Collision avoidance warning method and system for a ship lock based on visual detection

Abstract

The application discloses a kind of ship lock's anti-collision early warning method and system based on visual detection, and the application relates to the technical field of visual detection, the relative distance between ship lock and ship, the current state of ship lock and the current state of ship are determined to determine the multi-gradient early warning distribution diagram of ship lock, the multi-gradient early warning distribution diagram marks multiple early warning distribution points and corresponding early warning items, improve the early warning effect of ship lock.Therefore, based on the prediction state diagram and the real-time state diagram of ship lock, determine the state delay of ship lock, based on the state delay and the position of ship lock, predict the collision risk area of ship lock;Based on the collision risk area of ship lock and the real-time image of ship, determine the anti-collision measure of ship lock, determine the anti-collision early warning signal of ship lock to ship according to the execution video of anti-collision measure of ship lock and the current speed of ship, make full use of the anti-collision early warning signal of ship lock to ship, further improve the anti-collision early warning effect of ship lock.

Description

Technical Field

[0001] This invention relates to the field of visual inspection technology, and in particular to a collision avoidance warning method and system for ship locks based on visual inspection. Background Technology

[0002] With the development of technology and the continuous increase in the number of ships, the safety and passage efficiency of locks, as key channels connecting different waterways, are particularly important. Waterways have corresponding sub-channels and locks. Ships navigate through sub-channels and can only proceed further when the locks are open. In current technology, sub-channels accommodate multiple ships, and ships gradually approach the locks. However, when a lock is open, it may only be partially open; the lock is not fully open, yet ships still pass through. Furthermore, the locks do not provide multi-level early warning during ship navigation, resulting in poor warning effectiveness and making it easy for actual collision zones to exist between the locks and ships. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a collision avoidance warning method and system for ship locks based on visual detection.

[0004] This invention provides a visual detection-based collision avoidance warning method for ship locks, comprising: determining a sub-channel and a corresponding ship lock based on channel A, and determining a visual detection area based on the ship lock and the corresponding sub-channel;

[0005] The dynamic interaction image between the lock and the ship is determined based on the detection of the visual detection area. The current state of the lock and the ship is determined based on the dynamic interaction image. The dynamic interaction image shows the interactive action events of the lock and the ship at the same point in time.

[0006] A multi-gradient early warning distribution map of the lock is determined based on the relative distance between the lock and the ship, the current state of the lock, and the current state of the ship. This multi-gradient early warning distribution map marks multiple early warning distribution points and corresponding early warning items.

[0007] At each early warning distribution point, the predicted state map of the lock is determined based on the prediction of the early warning project. Based on the predicted state map and the real-time state map of the lock, the state delay of the lock is determined. Based on the state delay and the location of the lock, the collision risk area of ​​the lock is predicted. The state delay is the time delay or height difference. The collision risk area is marked as high-risk area and low-risk area.

[0008] Based on the collision risk area of ​​the lock and real-time images of the ship, the lock's anti-collision measures are determined, and the lock's anti-collision warning signal to the ship is determined based on the video of the lock's implementation of the anti-collision measures and the ship's current speed.

[0009] This invention provides a visual detection-based collision avoidance warning system for ship locks, which is applied to the aforementioned visual detection-based collision avoidance warning method for ship locks.

[0010] Compared with the prior art, the beneficial effects of the present invention are:

[0011] In this embodiment of the invention, the method determines a sub-channel and a corresponding lock based on channel A, and a visual detection area based on the lock and the corresponding sub-channel. A dynamic interaction image between the lock and the vessel is determined based on the detection in the visual detection area. The current state of the lock and the vessel is determined based on this dynamic interaction image. A multi-gradient early warning distribution map of the lock is determined based on the relative distance between the lock and the vessel, the current state of the lock, and the current state of the vessel. This multi-gradient early warning distribution map marks multiple early warning distribution points and corresponding early warning items, incorporating a holistic consideration of the relative distance between the lock and the vessel, the current state of the lock, and the current state of the vessel. This improves the accuracy of the multi-gradient early warning distribution map and allows for the management of corresponding early warning items at multiple early warning distribution points, thereby enhancing the early warning effect of the lock.

[0012] Therefore, at each early warning distribution point, the predicted state map of the lock is determined based on the prediction of the early warning project. Based on the predicted state map and the real-time state map of the lock, the state delay of the lock is determined. Based on the state delay and the location of the lock, the collision risk area of ​​the lock is predicted. Based on the collision risk area of ​​the lock and the real-time image of the ship, the anti-collision measures of the lock are determined. Based on the video of the lock's execution of the anti-collision measures and the current speed of the ship, the anti-collision warning signal of the lock to the ship is determined. The prediction of the collision risk area of ​​the lock is introduced, and the real-time prediction of the lock is realized. It is compatible with the overall consideration of the collision risk area of ​​the lock and the real-time image of the lock, which improves the effectiveness of the anti-collision measures of the lock. It also makes full use of the anti-collision warning signal of the lock to the ship, further improving the anti-collision warning effect of the lock and avoiding the existence of actual collision areas between the lock and the ship. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the collision avoidance warning method for ship locks based on visual detection in an embodiment of the present invention.

[0014] Figure 2 This is a flowchart illustrating step S11 of the visual detection-based anti-collision warning method for ship locks in this embodiment of the invention.

[0015] Figure 3 This is a flowchart illustrating step S12 of the visual detection-based anti-collision warning method for ship locks in this embodiment of the invention.

[0016] Figure 4 This is a flowchart illustrating step S13 of the visual detection-based anti-collision warning method for ship locks in an embodiment of the present invention.

[0017] Figure 5 This is a flowchart illustrating step S14 of the visual detection-based anti-collision warning method for ship locks in an embodiment of the present invention.

[0018] Figure 6 This is a flowchart illustrating step S15 of the visual detection-based anti-collision warning method for ship locks in an embodiment of the present invention.

[0019] Figure 7 This is a schematic diagram of the structural composition of a visual detection-based anti-collision warning system for a ship lock according to an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0021] Please see Figures 1 to 7 A vision-based collision avoidance warning method for ship locks is proposed and applied to collision avoidance warning scenarios in ship locks. The vision-based collision avoidance warning method for ship locks includes:

[0022] Step S11: Determine the sub-channel and corresponding lock based on channel A, and determine the visual detection area based on the lock and the corresponding sub-channel;

[0023] Step S12: Determine the dynamic interaction image between the lock and the ship based on the detection of the visual detection area, and determine the current state of the lock and the current state of the ship based on the dynamic interaction image.

[0024] Step S13: Based on the relative distance between the lock and the ship, the current state of the lock and the current state of the ship, determine the multi-gradient early warning distribution map of the lock. The multi-gradient early warning distribution map marks multiple early warning distribution points and corresponding early warning items.

[0025] Step S14: At each early warning distribution point, determine the predicted state map of the lock based on the prediction of the early warning project, determine the state delay of the lock based on the predicted state map and the real-time state map of the lock, and predict the collision risk area of ​​the lock based on the state delay and the location of the lock.

[0026] Step S15: Determine the anti-collision measures of the lock based on the collision risk area of ​​the lock and the real-time image of the ship, and determine the anti-collision warning signal of the lock to the ship based on the video of the lock's execution of the anti-collision measures and the ship's current speed.

[0027] refer to Figure 2In step S11, a sub-channel and the corresponding lock are determined based on channel A, and a visual detection area is determined based on the lock and the corresponding sub-channel.

[0028] In the specific implementation of this invention, the specific steps are as follows:

[0029] S111: Collect the location of Channel A, determine the channel area corresponding to Channel A based on the regional search of Channel A's location, and output the regional distribution map of the channel area corresponding to Channel A.

[0030] S112: Based on the identification of the regional distribution map, determine the sub-channels and corresponding locks, and mark the channel distance of the sub-channels and the spatial location of the locks in the sub-channels;

[0031] S113: Collect the channel morphology of the sub-channel, and determine the visual detection area based on the channel morphology, channel distance, and spatial location of the lock in the sub-channel.

[0032] In the embodiments of this application, the location of channel A is collected, and the channel area corresponding to channel A is determined based on the regional retrieval of the location of channel A. The regional distribution map of the channel area corresponding to channel A is then output, which takes into account the overall consideration of the regional retrieval of the location of channel A and ensures the accuracy of the channel area corresponding to channel A.

[0033] At this point, the precise geographical location information of waterway A is obtained, including the starting point, ending point, width, length, and curvature. Simultaneously, high-precision GPS positioning devices are placed at key locations along the waterway (such as the starting point, ending point, and turning points) to collect the latitude and longitude coordinates of these points. Combined with Geographic Information System (GIS) software, the collected GPS data is imported into the GIS system to generate a geographic information model of the waterway.

[0034] The complete waterway area where waterway A is located is determined through a regional retrieval algorithm, including information such as the waterway's boundaries, width, and length. At this point, spatial analysis tools in the GIS system are used to retrieve the complete area where the waterway is located based on the collected waterway location data. In the GIS system, the collected waterway coordinate points are connected to form a line, creating the centerline of the waterway. The waterway boundary line is generated based on the waterway width. Finally, a regional polygon of the waterway is generated, including both sides of the waterway.

[0035] The identified waterway area is displayed graphically to facilitate subsequent analysis and use. In the GIS system, the waterway area polygon is displayed graphically, and key information such as the starting point, ending point, width, and length of the waterway is marked. The regional distribution map can be output as a vector map (such as SVG format) or a raster map (such as PNG format) for easy viewing on different devices and platforms.

[0036] Furthermore, based on the identification of the regional distribution map, sub-channels and corresponding locks are determined, and the channel distance of the sub-channels and the spatial location of the locks in the sub-channels are marked. This is compatible with the overall consideration of the identification of the regional distribution map and ensures the accuracy of the sub-channels and corresponding locks.

[0037] At this point, sub-channels and corresponding locks are identified from the regional distribution map of channel A, and their positional relationship is clarified. Simultaneously, image processing algorithms (such as edge detection and contour extraction) are used to identify the contours of sub-channels and locks from the regional distribution map. Combined with the spatial analysis tools of GIS software, sub-channels are divided and locks are located based on the geometric shape and location information of the channel. For complex or difficult-to-identify areas, manual annotation is used for correction and supplementation.

[0038] Optionally, a regional distribution map of Channel A has already been generated. Now, it is necessary to identify the sub-channels and locks from it. Channel A is a Y-shaped channel, divided into two sub-channels, each with a lock at its end. Edge detection algorithms (such as Canny edge detection) are used to extract the channel boundaries from the regional distribution map. Contour extraction algorithms (such as OpenCV's findContours) are used to identify the contours of the sub-channels and locks. Using GIS software, based on the channel's geometry and location information, the two sub-channels are divided, and the locks at the ends of each sub-channel are located. The automatic identification results are checked, and an error is found in the boundary identification of one sub-channel. The boundaries are manually adjusted to ensure the accuracy of the sub-channel and lock contours.

[0039] Measure and record key distance information such as the length and width of each sub-channel; use the measurement tools in GIS software to calculate the length and width of the sub-channel based on its outline; for curved channels, use the curve length calculation formula. Further, determine the specific location of the lock within the sub-channel, including coordinates and relative position information; use the spatial analysis tools in GIS software to calculate the center coordinates of the lock based on its outline and the boundary of the sub-channel; determine the relative position of the lock within the sub-channel (e.g., distance from the starting point of the sub-channel).

[0040] Therefore, by collecting the channel morphology of the sub-channels and determining the visual inspection area based on the channel morphology, channel distance, and spatial position of the lock within the sub-channels, the overall consideration of the channel morphology, channel distance, and spatial position of the lock within the sub-channels is taken into account, ensuring the accuracy of the visual inspection area.

[0041] At this point, detailed morphological information of the sub-channel is acquired, including its shape, curvature, and obstacles, to provide a basis for the division of visual inspection areas. Simultaneously, aerial imagery is used: drones or satellite imagery are used to obtain a top-down view of the sub-channel, clearly showing its shape and curvature; underwater photography is used to capture the underwater morphology (such as shoals and reefs); LiDAR is used to acquire elevation information of the channel for analyzing depth changes; and a sonar system is used to measure the underwater topography of the channel, especially for deeper sections. The data collected by these multiple sensors are then fused to generate a complete channel morphology model.

[0042] Based on the channel morphology, channel distance, and lock location, the visual inspection area is rationally divided. The visual inspection area can comprehensively cover the ship's passage path and potential collision risk areas. At this time, for a straight channel (such as sub-channel 1), the visual inspection area can be evenly distributed on both sides of the channel centerline; for a curved channel (such as sub-channel 2), the visual inspection area needs to be adjusted according to the degree of curvature and the location of shoals to ensure coverage of the curved part and shoal area. The extent of the visual inspection area is determined according to the length and width of the channel. Using GIS software or image processing tools, the visual inspection area is divided according to the above analysis. The visual inspection area can be divided into multiple sub-regions, each sub-region corresponding to the detection range of one or more cameras.

[0043] refer to Figure 3 In step S12, a dynamic interaction image between the lock and the ship is determined based on the detection of the visual detection area, and the current state of the lock and the current state of the ship are determined based on the dynamic interaction image.

[0044] In the specific implementation of this invention, the specific steps are as follows:

[0045] S121: Collect the visual detection area and perform real-time detection on the visual detection area. Collect multiple images based on the real-time detection of the visual detection area. The multiple images include the current image of the lock, the current image of the ship, and the dynamic image between the lock and the ship.

[0046] S122: The dynamic interaction image between the lock and the ship is determined by the synthesis of multiple images. The dynamic interaction image between the lock and the ship presents the interactive action events of the lock and the ship at the same time point.

[0047] S123: Determine the lock's action content and the ship's action content based on the detection of the interactive action event, collect the ship's position relative to the sub-channel, and determine the lock's current state and the ship's current state based on the ship's position relative to the sub-channel, the ship's action content, and the lock's action content.

[0048] In the embodiments of this application, a visual detection area is acquired and the visual detection area is detected in real time. Multiple images are acquired based on the real-time detection of the visual detection area. The multiple images include the current image of the lock, the current image of the ship, and the dynamic image between the lock and the ship.

[0049] At this point, real-time image data within the visual inspection area is acquired to provide a basis for subsequent analysis. Simultaneously, multiple high-definition cameras are installed within the visual inspection area to ensure coverage of the lock, ships, and the interaction areas between them. The camera installation locations should include the lock entrance, exit, gate area, and key locations in the waterway.

[0050] Real-time image acquisition within the visual inspection area ensures the capture of dynamic changes in locks and vessels. This is achieved using industrial cameras or smart cameras with a suitable frame rate (e.g., 30fps or higher). The image acquisition system must operate stably and transmit the acquired images to the central processing unit. The acquired image data must be stored on a server or cloud storage for subsequent analysis. Data integrity and security must be ensured.

[0051] Acquiring images of the locks, vessels, and their dynamic interactions provides comprehensive data support for subsequent analysis. The acquired images are categorized into three types: current images of the locks (displaying gate status, position, etc.); current images of the vessels (displaying position, speed, direction, etc.); and dynamic interaction images between the locks and vessels (displaying interactive events between the locks and vessels at the same point in time). The image acquisition frequency is set according to actual needs. For example, acquiring 10 images per second ensures that dynamic changes are captured.

[0052] Optionally, a preset image matching table can be used, as shown in Table 1:

[0053] Table 1 Image Matching Table

[0054]

[0055] Furthermore, the dynamic interaction image between the lock and the ship is determined by synthesizing multiple images. The dynamic interaction image between the lock and the ship presents the interactive action events of the lock and the ship at the same time point, realizing the determination of the dynamic interaction between the lock and the ship by synthesizing multiple images, and further enabling precise control of the images.

[0056] At this point, images captured by multiple cameras are combined to generate an image that comprehensively displays the dynamic interaction between the lock and the ship. Simultaneously, preprocessing operations such as noise reduction and contrast enhancement are performed on the captured images to ensure image quality; the timestamps of all images are kept consistent for synchronous synthesis; image registration techniques (such as feature point matching) are used to align images captured by different cameras to the same coordinate system; and the spatial relationships between images are ensured to be correct, such as the positional relationship between the lock and the ship.

[0057] Multiple images are combined to generate a dynamic interactive image that shows the lock and the ship at the same point in time. At this time, image stitching technology (such as OpenCV's cv2.stitch) is used to stitch multiple images into a panoramic image; or image fusion technology (such as weighted average fusion) is used to fuse multiple images into a single image; the current image of the lock, the current image of the ship, and the dynamic interactive image are aligned according to timestamps; these images are then combined into a single image to ensure that the interactive action events between the lock and the ship can be clearly displayed.

[0058] The interaction events between the lock and the ship are identified from the synthesized dynamic interactive images. At this time, the positions of the lock and the ship are detected using object detection algorithms (such as YOLO and SSD) to ensure accurate identification of the lock gate status and the ship's motion status. The motion trajectory of the ship is tracked using trajectory tracking algorithms (such as Kalman filter). The ship's speed changes, direction changes and other motion events are analyzed. Based on the results of object detection and trajectory tracking, the interaction events between the lock and the ship are identified. For example, the ship's deceleration, acceleration, and turning, as well as the opening and closing of the lock gates, are identified.

[0059] Optionally, an image showing the dynamic interaction between lock 1 and vessel 1 has been synthesized. Now, action event recognition is performed: the gate of lock 1 is detected to be closed; the position of vessel 1 is detected, 100 meters away from the lock entrance; the movement trajectory of vessel 1 is tracked, and the speed of vessel 1 is found to decrease from 10 knots to 5 knots; it is recognized that vessel 1 is decelerating and preparing to enter lock 1; the gate of lock 1 is recognized to be closed. A schematic table of the finally recognized interactive action events is collected, as shown in Table 2:

[0060] Table 2. Diagram of Interactive Action Events

[0061]

[0062] Therefore, the lock's action content and the ship's action content are determined based on the detection of this interactive action event. The ship's position relative to the sub-channel is collected. The current state of the lock and the ship is determined based on the ship's position relative to the sub-channel, the ship's action content, and the lock's action content. This method takes into account the overall consideration of the ship's position relative to the sub-channel, the ship's action content, and the lock's action content, ensuring the accuracy of the lock's current state and the ship's current state.

[0063] At this point, the specific actions of the lock are extracted from the interactive action events, such as the opening and closing of the gate, to identify the gate status; whether the gate is in an open or closed state is detected; the changes in the gate state are analyzed, such as from open to closed or from closed to open; and the timestamps of the gate actions are recorded for subsequent analysis. Optionally, the following lock actions have been detected using image recognition technology: Lock 1: Current state: Gate closed; Action content: Gate changes from open to closed; Timestamp: 2025-06-23 10:00:00; Lock 2: Current state: Gate open; Action content: Gate remains open; Timestamp: 2025-06-23 10:05:00.

[0064] Extract specific ship actions from interactive events, such as changes in speed and direction, identify the ship's position, and use trajectory tracking algorithms (such as Kalman filters) to track the ship's trajectory. Analyze the ship's speed and direction changes; record the timestamps of the ship's actions for subsequent analysis. Optionally, the following ship actions have been detected using image recognition and trajectory tracking technology: Ship 1: Current state: decelerating; Action: speed reduced from 10 knots to 5 knots; Timestamp: 2025-06-23 10:00:00; Location: 100 meters from the entrance of Lock 1. Ship 2: Current state: passing at a constant speed; Action: speed maintained at 5 knots; Timestamp: 2025-06-23 10:05:00; Location: 300 meters from the entrance of Lock 2.

[0065] Determine the specific position of the vessel in the sub-channel in order to assess its relative position to the lock. Convert the vessel's position in the image to the actual coordinates of the sub-channel using known channel geometry information (such as width and length). Calculate the distance of the vessel relative to the starting point of the sub-channel. Determine the specific position of the vessel in the channel. Optionally, Vessel 1: Position: 1500 meters from the starting point of sub-channel 1; Relative position: 100 meters from the entrance of lock 1. Vessel 2: Position: 2000 meters from the starting point of sub-channel 2; Relative position: 300 meters from the entrance of lock 2.

[0066] By combining the actions of the vessel and the lock with the vessel's position information, their current states are determined. Based on the vessel's actions (such as speed and direction) and position, the vessel's current state is evaluated. Similarly, based on the lock's actions (such as gate status) and timestamps, the lock's current state is evaluated. The evaluation results are recorded in a state table for subsequent analysis and early warning. At this point, the actions and position information of the vessel and lock have been comprehensively analyzed, and a state matching table has been determined, as shown in Table 3.

[0067] Table 3 State Matching Table

[0068]

[0069] refer to Figure 4 In step S13, a multi-gradient early warning distribution map of the lock is determined based on the relative distance between the lock and the ship, the current state of the lock, and the current state of the ship. This multi-gradient early warning distribution map marks multiple early warning distribution points and corresponding early warning items.

[0070] In the specific implementation of this invention, the specific steps are as follows:

[0071] S131: Collect the relative distance between the lock and the ship, determine the control area of ​​the lock relative to the ship based on the relative distance and the distribution map corresponding to the sub-channel, and determine multiple early warning distribution nodes according to the division of the control area;

[0072] S132: In each early warning distribution node, a first-gradient early warning dynamic map is determined based on the location of the early warning distribution node, the distance between two adjacent early warning distribution nodes, and the current state of the lock; a second-gradient early warning dynamic map is determined based on the location of the early warning distribution node, the distance between two adjacent early warning distribution nodes, and the current state of the ship.

[0073] S133: The multi-gradient early warning distribution map of the lock is determined based on the synthesis of the first-gradient early warning dynamic map and the second-gradient early warning dynamic map. At this time, the early warning item corresponding to the early warning distribution point is determined according to the location of the early warning distribution node, the current state of the lock and the current state of the ship. The multi-gradient early warning distribution map is marked with each early warning item.

[0074] In the embodiments of this application, the relative distance between the lock and the ship is collected, and the control area of ​​the lock relative to the ship is determined based on the relative distance and the distribution map corresponding to the sub-channel. Multiple early warning distribution nodes are determined according to the division of the control area, which is compatible with the overall consideration of the relative distance and the distribution map corresponding to the sub-channel, and ensures the accuracy of the control area of ​​the lock relative to the ship.

[0075] At this point, the real-time distance between the lock and the ship is obtained to provide basic data for subsequent control area division and early warning node setting. Deep learning algorithms (such as YOLO and SSD) are used to detect the position of the lock and the ship from the dynamic interactive image and calculate the Euclidean distance between the lock and the ship. At the same time, more accurate distance data is obtained by combining LiDAR or ultrasonic sensors, and the sensor data is fused with the image processing results to improve the accuracy of distance measurement.

[0076] Optionally, the following relative distances have been calculated using image processing technology: 100 meters between lock 1 and ship 1; 300 meters between lock 2 and ship 2. Simultaneously, using a lidar sensor for assisted measurement, the following data were obtained: 102 meters between lock 1 and ship 1; 305 meters between lock 2 and ship 2. The average of these two distances is taken as the relative distance: 101 meters between lock 1 and ship 1; (100 + 102) / 2 = 101 meters; (300 + 305) / 2 = 302.5 meters.

[0077] Based on relative distances and sub-channel distribution maps, the control area of ​​the lock is delineated; the positional relationship between the lock and vessels is determined using the sub-channel distribution maps; the control area is delineated based on the geometry of the sub-channels and the location of the lock; based on relative distances, the control area is divided into multiple intervals, such as 0-50 meters, 50-100 meters, 100-200 meters, etc.; each interval corresponds to a different risk level (e.g., high-risk area, medium-risk area, low-risk area); optionally, the control areas have already been delineated based on the sub-channel distribution maps as follows:

[0078] Sub-channel 1: High-risk area: 0-50 meters; Medium-risk area: 50-100 meters; Low-risk area: 100-200 meters; Sub-channel 2: High-risk area: 0-50 meters; Medium-risk area: 50-100 meters; Low-risk area: 100-200 meters; Based on relative distance: Lock 1 and Ship 1: Relative distance is 101 meters, located in the medium-risk area; Lock 2 and Ship 2: Relative distance is 302.5 meters, located in the low-risk area.

[0079] Multiple early warning distribution nodes are defined within the controlled area for subsequent generation of early warning dynamic maps; multiple early warning nodes are evenly distributed within each controlled area; the location of the early warning nodes can be adjusted according to actual needs, for example, one node can be set every 10 meters or 20 meters; the location and corresponding risk level of each early warning node are marked; the location of the early warning nodes should correspond to the distribution of sub-channels. Figure 1To ensure accuracy, the warning nodes within the controlled area can be divided as follows: Sub-channel 1: High-risk area (0-50 meters): Node 1 (10 meters); Node 2 (20 meters); Node 3 (30 meters); Node 4 (40 meters); Node 5 (50 meters); Medium-risk area (50-100 meters): Node 6 (60 meters); Node 7 (70 meters); Node 8 (80 meters); Node 9 (90 meters); Node 10 (100 meters); Low-risk area (100-200 meters): Node 11 (110 meters); Node 12 (120 meters); Node 13 (130 meters); Node 14 (140 meters); Node 15 (150 meters); A preset risk level matching table is collected, as shown in Table 4.

[0080] Table 4 Risk Level Matching Table

[0081]

[0082] Furthermore, in each early warning distribution node, a first-gradient early warning dynamic map is determined based on the location of the early warning distribution node, the distance between two adjacent early warning distribution nodes, and the current state of the lock. A second-gradient early warning dynamic map is determined based on the location of the early warning distribution node, the distance between two adjacent early warning distribution nodes, and the current state of the ship. This approach takes into account the overall consideration of the location of the early warning distribution node, the distance between two adjacent early warning distribution nodes, and the current state of the ship, ensuring the accuracy of the second-gradient early warning dynamic map.

[0083] At this point, based on the location of the early warning distribution nodes, the distance between adjacent nodes, and the current state of the lock, a first-gradient early warning dynamic map is generated. The specific location of each early warning distribution node is determined, including its coordinates in the waterway, and the distance between two adjacent early warning distribution nodes is calculated. The current state of the lock (such as whether the gate is open or closed) is obtained. Based on the current state of the lock and the location of the early warning distribution nodes, the risk level of each node is evaluated. Optionally, the current state of lock 1 is "gate closed". The location and risk level of the early warning distribution nodes are as follows: Lock 1: Node 6 (60 meters): High-risk area → Risk level: High risk; Node 7 (70 meters): Medium-risk area → Risk level: Medium risk; Node 8 (80 meters): Medium-risk area → Risk level: Medium risk; Node 9 (90 meters): Low-risk area → Risk level: Low risk; Node 10 (100 meters): Low-risk area → Risk level: Low risk.

[0084] Based on the location of the early warning distribution nodes, the distance between adjacent nodes, and the current state of the ship, a second-gradient early warning dynamic map is generated. The specific location of each early warning distribution node is determined, including its coordinates in the channel. The distance between two adjacent early warning distribution nodes is calculated, and the current state of the ship (such as speed and direction) is obtained. Based on the current state of the ship and the location of the early warning distribution nodes, the risk level of each node is assessed. Optionally, if ship 1's speed is 10 knots, the location and risk level of the early warning distribution nodes are as follows: Ship 1: Node 6 (60 meters): High-risk area → Risk level: High risk; Node 7 (70 meters): Medium-risk area → Risk level: Medium risk; Node 8 (80 meters): Medium-risk area → Risk level: Medium risk; Node 9 (90 meters): Low-risk area → Risk level: Low risk; Node 10 (100 meters): Low-risk area → Risk level: Low risk; Ship 2's speed is 5 knots, and the location and risk level of the early warning distribution nodes are as follows: Ship 2: Node 26 (110 meters): High-risk area → Risk level: Medium risk; Node 27 (120 meters): Medium-risk area → Risk level: Low risk; Node 28 (130 meters): Medium risk area → Risk level: Low risk; Node 29 (140 meters): Low risk area → Risk level: Low risk; Node 30 (150 meters): Low risk area → Risk level: Low risk; Visualize the assessment results and generate an early warning dynamic map. Use GIS software or image processing tools (such as Matplotlib) to generate the early warning dynamic map. Mark the risk level of each node on the map with different colors or symbols. The generated early warning dynamic map should clearly show the location and risk level of each node.

[0085] Optionally, a preset warning animation can be collected, as shown in Table 5:

[0086] Table 5. Early Warning Dynamic Chart

[0087]

[0088] Therefore, a multi-gradient early warning distribution map of the lock is determined by synthesizing the first-gradient early warning dynamic map and the second-gradient early warning dynamic map. At this time, the early warning item corresponding to the early warning distribution point is determined according to the location of the early warning distribution node, the current state of the lock, and the current state of the ship. The multi-gradient early warning distribution map is marked with each early warning item, which takes into account the overall consideration of the location of the early warning distribution node, the current state of the lock, and the current state of the ship, ensuring the accuracy of the early warning item corresponding to the early warning distribution point. At the same time, it takes into account the overall consideration of the relative distance between the lock and the ship, the current state of the lock, and the current state of the ship, improving the accuracy of the multi-gradient early warning distribution map. Furthermore, the corresponding early warning items are controlled at multiple early warning distribution points, thereby improving the early warning effect of the lock.

[0089] At this point, the first-gradient (based on lock status) and second-gradient (based on ship status) early warning dynamic maps are synthesized to generate a comprehensive multi-gradient early warning distribution map; the risk scores of each node in the first-gradient and second-gradient early warning dynamic maps are weighted and averaged or directly added together; the comprehensive risk score of each node after synthesis is determined; thresholds can be set to classify risk levels, for example: high risk: score ≥ 4; medium risk: 2 ≤ score < 4; low risk: score < 2.

[0090] Optionally, the risk scores for each node in the first and second gradient early warning dynamic diagrams have been calculated as follows: Lock 1: First gradient: Node 6: 3.1; Node 7: 2.7; Node 8: 2.3; Node 9: 1.9; Node 10: 1.5; Second gradient: Node 6: 4.9; Node 7: 4.5; Node 8: 4.1; Node 9: 3.7; Node 10: 3.3; Lock 2: First gradient: Node 26: 4.9; Node 27: 4.5; Node 28: 4.1; Node 29: 3.7; Node 30: 3.3; Second gradient: Node 26: 3.9; Node 27: 3.5; Node 28: 3.1; Node 29: 2.7; Node 30: 2.3; Comprehensive risk score calculation: Lock 1: Node 6: (3.1 + 4.9) / 2 = 4.0; Node 7: (2.7 + 4.5) / 2 = 3.6; Node 8: (2.3 + 4.1) / 2 = 3.2; Node 9: (1.9 + 3.7) / 2 = 2.8; Node 10: (1.5 + 3.3) / 2 = 2.4; Lock 2: Node 26: (4.9 + 3.9) / 2 = 4.4; Node 27: (4.5 + 3.5) / 2 = 4.0; Node 28: (4.1 + 3.1) / 2 = 3.6; Node 29: (3.7 + 2.7) / 2 = 3.2; Node 30: (3.3 + 2.3) / 2 = 2.8.

[0091] Based on the comprehensive risk score and node location, the specific warning items for each warning distribution point are determined. Based on the comprehensive risk score, each node is divided into different risk levels, and specific warning items are assigned according to the risk level and node location; for example: high risk: emergency deceleration and avoidance; medium risk: deceleration and course adjustment; low risk: maintain the current state.

[0092] Optionally, risk levels can be categorized based on the comprehensive risk score as follows: High risk: score ≥ 4; Medium risk: 2 ≤ score < 4; Low risk: score < 2; Lock 1: Node 6: 4.0 (High risk) → Warning item: Emergency deceleration; Node 7: 3.6 (High risk) → Warning item: Emergency deceleration; Node 8: 3.2 (Medium risk) → Warning item: Deceleration; Node 9: 2.8 (Medium risk) → Warning item: Deceleration; Node 10: 2.4 (Low risk) → Warning item: Maintain current state;

[0093] Lock 2: Node 26: 4.4 (High Risk) → Warning Item: Emergency Deceleration; Node 27: 4.0 (High Risk) → Warning Item: Emergency Deceleration; Node 28: 3.6 (Medium Risk) → Warning Item: Deceleration; Node 29: 3.2 (Medium Risk) → Warning Item: Deceleration; Node 30: 2.8 (Low Risk) → Warning Item: Maintain Current Status.

[0094] Mark the risk level and warning items for each warning distribution point on a multi-gradient warning distribution map. Use GIS software or image processing tools (such as Matplotlib) to generate the multi-gradient warning distribution map, marking the risk level and warning items for each node on the map with different colors or symbols. The generated multi-gradient warning distribution map should clearly show the location, risk level, and warning items for each node. Optionally, collect a preset multi-gradient warning distribution map, as shown in Table 6:

[0095] Table 6 Multi-gradient early warning distribution map

[0096]

[0097] refer to Figure 5 In step S14, at each early warning distribution point, the predicted state map of the lock is determined based on the prediction of the early warning item, the state delay of the lock is determined based on the predicted state map and the real-time state map of the lock, and the collision risk area of ​​the lock is predicted based on the state delay and the location of the lock.

[0098] In the specific implementation of this invention, the specific steps are as follows:

[0099] S141: Monitor each early warning distribution point in real time, and determine the control event of the lock based on the analysis of the early warning item. Determine the predicted state diagram of the lock based on the current state of the lock and the prediction of the control event. The predicted state diagram of the lock serves as the theoretical state diagram of the lock in the early warning distribution point.

[0100] S142: If a ship passes through the warning distribution point during its navigation in the sub-channel, the lock status detection is triggered at the same time node to collect the real-time status map of the lock. The lock status delay is determined by comparing the predicted status map with the real-time status map of the lock. The lock status delay reflects the height difference of the lock under height adjustment.

[0101] S143: Determine the first sub-collision risk area of ​​the lock based on the state delay amount and the current position of the vessel, determine the second sub-collision risk area of ​​the lock based on the state delay amount and the position of the lock, and determine the collision risk area of ​​the lock based on the combination of the first sub-collision risk area and the second sub-collision risk area of ​​the lock.

[0102] In the embodiments of this application, each early warning distribution point is monitored in real time, and the control event of the lock is determined based on the analysis of the early warning item. The predicted state diagram of the lock is determined based on the current state of the lock and the prediction of the control event. The predicted state diagram of the lock serves as the theoretical state diagram of the lock in the early warning distribution point, which takes into account the overall consideration of the current state of the lock and the prediction of the control event, and ensures the accuracy of the predicted state diagram of the lock.

[0103] At this time, images and data from the warning distribution points are acquired in real time to promptly detect potential collision risks. High-definition cameras and sensors (such as lidar and ultrasonic sensors) are installed at each warning distribution point to ensure that the cameras and sensors cover the key areas of the locks and vessels. Image acquisition systems are used to acquire image data in real time, and the acquired data is transmitted to a central processing unit for analysis. Image processing algorithms (such as target detection and trajectory tracking) are used to analyze the image data in real time and detect changes in the state of the locks and vessels. Optionally, cameras and sensors are installed at the warning distribution points of locks 1 and 2 for real-time monitoring as follows:

[0104] Lock 1: Node 6: Real-time camera image shows that the gate of Lock 1 is closed; Node 7: Real-time camera image shows that the gate of Lock 1 is closed; Node 8: Real-time camera image shows that the gate of Lock 1 is closed; Node 9: Real-time camera image shows that the gate of Lock 1 is closed; Node 10: Real-time camera image shows that the gate of Lock 1 is closed.

[0105] Lock 2: Node 26: Real-time camera image shows that the gate of Lock 2 is open; Node 27: Real-time camera image shows that the gate of Lock 2 is open; Node 28: Real-time camera image shows that the gate of Lock 2 is open; Node 29: Real-time camera image shows that the gate of Lock 2 is open; Node 30: Real-time camera image shows that the gate of Lock 2 is open.

[0106] Based on the analysis of the early warning items, the control events required for the lock are determined. According to the early warning items marked on the multi-gradient early warning distribution map, the specific requirements for each early warning distribution point are analyzed. Early warning items include emergency deceleration, avoidance, and course adjustment. Based on the early warning items, the control events required for the lock are determined. Control events include gate opening, closing, or other operations. Optional early warning items are as follows:

[0107] Lock 1: Node 6: Warning item is "Emergency deceleration", control event is "Keep gate closed"; Node 7: Warning item is "Emergency deceleration", control event is "Keep gate closed"; Node 8: Warning item is "Deceleration", control event is "Keep gate closed"; Node 9: Warning item is "Deceleration", control event is "Keep gate closed"; Node 10: Warning item is "Maintain current state", control event is "Keep gate closed".

[0108] Lock 2: Node 26: Warning item is "Emergency deceleration", control event is "Keep gate open"; Node 27: Warning item is "Emergency deceleration", control event is "Keep gate open"; Node 28: Warning item is "Deceleration", control event is "Keep gate open"; Node 29: Warning item is "Deceleration", control event is "Keep gate open"; Node 30: Warning item is "Maintain current state", control event is "Keep gate open".

[0109] Based on the current state of the lock and the prediction of control events, a predicted state diagram of the lock is generated. Based on the current state of the lock and the control events, the state changes of the lock in the future are predicted. Time series analysis or machine learning models are used for state prediction. Based on the prediction results, a predicted state diagram is generated to show the theoretical state of the lock. The predicted state diagram should include information such as the opening and closing status of the gates and the position of the ships.

[0110] Optionally, the current state of Lock 1 is "Gate Closed", the control event is "Keep Gate Closed", and the predicted state diagram is as follows: Lock 1: Node 6: Predicted state is "Gate Closed"; Node 7: Predicted state is "Gate Closed"; Node 8: Predicted state is "Gate Closed"; Node 9: Predicted state is "Gate Closed"; Node 10: Predicted state is "Gate Closed"; The current state of Lock 2 is "Gate Open", the control event is "Keep Gate Open", and the predicted state diagram is as follows: Lock 2: Node 26: Predicted state is "Gate Open"; Node 27: Predicted state is "Gate Open"; Node 28: Predicted state is "Gate Open"; Node 29: Predicted state is "Gate Open"; Node 30: Predicted state is "Gate Open".

[0111] Furthermore, if a vessel passes through the warning distribution point during its navigation in the sub-channel, the lock's status detection is triggered at the same time node to collect the lock's real-time status map. The lock's status delay is determined by comparing the predicted status map with the lock's real-time status map. This lock's status delay reflects the height difference of the lock under height adjustment, taking into account the overall consideration of comparing the predicted status map and the lock's real-time status map, thus ensuring the accuracy of the lock's status delay.

[0112] At this time, the lock status is monitored in real time as the vessel passes the warning distribution point; the vessel's position is monitored in real time using GPS or other positioning technologies. When the vessel approaches or passes the warning distribution point, status detection is triggered. At the same time, the lock status detection system is activated, and cameras and sensors are used to collect real-time status images of the lock. Optionally, when vessel 1 is navigating in sub-channel 1 and approaches node 6, the status detection of lock 1 is triggered: Vessel 1: Current position: 100 meters from the entrance of lock 1; Status detection triggered: The system automatically activates the camera and sensors of lock 1. When vessel 2 is navigating in sub-channel 2 and approaches node 26, the status detection of lock 2 is triggered: Vessel 2: Current position: 300 meters from the entrance of lock 2; Status detection triggered: The system automatically activates the camera and sensors of lock 2.

[0113] The actual status of the locks is obtained when a ship passes through the warning distribution points. High-definition cameras are used to capture images of the locks in real time, ensuring the images clearly show the lock gate status (e.g., open, closed). The captured image data is recorded for subsequent analysis, ensuring the image timestamps match the ship's position data. Optionally, the real-time status diagrams for Lock 1 and Lock 2 are as follows: Lock 1: Node 6: Real-time status diagram "Gate Closed"; Node 7: Real-time status diagram "Gate Closed"; Node 8: Real-time status diagram "Gate Closed"; Node 9: Real-time status diagram "Gate Closed"; Node 10: Real-time status diagram "Gate Closed". Lock 2: Node 26: Real-time status diagram "Gate Open"; Node 27: Real-time status diagram "Gate Open"; Node 28: Real-time status diagram "Gate Open"; Node 29: Real-time status diagram "Gate Open"; Node 30: Real-time status diagram "Gate Open".

[0114] The state delay of the lock is determined by comparing the predicted state diagram and the real-time state diagram. The real-time state diagram is compared with the predicted state diagram to check if the gate states are consistent (e.g., open or closed). If the gate states are inconsistent, the state delay is calculated. The state delay can be either a time delay or a height difference. The predicted and real-time state diagrams are as follows: Lock 1: Node 6: Predicted state is "Gate Closed", Real-time state is "Gate Closed" → State delay is 0; Node 7: Predicted state is "Gate Closed", Real-time state is "Gate Closed" → State delay is 0; Node 8: Predicted state is "Gate Closed", Real-time state is "Gate Closed" → State delay is 0; Node 9: Predicted state is "Gate Closed", Real-time state is "Gate Closed" → State delay is 0; Node 10: Predicted state is "Gate Closed", Real-time state is "Gate Closed" → State delay is 0. Lock 2: Node 26: Predicted state is "Gate Open", real-time state is "Gate Open" → state delay is 0; Node 27: Predicted state is "Gate Open", real-time state is "Gate Open" → state delay is 0; Node 28: Predicted state is "Gate Open", real-time state is "Gate Open" → state delay is 0; Node 29: Predicted state is "Gate Open", real-time state is "Gate Open" → state delay is 0; Node 30: Predicted state is "Gate Open", real-time state is "Gate Open" → state delay is 0.

[0115] If there is a status delay, calculate the height difference of the lock under height adjustment; use lidar or other height measuring equipment to measure the actual height of the lock, compare the predicted height and the actual height, and calculate the height difference; record the height difference for subsequent analysis; optionally, the height difference between lock 1 and lock 2 is as follows: Lock 1: Node 6: Predicted height is 0 meters (gate closed), actual height is 0 meters → height difference is 0 meters; Node 7: Predicted height is 0 meters (gate closed), actual height is 0 meters → height difference is 0 meters; Node 8: Predicted height is 0 meters (gate closed), actual height is 0 meters → height difference is 0 meters; Node 9: Predicted height is 0 meters (gate closed), actual height is 0 meters → height difference is 0 meters; Node 10: Predicted height is 0 meters (gate closed), actual height is 0 meters → height difference is 0 meters. Lock 2: Node 26: Predicted height 5 meters (gate open), actual height 5 meters → height difference 0 meters; Node 27: Predicted height 5 meters (gate open), actual height 5 meters → height difference 0 meters; Node 28: Predicted height 5 meters (gate open), actual height 5 meters → height difference 0 meters; Node 29: Predicted height 5 meters (gate open), actual height 5 meters → height difference 0 meters; Node 30: Predicted height 5 meters (gate open), actual height 5 meters → height difference 0 meters.

[0116] Therefore, the first sub-collision risk area of ​​the lock is determined based on the state delay amount and the current position of the ship, and the second sub-collision risk area of ​​the lock is determined based on the state delay amount and the position of the lock. The collision risk area of ​​the lock is determined based on the synthesis of the first and second sub-collision risk areas, which takes into account the overall consideration of the synthesis of the first and second sub-collision risk areas of the lock, and ensures the accuracy of the collision risk area of ​​the lock.

[0117] At this point, based on the state delay and the ship's current position, the first sub-collision risk area of ​​the lock is determined; the state delay is analyzed to determine whether there is a potential collision risk. If the state delay is 0, it means that the lock state is consistent with the prediction and the risk is low. If the state delay is not 0, it means that the lock state is inconsistent with the prediction and there is a collision risk; the ship's current position is obtained, and combined with the state delay, the potential collision area between the ship and the lock is determined. The first sub-collision risk area is determined as follows: based on the state delay and the ship's position, the potential collision area between the ship and the lock is calculated. This area represents the area where the ship and the lock may collide.

[0118] Optionally, when vessel 1 is navigating in sub-channel 1 and passes node 6, the state delay is 0, and the vessel's current position is 100 meters from the entrance of lock 1; Lock 1: Node 6: State delay is 0, and the vessel's current position is 100 meters from the entrance of lock 1; First sub-collision risk area: 100 meters from the entrance of lock 1, with a range of ±10 meters; When vessel 2 is navigating in sub-channel 2 and passes node 26, the state delay is 0, and the vessel's current position is 300 meters from the entrance of lock 2; Lock 2: Node 26: State delay is 0, and the vessel's current position is 300 meters from the entrance of lock 2; First sub-collision risk area: 300 meters from the entrance of lock 2, with a range of ±10 meters.

[0119] Based on the state delay and the location of the lock, the second sub-collision risk area of ​​the lock is determined. The state delay is analyzed to determine whether there is a potential collision risk. If the state delay is 0, it means that the lock state is consistent with the prediction and the risk is low. If the state delay is not 0, it means that the lock state is inconsistent with the prediction and there is a collision risk. The location of the lock is obtained, and combined with the state delay, the potential collision area between the lock and the ship is determined. The second sub-collision risk area is determined as follows: Based on the state delay and the location of the lock, the potential collision area between the lock and the ship is calculated. This area represents the area where the lock and the ship may collide. Optionally, the state delay of lock 1 is 0, and the lock position is 1500 meters from the starting point of sub-channel 1: Lock 1: Node 6: State delay is 0, and the lock position is 1500 meters from the starting point of sub-channel 1; Second sub-collision risk area: 1500 meters from the starting point of sub-channel 1, with a range of ±10 meters; The state delay of lock 2 is 0, and the lock position is 2000 meters from the starting point of sub-channel 2: Lock 2: Node 26: State delay is 0, and the lock position is 2000 meters from the starting point of sub-channel 2; Second sub-collision risk area: 2000 meters from the starting point of sub-channel 2, with a range of ±10 meters.

[0120] The first and second sub-collision risk areas are combined to determine the final collision risk area. The first and second sub-collision risk areas are then merged to ensure that the merged area can cover all potential collision risk points. A risk assessment is performed on the merged collision risk area, which is marked with high-risk and low-risk areas. Areas with different risk levels can be marked with colors or symbols.

[0121] Optionally, the first and second sub-collision risk regions are as follows:

[0122] Lock 1: First sub-collision risk area: 100 meters from the entrance of Lock 1, with a range of ±10 meters; Second sub-collision risk area: 1500 meters from the starting point of Sub-channel 1, with a range of ±10 meters; Combined collision risk area: Covers the area from 1490 meters to 1510 meters from the starting point of Sub-channel 1.

[0123] Lock 2: First sub-collision risk area: 300 meters from the entrance of Lock 2, with a range of ±10 meters; Second sub-collision risk area: 2000 meters from the starting point of Sub-channel 2, with a range of ±10 meters; Combined collision risk area: Covers the area from 1990 meters to 2010 meters from the starting point of Sub-channel 2.

[0124] refer to Figure 6 In step S15, collision prevention measures of the lock are determined based on the collision risk area of ​​the lock and the real-time image of the ship, and collision prevention warning signal of the lock to the ship is determined based on the video of the lock's execution of the collision prevention measures and the ship's current speed.

[0125] In the specific implementation of this invention, the specific steps are as follows:

[0126] S151: Collect the ship's navigation direction in the sub-channel, determine multiple collision contact points in the collision risk area based on the collision risk area of ​​the lock, the ship's navigation direction and the ship's current navigation position, determine the lock's anti-collision event based on the comparison of multiple collision contact points and the ship's real-time image, and match the corresponding lock's anti-collision measures.

[0127] S152: Real-time monitoring of the lock's execution of anti-collision measures via video, and output of the current height of multiple collision contact points. Based on the current height of multiple collision contact points and the shape of the ship, the lock's height difference relative to the ship is determined.

[0128] S153: Based on the height difference, the ship's current navigation position, and the ship's current speed, determine the collision avoidance warning signal of the lock for the ship. The collision avoidance warning signal includes the ship's stop navigation signal, the ship's deceleration navigation signal, and the lock's lifting and lowering control signal.

[0129] In the embodiments of this application, the navigation direction of the ship in the sub-channel is collected, and multiple collision contact positions in the collision risk area are determined based on the collision risk area of ​​the lock, the navigation direction of the ship, and the current navigation position of the ship. The anti-collision event of the lock is determined by comparing the multiple collision contact positions with the real-time image of the ship, and the corresponding anti-collision measures of the lock are matched. This overall consideration of comparing multiple collision contact positions with the real-time image of the ship ensures the accuracy of the anti-collision event of the lock.

[0130] At this time, the ship's navigation direction in the sub-channel is obtained in order to accurately predict its trajectory; the ship's position and heading are monitored in real time using GPS or other positioning technologies; real-time images of the ship are collected using cameras, and the ship's navigation direction is extracted through target detection and trajectory tracking algorithms; optionally, ship 1 is sailing in sub-channel 1, and its current heading is positive (approaching lock 1): Ship 1: Current position: 100 meters from the entrance of lock 1; heading: positive; Ship 2 is sailing in sub-channel 2, and its current heading is positive (approaching lock 2): Ship 2: Current position: 300 meters from the entrance of lock 2; heading: positive.

[0131] Based on the collision risk area, the ship's course, and current position, multiple potential collision contact locations are determined. A geometric method is used to calculate these potential collision contact locations by combining the ship's current position, course, and collision risk area. A Kalman filter or other prediction algorithm is then used to predict the ship's future position. Optionally, the collision risk area for lock 1 is 1490 meters to 1510 meters from the starting point of sub-channel 1: Lock 1: Ship 1's current position: 100 meters from the entrance of lock 1; Course: Forward; Collision contact location: 1500 meters from the starting point of sub-channel 1; The collision risk area for lock 2 is 1990 meters to 2010 meters from the starting point of sub-channel 2: Lock 2: Ship 2's current position: 300 meters from the entrance of lock 2; Course: Forward; Collision contact location: 2000 meters from the starting point of sub-channel 2.

[0132] By comparing the collision contact point with real-time images of the ships, it is determined whether there is a collision risk. The images of the ships are captured in real time using cameras, and the real-time images are compared with the collision contact point to check whether the ships are close to the collision risk area. Optionally, if ship 1 is 1500 meters from the collision contact point, the real-time image shows that the ship is traveling at a relatively high speed: Lock 1: The real-time image shows that the ship is traveling at 10 knots and is close to the collision risk area; if ship 2 is 2000 meters from the collision contact point, the real-time image shows that the ship is traveling at a moderate speed: Lock 2: The real-time image shows that the ship is traveling at 5 knots and is close to the collision risk area.

[0133] Based on the comparison results, it is determined whether a collision avoidance event exists. The collision risk is assessed based on the vessel's speed, position, and collision risk zone. If the vessel is approaching the collision risk zone at a relatively high speed, it is classified as a high-risk event; if the vessel's speed is moderate and its position is within the collision risk zone, it is classified as a medium-risk event. Optionally, Lock 1: Collision Avoidance Event: High Risk; Vessel 1 is at a relatively high speed and approaching the collision risk zone. Lock 2: Collision Avoidance Event: Medium Risk; Vessel 2 is at a moderate speed and approaching the collision risk zone.

[0134] Based on the severity of the collision avoidance event, appropriate collision avoidance measures are matched. These measures include stopping navigation, reducing speed, and adjusting lock height. A preset collision avoidance measure matching table is used to match collision avoidance events with corresponding measures. The collision avoidance measure matching table is shown in Table 7.

[0135] Table 7 Collision Avoidance Measures Matching Table

[0136]

[0137] Furthermore, the system monitors the lock's execution of anti-collision measures in real time and outputs the current height of multiple collision contact points. Based on the current height of multiple collision contact points and the shape of the ship, the system determines the height difference of the lock relative to the ship, taking into account the overall consideration of the current height of multiple collision contact points and the shape of the ship, thus ensuring the accuracy of the height difference of the lock relative to the ship.

[0138] At this time, the process of implementing anti-collision measures in the lock is monitored in real time to ensure the effectiveness of the measures. High-definition cameras are used to collect real-time video of the lock's execution, ensuring that the cameras cover key areas of the lock, such as the gates and lifting mechanisms. The collected video data is transmitted to the central processing unit in real time for storage, so as to facilitate subsequent analysis and playback. Image processing algorithms (such as target detection and trajectory tracking) are used to analyze the video data in real time to detect the execution status of the lock, such as the opening and closing status of the gates and the lifting height. Optionally, the implementation status of anti-collision measures for lock 1 and lock 2 is as follows: Lock 1: Anti-collision measures: The deceleration navigation signal has been issued, and the lock height has been adjusted to 6 meters; the real-time monitoring video shows that the lock height adjustment is complete. Lock 2: Anti-collision measures: The deceleration navigation signal has been issued, and the lock height remains at 5 meters; the real-time monitoring video shows that the lock height remains unchanged.

[0139] Determine the current height of the collision contact point to calculate the height difference between the lock and the ship. Use lidar or other height measurement equipment to measure the lock height in real time. Extract the lock height information from real-time monitoring video and record the measured height data for subsequent calculations. Optionally, the collision contact point heights are as follows: Lock 1: 1500 meters from the collision contact point, current lock height is 6 meters; Lock 2: 2000 meters from the collision contact point, current lock height is 5 meters.

[0140] Based on the current height of the lock and the shape of the ship, calculate the height difference between the lock and the ship. Using real-time images of the ship, extract the ship's draft. The ship's draft can be obtained through image processing algorithms or pre-recorded data. Calculate the difference between the current height of the lock and the ship's draft: Height difference = Lock height - Ship draft. Optionally, if the draft of ship 1 is 3 meters and the height of lock 1 is 6 meters, then the height difference for lock 1 is 6 meters - 3 meters = 3 meters. If the draft of ship 2 is 3 meters and the height of lock 2 is 5 meters, then the height difference for lock 2 is 5 meters - 3 meters = 2 meters.

[0141] Therefore, based on the height difference, the ship's current navigation position, and the ship's current speed, the lock's anti-collision warning signal is determined. This anti-collision warning signal includes the ship's stop navigation signal, the ship's deceleration navigation signal, and the lock's lifting and lowering control signal. It takes into account the overall consideration of the height difference, the ship's current navigation position, and the ship's current speed, ensuring the accuracy of the lock's anti-collision warning signal. At the same time, it introduces the prediction of the lock's collision risk area and realizes the lock's real-time prediction. It takes into account the overall consideration of the lock's collision risk area and the lock's real-time image, improving the effectiveness of the lock's anti-collision measures. It also makes full use of the lock's anti-collision warning signal to further improve the lock's anti-collision warning effect and avoids the existence of an actual collision area between the lock and the ship.

[0142] At this point, considering the height difference, the ship's current position, and speed, the collision risk is assessed. Based on the height difference between the lock and the ship, the existence of a collision risk is evaluated. For example, if the height difference is less than a safety threshold (e.g., 1 meter), a high risk is considered. Combining the ship's current position and speed, the time it would take for the ship to reach the collision risk area is assessed. If the ship is traveling at a high speed and is approaching the collision risk area, emergency measures are required. Optionally, the height difference, position, and speed of ship 1 and ship 2 are as follows: Ship 1: Height difference: 3 meters; Current position: 100 meters from the entrance of lock 1; Speed: 10 knots; Ship 2: Height difference: 2 meters; Current position: 300 meters from the entrance of lock 2; Speed: 5 knots.

[0143] Based on the assessment results, corresponding collision avoidance warning signals are generated. These signals include stop navigation signals, deceleration signals, and lock lifting / lowering control signals. A pre-set warning signal matching table is collected, and the appropriate warning signal is selected according to different risk levels. This warning signal matching table is shown in Table 8.

[0144] Table 8 Early Warning Signal Matching Table

[0145]

[0146] Based on the warning signal matching table, a specific anti-collision warning signal is generated. According to the height difference, ship speed, and position, the warning signal matching table is consulted to generate the corresponding warning signal, which is then sent to the ship and the lock control system. Optional examples: Ship 1: Height difference: 3 meters; Speed: 10 knots; According to the warning signal matching table, the height difference is greater than 2 meters, and the speed is relatively high, so a warning signal of "maintain current speed, maintain lock height unchanged" is generated. Ship 2: Height difference: 2 meters; Speed: 5 knots; According to the warning signal matching table, the height difference is between 1 and 2 meters, and the speed is moderate, so a warning signal of "decelerate and maintain lock height unchanged" is generated.

[0147] Please see Figure 7 , Figure 7 This is a schematic diagram of the structural composition of a vision-based collision avoidance warning system for a ship lock according to an embodiment of the present invention; the vision-based collision avoidance warning system for a ship lock includes:

[0148] The visual detection area module 21 is used to determine the sub-channel and the corresponding lock based on channel A, and to determine the visual detection area based on the lock and the corresponding sub-channel.

[0149] The status module 22 is used to determine the dynamic interaction image between the lock and the ship based on the detection of the visual detection area, and to determine the current status of the lock and the current status of the ship based on the dynamic interaction image.

[0150] The early warning module 23 is used to determine a multi-gradient early warning distribution map of the lock based on the relative distance between the lock and the ship, the current state of the lock and the current state of the ship. The multi-gradient early warning distribution map marks multiple early warning distribution points and corresponding early warning items.

[0151] The prediction module 24 is used to determine the predicted state map of the lock based on the prediction of the warning items at each warning distribution point, determine the state delay of the lock based on the predicted state map and the real-time state map of the lock, and predict the collision risk area of ​​the lock based on the state delay and the location of the lock.

[0152] The collision avoidance warning signal module 25 is used to determine the collision avoidance measures of the lock based on the collision risk area of ​​the lock and the real-time image of the ship, and to determine the collision avoidance warning signal of the lock to the ship based on the video of the lock's execution of the collision avoidance measures and the ship's current speed.

[0153] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for collision avoidance warning of a ship lock based on visual detection, characterized by, The method comprises the steps of: determining sub-channels and corresponding ship locks based on A channel, determining a visual detection area based on the ship lock and the corresponding sub-channel; determining a dynamic interaction image between the ship lock and the ship according to the detection of the visual detection area, determining the current state of the ship lock and the current state of the ship based on the dynamic interaction image; the dynamic interaction image shows the interactive action event of the ship lock and the ship at the same time point; determining a multi-gradient warning distribution map of the ship lock based on the relative distance between the ship lock and the ship, the current state of the ship lock and the current state of the ship, the multi-gradient warning distribution map marking a plurality of warning distribution points and corresponding warning items; in each warning distribution point, determining a predicted state map of the ship lock based on the prediction of the warning item, determining a state delay amount of the ship lock based on the predicted state map and a real-time state map of the ship lock, predicting a collision risk area of the ship lock based on the state delay amount and the position of the ship lock, the state delay amount being a time delay amount or a height difference amount; the collision risk area is marked with a high-risk area and a low-risk area; determining a collision avoidance measure of the ship lock based on the collision risk area of the ship lock and a real-time image of the ship, determining a collision avoidance warning signal of the ship lock to the ship according to a video of the ship lock executing the collision avoidance measure and a current speed of the ship.

2. The visual detection-based collision-prevention warning method for a ship lock according to claim 1, characterized by, The method comprises the steps of: collecting the position of A channel, and determining the channel area corresponding to A channel according to the area retrieval of the position of A channel, and outputting the area distribution map of the channel area corresponding to A channel; determining sub-channels and corresponding ship locks based on A channel, determining a visual detection area based on the ship lock and the corresponding sub-channel; collecting the channel form of the sub-channel, and determining the visual detection area based on the channel form of the sub-channel, the channel distance of the sub-channel and the spatial position of the ship lock in the sub-channel.

3. The visual detection-based collision-prevention warning method for a ship lock according to claim 1, characterized by, The method comprises the steps of: collecting the visual detection area, and performing real-time detection on the visual detection area, collecting a plurality of images according to the real-time detection of the visual detection area, the plurality of images comprising a current image of the ship lock, a current image of the ship and a dynamic image between the ship lock and the ship; determining a dynamic interaction image between the ship lock and the ship according to the synthesis of the plurality of images, the dynamic interaction image showing the interactive action event of the ship lock and the ship at the same time node; determining the action content of the ship lock and the action content of the ship according to the detection of the interactive action event, collecting the position of the ship relative to the sub-channel, and determining the current state of the ship lock and the current state of the ship according to the position of the ship relative to the sub-channel, the action content of the ship and the action content of the ship.

4. The visual detection-based collision-prevention warning method for a ship lock according to claim 1, characterized by, The method comprises the steps of: Collect the relative distance between the ship lock and the ship, determine the control area of the ship lock relative to the ship based on the relative distance and the distribution map corresponding to the sub-channel, and determine a plurality of early warning distribution nodes according to the division of the control area.

5. The visual detection-based collision-prevention warning method for a ship lock according to claim 4, characterized by, The multi-gradient early warning distribution map of the ship lock is determined based on the relative distance between the ship lock and the ship, the current state of the ship lock, and the current state of the ship. The multi-gradient early warning distribution map marks a plurality of early warning distribution points and corresponding early warning items, and further comprises: In each early warning distribution node, a first gradient early warning dynamic map is determined based on the position of the early warning distribution node, the distance between the two adjacent early warning distribution nodes, and the current state of the ship lock, and a second gradient early warning dynamic map is determined based on the position of the early warning distribution node, the distance between the two adjacent early warning distribution nodes, and the current state of the ship; The multi-gradient early warning distribution map of the ship lock is determined based on the synthesis of the first gradient early warning dynamic map and the second gradient early warning dynamic map. At this time, the early warning item corresponding to the early warning distribution point is determined according to the position of the early warning distribution node, the current state of the ship lock, and the current state of the ship. The multi-gradient early warning distribution map marks each early warning item.

6. The visual detection-based collision-prevention warning method for a ship lock according to claim 1, characterized by, In each early warning distribution point, the predicted state map of the ship lock is determined based on the prediction of the early warning item, the state delay of the ship lock is determined based on the predicted state map and the real-time state map of the ship lock, and the collision risk area of the ship lock is predicted based on the state delay and the position of the ship lock, comprising: Real-time monitoring of each early warning distribution point, and determining the regulation and control event of the ship lock based on the analysis of the early warning item, determining the predicted state map of the ship lock based on the current state of the ship lock and the prediction of the regulation and control event, and the predicted state map of the ship lock as the theoretical state map of the ship lock in the early warning distribution point.

7. The visual detection-based collision-prevention warning method for a ship lock according to claim 6, characterized by, In each early warning distribution point, the predicted state map of the ship lock is determined based on the prediction of the early warning item, the state delay of the ship lock is determined based on the predicted state map and the real-time state map of the ship lock, and the collision risk area of the ship lock is predicted based on the state delay and the position of the ship lock, further comprising: If the ship passes through the early warning distribution point during the navigation process in the sub-channel, the state detection of the ship lock is triggered at the same time node to collect the real-time state map of the ship lock, and the state delay of the ship lock is determined according to the comparison between the predicted state map and the real-time state map of the ship lock. The state delay of the ship lock presents the height difference of the ship lock under height adjustment; According to the state delay and the current position of the ship, a first sub-collision risk area of the ship lock is determined, according to the state delay and the position of the ship lock, a second sub-collision risk area of the ship lock is determined, and based on the synthesis of the first sub-collision risk area and the second sub-collision risk area of the ship lock, the collision risk area of the ship lock is determined.

8. The visual detection-based collision-prevention warning method for a ship lock according to claim 1, characterized by, The anti-collision measure of the ship lock is determined based on the collision risk area of the ship lock and the real-time image of the ship, and the anti-collision warning signal of the ship lock to the ship is determined according to the execution video of the anti-collision measure of the ship lock and the current speed of the ship, comprising: The ship's sailing direction in the sub-channel is collected, a plurality of collision contact positions in the collision risk area of the ship lock are determined according to the collision risk area of the ship lock, the sailing direction of the ship and the current sailing position of the ship, a collision avoidance event of the ship lock is determined according to the comparison between the plurality of collision contact positions and the real-time image of the ship, and the corresponding collision avoidance measure of the ship lock is matched.

9. The visual detection-based collision-prevention warning method for a ship lock according to claim 8, characterized by, The collision avoidance measure of the ship lock is determined based on the collision risk area of the ship lock and the real-time image of the ship, the collision avoidance warning signal of the ship lock to the ship is determined according to the execution video of the collision avoidance measure by the ship lock and the current speed of the ship, and the collision avoidance warning system of the ship lock further comprises: The execution video of the collision avoidance measure by the ship lock is monitored in real time, and the current height of the plurality of collision contact positions is output, the height difference of the ship lock relative to the ship is determined according to the current height of the plurality of collision contact positions and the shape of the ship; The collision avoidance warning signal of the ship lock to the ship is determined according to the height difference, the current sailing position of the ship and the current speed of the ship, and the collision avoidance warning signal comprises a stop sailing signal of the ship, a slow sailing signal of the ship and a lifting control signal of the ship lock.

10. A visual detection based anti-collision warning system for a ship lock, characterized by The collision avoidance warning system of the ship lock based on visual detection is applied to the collision avoidance warning method of the ship lock based on visual detection as claimed in any one of claims 1-9.

Images