trained image recognition model, event recognition is performed on each inspection image in the inspection image sequence to obtain the possible event type and the probability of occurrence of each event type corresponding to each inspection image;

[0022] Based on the recognition results of all inspection images in the inspection image sequence, determine the comprehensive probability of each possible event type;

[0023] Summarize all possible event types and the comprehensive probability of event types occurring to build a real-time event set;

[0024] Specifically, the calculation process is as follows:

[0025]

[0026] Where, is the probability of the Xth possible event type occurring, is the inspection image sequence, is the e-th element in the inspection image sequence, is the probability of occurrence of the Xth possible event type corresponding to the eth element in the inspection image sequence.

[0027] Preferably, the calculating the probability of occurrence of other events within a specified future time window based on other events causally associated with the current real-time event in the causal knowledge graph and the corresponding causal association strength and / or temporal probability specifically includes:

[0028] Perform event chain reasoning based on the current real-time event set and the causal knowledge graph, and calculate the probability of occurrence of subsequent node events on the event chain within the future time window. The probability of occurrence is the combination of the causal weights of each link on the path;

[0029] The specific calculation formula is:

[0030]

[0031] Where, is the probability of the Yth subsequent node event occurring in the future time window, is the probability of occurrence of the Lth possible event type, is the event chain source set corresponding to the Yth subsequent node event, is the event chain from event L to event Y, For the adjacent node events in , for The causal weight between them.

[0032] Preferably, the locating of the potential source location or associated area of the abnormality based on the output potential abnormal risk event and its corresponding causal association path specifically includes:

[0033] Based on the source events of all event chains corresponding to potential abnormal risk events, analyze the potential risk contribution value of each real-time event to the potential abnormal risk event;

[0034] Determine the associated area corresponding to each real-time event;

[0035] Accumulate the potential risk contribution values of all real-time events corresponding to the associated area to the potential abnormal risk items to obtain the potential risk value of the associated area;

[0036] Output the associated areas in order from largest to smallest.

[0037] Furthermore, a ferry cabin inspection system based on image monitoring is proposed, which is used to implement the above-mentioned ferry cabin abnormality positioning method based on image monitoring, including:

[0038] Image acquisition units are deployed in key areas of the ferry cabin to collect real-time inspection image sequences inside the cabin;

[0039] A historical log storage unit for storing the ferry cabin inspection log including historical event types and their timestamps;

[0040] A causal graph construction unit, connected to the historical log storage unit, is used to calculate the causal association strength between events based on the temporal relationship of historical events, and generate a weighted causal knowledge graph with events as nodes and causal relationships as edges;

[0041] A real-time event recognition unit, connected to the image acquisition unit and configured with an image recognition model, for identifying the current event type in the image sequence and outputting a real-time event set with probability weights;

[0042] An event prediction unit, connected to the causal graph construction unit and the real-time event identification unit, is used to deduce the probability of other events occurring within a specified future time window based on the association path between the real-time event set and the causal knowledge graph;

[0043] an anomaly determination unit, connected to the event prediction unit, for screening future events whose probability of occurrence exceeds a preset threshold, and marking potential abnormal risk events based on a predefined anomaly type library;

[0044] An anomaly locating unit, connected to the anomaly determining unit, is used to trace the associated areas based on the causal association path of potential abnormal risk events and output a ranked list of risk areas;

[0045] The alarm output unit is used to publish alarm information containing potential abnormal risk events and risk areas.

[0046] Optionally, the causal graph construction unit includes:

[0047] The event pre-processing sub-unit is used to classify the event types in the historical logs;

[0048] The causal analysis subunit is used to count the temporal frequency of events and calculate the causal weight.

[0049] Optionally, the real-time event recognition unit includes:

[0050] A separate identification subunit is used to perform separate event identification on each inspection image in the inspection image sequence to obtain a possible event type and a probability of occurrence of the event type corresponding to each inspection image;

[0051] The event probability fusion subunit is used to integrate the recognition results of multiple frames of images to generate a comprehensive probability value of the event type.

[0052] Optionally, the abnormality locating unit includes:

[0053] The source contribution calculation subunit is used to quantify the contribution of the source events in the causal chain to the risk;

[0054] The area mapping subunit associates events with pre-divided physical areas of the cabin;

[0055] The regional risk ranking subunit aggregates risk contribution values by region and outputs them in descending order.

[0056] Optionally, the image acquisition unit includes:

[0057] A multi-view camera group with time and space tags is deployed on the cabin roof and equipment nodes;

[0058] Image synchronization component, which adds a location tag and timestamp to each frame of image.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] The present invention revolutionizes the safety monitoring mechanism inside the ferry cabin by constructing a causal knowledge graph of cabin events and integrating real-time image sequence analysis. It not only identifies various dynamic abnormal states in real time, but also predicts the evolution trend of potential risk chains through a causal reasoning model between historical events and real-time events, enabling the system to proactively warn before secondary disasters manifest, significantly advancing the risk handling window period. At the same time, based on the reverse tracing capability of the event causal chain, it accurately locates the physical source of the hidden danger, completely solving the blind spot problem of traditional monitoring that is difficult to trace the root cause of abnormalities, and significantly improving the efficiency of accident investigation. In addition, the collaborative perception of multi-dimensional states such as equipment operation abnormalities, environmental state changes, and personnel behavior risks realizes all-round automated coverage of the core area of the cabin, effectively reducing dependence on manual inspections and comprehensively enhancing the active protection capability of ferry operation safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a flow chart of the method for locating abnormalities in a ferry cabin based on image monitoring proposed by the present invention;

[0062] Figure 2This is a flow chart of the method for constructing a causal knowledge graph proposed in the present invention;

[0063] Figure 3 This is a flow chart of the method proposed by the present invention for identifying a set of real-time events currently occurring in a cabin;

[0064] Figure 4 This is a flow chart of the method proposed by the present invention for locating potential source locations or associated areas where anomalies occur. DETAILED DESCRIPTION

[0065] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0066] Reference Figure 1 As shown, a method for locating abnormalities in a ferry cabin based on image monitoring includes:

[0067] Obtain the historical inspection log of the ferry cabin, which contains various historical events that occurred in the cabin and their corresponding occurrence timestamps;

[0068] Based on long-term accumulated historical event data, we deeply explore the operating patterns within the cabin, providing a reliable data foundation for building causal association models and significantly improving the accuracy and reliability of subsequent risk predictions.

[0069] Based on historical inspection logs, the causal relationship strength and / or temporal probability between different historical events are calculated to construct a causal knowledge graph that characterizes the causal relationships between various events in the cabin.

[0070] By quantifying the causal relationships between events, discrete historical events are transformed into structured knowledge networks, breaking through the limitations of traditional monitoring methods that rely solely on single event responses and laying the core logical framework for accurate deduction of real-time risk evolution.

[0071] Multiple image acquisition devices deployed in key areas of the ferry cabin collect inspection image sequences in real time;

[0072] Multi-view, high-density coverage of core areas captures dynamic environmental changes in real time, overcomes sensor monitoring blind spots and manual inspection lags, and achieves continuous perception of the entire cabin's operating status.

[0073] The image recognition model is used to process the real-time inspection image sequence to identify the real-time event set currently occurring in the cabin. The real-time event set includes the possible event types and the comprehensive probability of the event type occurring.

[0074] Fusion of multi-frame image recognition results generates probabilistic event conclusions, significantly reducing false alarm rates, such as distinguishing normal water mist from leaking steam, while enhancing recognition robustness in complex scenes.

[0075] Calculate the probability of other events occurring within a specified future time window based on other events in the causal knowledge graph that are causally related to the current real-time event, as well as the corresponding causal association strength and / or temporal probability;

[0076] Combining real-time dynamics with historical causal chains, we proactively deduce the evolution path of secondary disasters, such as leakage → short circuit → fire, to achieve a fundamental shift from post-event warning to pre-event prediction, significantly extending the risk management window.

[0077] Output the abnormal time and probability of occurrence of future events with a probability exceeding the set threshold within a specified time window in the future as potential events;

[0078] Determine whether the potential event is of the defined abnormal event type. If so, mark the potential event as a potential abnormal risk event. If not, do not respond to the potential event output.

[0079] Filter low-risk events based on a preset anomaly type library and focus on real high-risk targets, such as electrical fires and structural collapses, to avoid invalid alarm interference and improve system decision-making efficiency and credibility;

[0080] Based on the output potential abnormal risk events and their corresponding causal association paths, the potential source location or associated area where the abnormality occurs is located.

[0081] Trace the risk transmission chain back to the source in physical space, such as locating leakage points or faulty equipment, completely solving the pain point of traditional systems where "fire is seen but not the source" and guiding accurate and rapid disposal to minimize accident losses.

[0082] Reference Figure 2 As shown in the figure, constructing a causal knowledge graph that represents the causal relationships between various events in the cabin specifically includes:

[0083] Preprocess and define events in historical inspection logs;

[0084] Based on the timestamps of events, statistically analyze the time interval distribution and frequency of event A occurring before event B in the event sequence;

[0085] The causal weight of event A as the cause of event B is calculated based on the set time window and significance principle. Specifically, in some preferred embodiments, the causal weight of event A as the cause of event B can be calculated by comparing the number of times event B occurs within the time window after event A occurs with the sum of the number of times all events occur within the time window after event A occurs.

[0086] With events as nodes and causal relationships represented by causal weights that meet preset conditions as edges, a directed weighted graph is constructed to form a causal knowledge graph.

[0087] By quantifying the temporal dependencies between historical events and constructing a causal knowledge graph, three core breakthroughs were achieved: First, the proportion of causal associations within the event window was used as the weight basis. For example, the frequency of event B within a specific time period after event A occurred was the percentage. This objectively separated the accidental temporal relationship from a statistical perspective, significantly improving the interpretability and accuracy of causal judgments, and avoiding the bias of traditional expert systems that rely on subjective experience rules. Second, the constructed dynamic directed weighted graph can adaptively integrate newly added historical data, giving the knowledge graph the ability to continuously evolve and continuously optimizing the modeling accuracy of the risk propagation laws of complex systems in the cabin. Finally, the graph provides a physical and logically complete reasoning framework for the risk deduction of subsequent real-time events, ensuring that the prediction results of key causal chains such as "equipment leakage" and "electrical short circuit" have clear engineering traceability, fundamentally resolving the trust crisis in operation and maintenance decisions caused by black-box prediction models, and laying a technical foundation for the engineering implementation of active protection systems.

[0088] Reference Figure 3 As shown in the figure, the image recognition model is used to process the real-time inspection image sequence to identify the real-time event set currently occurring in the cabin. The real-time event set includes the possible event types and the comprehensive probability of the event types occurring. Specifically, it includes:

[0089] Based on the trained image recognition model, event recognition is performed on each inspection image in the inspection image sequence to obtain the possible event type and the probability of occurrence of the event type corresponding to each inspection image. The image recognition model can be trained using a convolutional neural network. The softmax layer is used to convert the probability of occurrence of the event recognized by the image recognition model into a probability value between 0 and 1. Convolutional neural networks are a mature technology in this field. The training process of the image recognition model does not involve the core improvement of this solution, so it will not be described in detail here.

[0090] Based on the recognition results of all inspection images in the inspection image sequence, determine the comprehensive probability of each possible event type;

[0091] Summarize all possible event types and the comprehensive probability of event types occurring to build a real-time event set;

[0092] Specifically, the calculation process is as follows:

[0093]

[0094] Where, is the probability of the Xth possible event type occurring, is the inspection image sequence, is the e-th element in the inspection image sequence, is the probability of occurrence of the Xth possible event type corresponding to the eth element in the inspection image sequence.

[0095] Through the probabilistic fusion mechanism of multiple frames of images, the robustness and confidence of event recognition in complex cabin environments are significantly improved: First, comprehensive probability calculations are performed based on the recognition results of sequential images, such as capturing the continuous state changes of equipment leakage from multiple angles, effectively eliminating misjudgments of single-frame images due to perspective occlusion, light interference, or instantaneous movements, and improving the confidence level of key event judgments to an operational level; second, the probabilistic fusion model can adapt to complex interference environments such as dynamic vibration, sudden light changes, and equipment occlusion in ferry operations, ensuring the output of high-reliability event sets in mobile scenarios; finally, the generated probabilistic event description provides quantitative input for subsequent causal chain deduction, making "78% probability of event occurrence" more accurate in supporting risk grading decisions than the binary judgment of "the existence of an event", fundamentally solving the problem of false alarms in the application of image monitoring technology in industrial scenarios, and providing a highly reliable original decision-making basis for predictive protection systems.

[0096] Based on other events in the causal knowledge graph that are causally related to the current real-time event, as well as the corresponding causal relationship strength and / or temporal probability, the probability of other events occurring within a specified future time window is calculated. Specifically, the following steps are involved:

[0097] Based on the current real-time event set and causal knowledge graph, event chain reasoning is performed to calculate the probability of subsequent node events on the event chain within the future time window. The probability of occurrence is the combination of the causal weights of each link on the path;

[0098] The specific calculation formula is:

[0099]

[0100] Where, is the probability of the Yth subsequent node event occurring in the future time window, is the probability of occurrence of the Lth possible event type, is the event chain source set corresponding to the Yth subsequent node event, is the event chain from event L to event Y, For the adjacent node events in , for The causal weight between them.

[0101] Through the multiplication and aggregation mechanism of event chain causal weights, the quantitative prediction of complex cascade risks has been achieved for the first time in the field of ship monitoring: based on the dynamic coupling of real-time event probability and historical causal weights, the evolution path of multi-level secondary disasters is accurately captured, such as the transmission chain of equipment leakage → electrical short circuit → cabin smoke, breaking through the technical bottleneck that traditional models can only judge direct causal relationships; at the same time, the mathematical essence of probability product is used to enhance the perception sensitivity of long-term, weakly correlated risks. For example, the probability of an initial weak leakage triggering a fire jumps after three levels of transmission, enabling the system to warn of systemic risks that cannot be identified by traditional means at the budding stage of hidden dangers; finally, the output event chain probability value provides a quantitative basis for operation and maintenance decision-making, such as giving priority to handling "leakage-short circuit" combination events with weighted probabilities exceeding the threshold, avoiding the mismatch of protective resources caused by the misjudgment of discrete events from the root, and greatly improving the active prevention and control accuracy of high-risk scenarios.

[0102] Reference Figure 4 As shown, based on the output potential abnormal risk events and their corresponding causal association paths, the potential source location or associated area where the abnormality occurs is located, specifically including:

[0103] Based on the source events of all event chains corresponding to potential abnormal risk events, analyze the potential risk contribution value of each real-time event to the potential abnormal risk event;

[0104] Determine the associated area corresponding to each real-time event;

[0105] Accumulate the potential risk contribution values of all real-time events corresponding to the associated area to the potential abnormal risk items to obtain the potential risk value of the associated area;

[0106] Output the associated areas in order from largest to smallest.

[0107] Specifically, the potential risk contribution of each real-time event to the potential abnormal risk event is The potential risk contribution value reflects the risk probability guidance value of the real-time event with the real-time event as the source ultimately leading to the potential abnormal risk event. The larger the value, the greater the causal guidance relationship between the real-time event and the occurrence of the potential abnormal risk event.

[0108] Through the regional accumulation and sorting mechanism of risk contribution values, a leap forward from abstract event warning to precise positioning in physical space is achieved: First, based on the causal chain, the contribution intensity of each source event to the terminal risk is calculated by reverse tracing, such as the weighted guiding value of oil pump leakage to electrical fire, and minor related events are separated in a quantitative manner to focus on the core risk source. For example, the leakage source that causes a short circuit is accurately distinguished from the unrelated pipeline condensate; secondly, by mapping the regional risk value accumulation to the physical space, the industry has achieved the first cognitive upgrade from "where the abnormality occurs" to "where the source of the hidden danger is", completely solving the pain point that traditional monitoring alarms still require hours of manual investigation; finally, the regional sequence is output in descending order of risk value, which directly guides the efficient deployment of operation and maintenance forces, improves the efficiency of maintenance response, and transforms the time limit for handling accident hazards from "post-event rescue" to "source elimination".

[0109] Furthermore, based on the above-mentioned method for locating abnormalities in a ferry cabin based on image monitoring, this solution also proposes a ferry cabin inspection system based on image monitoring, including:

[0110] Image acquisition units are deployed in key areas of the ferry cabin to collect real-time inspection image sequences inside the cabin;

[0111] The image acquisition unit includes a multi-view camera group with time and space tags, deployed on the cabin roof and equipment nodes, and an image synchronization component to add a location tag and time stamp to each frame of the image;

[0112] A historical log storage unit for storing the ferry cabin inspection log including historical event types and their timestamps;

[0113] The causal graph construction unit is connected to the historical log storage unit and is used to calculate the causal relationship strength between events based on the temporal relationship of historical events, and generate a weighted causal knowledge graph with events as nodes and causal relationships as edges;

[0114] The causal graph construction units include:

[0115] The event pre-processing sub-unit is used to classify the event types in the historical logs;

[0116] The causal analysis subunit is used to count the temporal frequency of events and calculate the causal weight;

[0117] A real-time event recognition unit, connected to the image acquisition unit and configured with an image recognition model, is used to identify the current event type in the image sequence and output a real-time event set with probability weights;

[0118] Real-time event recognition unit:

[0119] The individual recognition subunit is used to perform individual event recognition on each inspection image in the inspection image sequence, and obtain the possible event type and the probability of occurrence of the event type corresponding to each inspection image;

[0120] The event probability fusion subunit is used to integrate the recognition results of multiple frames of images to generate a comprehensive probability value of the event type;

[0121] The event prediction unit connects the causal graph construction unit and the real-time event recognition unit, and is used to deduce the probability of other events occurring within a specified future time window based on the association path between the real-time event set and the causal knowledge graph;

[0122] The anomaly determination unit, connected to the event prediction unit, is used to screen future events whose probability of occurrence exceeds a preset threshold and mark potential abnormal risk events based on a predefined anomaly type library;

[0123] The anomaly location unit is connected to the anomaly determination unit and is used to trace the associated areas based on the causal association path of potential abnormal risk events and output a ranked list of risk areas;

[0124] The abnormality location unit includes:

[0125] The source contribution calculation subunit is used to quantify the contribution of the source events in the causal chain to the risk;

[0126] The area mapping subunit associates events with pre-divided physical areas of the cabin;

[0127] The regional risk ranking subunit aggregates risk contribution values by region and outputs them in descending order;

[0128] The alarm output unit is used to publish alarm information containing potential abnormal risk events and risk areas.

[0129] In summary, the advantages of the present invention are: by constructing a causal knowledge graph of cabin events and integrating real-time image sequence analysis, it innovates the safety monitoring mechanism in the ferry cabin, not only identifying various dynamic abnormal states in real time, but also predicting the evolution trend of the potential risk chain through the causal reasoning model between historical events and real-time events, so that the system can actively warn before secondary disasters manifest, and significantly advance the risk disposal window period; at the same time, based on the reverse tracing capability of the event causal chain, it accurately locates the source of the physical area where the hidden danger occurs, completely solving the blind spot problem of traditional monitoring that is difficult to trace the root cause of the abnormality, and significantly improving the efficiency of accident investigation; coupled with the collaborative perception of multi-dimensional states such as equipment operation abnormalities, environmental state changes, and personnel behavior risks, it realizes all-round automated coverage of the core area of the cabin, effectively reduces dependence on manual inspections, and comprehensively enhances the active protection capability of ferry operation safety.

[0130] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for locating abnormalities in a ferry cabin based on image monitoring, characterized in that: include: Obtaining a historical inspection log in the ferry cabin, wherein the historical inspection log includes various historical events that occurred in the cabin and their corresponding occurrence timestamps; Based on the historical inspection logs, the causal correlation strength and / or temporal probability between different historical events are calculated to construct a causal knowledge graph that characterizes the causal correlation between various events in the cabin; Multiple image acquisition devices deployed in key areas of the ferry cabin collect inspection image sequences in real time; Using an image recognition model to process the inspection image sequence collected in real time, identifying a set of real-time events currently occurring in the cabin, the real-time event set including possible event types and comprehensive probabilities of occurrence of event types; Calculate the probability of occurrence of other events in the causal knowledge graph that are causally related to the current real-time event, as well as the corresponding causal association strength and / or temporal probability, within a specified future time window; Output the abnormal time and probability of occurrence of future events with a probability exceeding the set threshold within a specified time window in the future as potential events; Determine whether the potential event is of the defined abnormal event type. If so, mark the potential event as a potential abnormal risk event. If not, do not respond to the potential event output. Based on the output potential abnormal risk events and their corresponding causal association paths, the potential source location or associated area where the abnormality occurs is located.

2. The method for locating abnormalities in a ferry cabin based on image monitoring according to claim 1, characterized in that: The construction of a causal knowledge graph representing the causal relationships between various events in the cabin specifically includes: Preprocess and define events in historical inspection logs; Based on the timestamps of events, statistically analyze the time interval distribution and frequency of event A occurring before event B in the event sequence; Calculate the causal weight of event A as the cause of event B based on the set time window and significance principle; A directed weighted graph is constructed with events as nodes and causal relationships represented by causal weights that meet preset conditions as edges to form the causal knowledge graph.

3. The method for locating abnormalities in a ferry cabin based on image monitoring according to claim 2, characterized in that: The image recognition model is used to process the inspection image sequence collected in real time to identify a set of real-time events currently occurring in the cabin. The real-time event set includes possible event types and the comprehensive probability of occurrence of event types, specifically including: Based on the trained image recognition model, event recognition is performed on each inspection image in the inspection image sequence to obtain the possible event type and the probability of occurrence of each event type corresponding to each inspection image; Based on the recognition results of all inspection images in the inspection image sequence, determine the comprehensive probability of each possible event type; Summarize all possible event types and the comprehensive probability of event types occurring to build a real-time event set; Specifically, the calculation process is as follows: ； Where, is the probability of the Xth possible event type occurring, is the inspection image sequence, is the e-th element in the inspection image sequence, is the probability of occurrence of the Xth possible event type corresponding to the eth element in the inspection image sequence.

4. The method for locating abnormalities in a ferry cabin based on image monitoring according to claim 3 is characterized in that: Calculating the probability of occurrence of other events in a future specified time window based on other events causally associated with the current real-time event in the causal knowledge graph and the corresponding causal association strength and / or temporal probability specifically includes: Perform event chain reasoning based on the current real-time event set and the causal knowledge graph, and calculate the probability of occurrence of subsequent node events on the event chain within the future time window. The probability of occurrence is the combination of the causal weights of each link on the path; The specific calculation formula is: ； Where, is the probability of the Yth subsequent node event occurring in the future time window, is the probability of occurrence of the Lth possible event type, is the event chain source set corresponding to the Yth subsequent node event, is the event chain from event L to event Y, For the adjacent node events in , for The causal weight between them.

5. The method for locating abnormalities in a ferry cabin based on image monitoring according to claim 4 is characterized in that: The method of locating the potential source location or associated area of the abnormality based on the output potential abnormal risk event and its corresponding causal association path specifically includes: Based on the source events of all event chains corresponding to potential abnormal risk events, analyze the potential risk contribution value of each real-time event to the potential abnormal risk event; Determine the associated area corresponding to each real-time event; Accumulate the potential risk contribution values of all real-time events corresponding to the associated area to the potential abnormal risk items to obtain the potential risk value of the associated area; Output the associated areas in order from largest to smallest.

6. A ferry cabin inspection system based on image monitoring, characterized in that: The method for locating anomalies in a ferry cabin based on image monitoring according to any one of claims 1 to 5 comprises: Image acquisition units are deployed in key areas of the ferry cabin to collect real-time inspection image sequences inside the cabin; A historical log storage unit for storing the ferry cabin inspection log including historical event types and their timestamps; A causal graph construction unit, connected to the historical log storage unit, is used to calculate the causal association strength between events based on the temporal relationship of historical events, and generate a weighted causal knowledge graph with events as nodes and causal relationships as edges; A real-time event recognition unit, connected to the image acquisition unit and configured with an image recognition model, for identifying the current event type in the image sequence and outputting a real-time event set with probability weights; An event prediction unit, connected to the causal graph construction unit and the real-time event identification unit, is used to deduce the probability of other events occurring within a specified future time window based on the association path between the real-time event set and the causal knowledge graph; an anomaly determination unit, connected to the event prediction unit, for screening future events whose probability of occurrence exceeds a preset threshold, and marking potential abnormal risk events based on a predefined anomaly type library; An anomaly locating unit, connected to the anomaly determining unit, is used to trace the associated areas based on the causal association path of potential abnormal risk events and output a ranked list of risk areas; The alarm output unit is used to publish alarm information containing potential abnormal risk events and risk areas.

7. The image monitoring-based ferry cabin inspection system according to claim 6, characterized in that: The causal graph construction unit includes: The event pre-processing sub-unit is used to classify the event types in the historical logs; The causal analysis subunit is used to count the temporal frequency of events and calculate the causal weight.

8. The image monitoring-based inspection system for ferry cabins according to claim 6, characterized in that: The real-time event recognition unit includes: A separate identification subunit is used to perform separate event identification on each inspection image in the inspection image sequence to obtain a possible event type and a probability of occurrence of the event type corresponding to each inspection image; The event probability fusion subunit is used to integrate the recognition results of multiple frames of images to generate a comprehensive probability value of the event type.

9. The image monitoring-based ferry cabin inspection system according to claim 6, characterized in that: The abnormality locating unit includes: The source contribution calculation subunit is used to quantify the contribution of the source events in the causal chain to the risk; The area mapping subunit associates events with pre-divided physical areas of the cabin; The regional risk ranking subunit aggregates risk contribution values by region and outputs them in descending order.

10. The ferry cabin inspection system based on image monitoring according to claim 6, characterized in that: The image acquisition unit includes: A multi-view camera group with time and space tags is deployed on the cabin roof and equipment nodes; Image synchronization component, which adds a location tag and timestamp to each frame of image.

Images