Method and system for multi-source obstacle perception and operation trajectory planning of a trailing suction dredger

Abstract

The present application relates to the technical field of drag suction dredger, and discloses a method and system for multi-source obstacle perception and operation trajectory planning of drag suction dredger, the method comprising: acquiring multi-source sensor sampling data and environment images at a set monitoring time, constructing an obstacle perception model and an operation trajectory model and correlating output results; collecting multi-source sensor data and environment images of a current operation area in real time, identifying obstacle positions and operation trajectory information through the model; acquiring ship state data, generating obstacle distribution information and a set operation path; and generating a feasible trajectory or an obstacle avoidance trajectory according to path and trajectory matching conditions and safety area information. The system comprises model construction, data collection, information identification and other units. Through multi-source data fusion and dynamic trajectory planning, the present application improves the accuracy of obstacle perception and the rationality of operation trajectory planning of the dredger in a complex environment, thereby ensuring operation safety and efficiency.

Description

Technical Field

[0001] This invention relates to the field of trailing suction hopper dredger technology, specifically to a method and system for multi-source obstacle perception and operation trajectory planning for trailing suction hopper dredgers. Background Technology

[0002] In the field of marine engineering and waterway dredging, trailing suction hopper dredgers are important operational equipment, and their operational safety and efficiency have always been core concerns of the industry. Traditional trailing suction hopper dredgers face challenges such as incomplete obstacle perception and unreasonable operation trajectory planning when operating in complex waters, which seriously restricts their operational capabilities and safety.

[0003] From the perspective of obstacle perception, early dredgers mainly relied on single-type sensors, such as sonar devices alone, for underwater obstacle detection. However, single sensors are susceptible to environmental interference. For example, sonar accuracy drops significantly in turbid waters or environments with a large amount of suspended particles, easily missing small obstacles or misjudging their locations. Furthermore, traditional perception methods lack comprehensive utilization of environmental images, failing to visually assess the obstacle distribution in the work area, making it difficult for operators to fully grasp the working environment. In addition, the low degree of data fusion between different sensors, with each sensor operating independently and lacking correlation and coordination, prevents the formation of a complete obstacle perception model. This results in insufficient accuracy and reliability of obstacle perception when dredgers operate in complex terrain (such as reef areas or areas with underwater structure debris), easily leading to collisions, equipment damage, and personnel injuries.

[0004] In terms of operational trajectory planning, traditional methods typically rely on pre-defined fixed paths, failing to adequately consider real-time obstacle information and vessel status data. When sudden obstacles or environmental changes occur in the work area, fixed path planning cannot be adjusted in a timely manner, forcing dredgers to suspend operations for manual intervention, severely impacting operational efficiency. Furthermore, traditional trajectory planning lacks an effective correlation between obstacle perception models and operational trajectory models, making it impossible to dynamically optimize operational paths based on real-time obstacle distribution. For example, when encountering dense obstacle clusters, traditional methods may fail to accurately calculate safe passage distances and angles, resulting in operational paths that are either overly conservative, increasing unnecessary travel distance and time costs, or overly aggressive, posing a collision risk. In addition, the lack of real-time analysis of vessel status data (such as vessel attitude, speed, and draft) prevents adjustments to trajectory planning based on the actual vessel condition, further reducing operational adaptability and safety.

[0005] With the deepening of marine development, trailing suction hopper dredgers face increasingly complex operating environments, placing higher demands on obstacle perception and trajectory planning. On the one hand, it is necessary to integrate data from multiple sensors to achieve multi-dimensional and all-round perception of obstacles in the operating area, improving the accuracy and reliability of obstacle detection. On the other hand, it is necessary to establish a dynamic trajectory planning model to adjust the operating path in real time based on real-time perceived obstacle information and hull status data, ensuring the safe and efficient operation of the dredger in complex environments. However, existing technologies lack a method and system that can effectively integrate multi-source sensor data, establish a correlation between obstacle perception models and trajectory models, and dynamically plan the trajectory based on real-time data, making it difficult to meet the needs of modern trailing suction hopper dredgers operating in complex waters. Therefore, a new method and system for multi-source obstacle perception and trajectory planning for trailing suction hopper dredgers is urgently needed to solve the problems existing in traditional technologies and improve the operational safety and efficiency of dredgers. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for multi-source obstacle perception and operation trajectory planning for trailing suction hopper dredgers, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for multi-source obstacle perception and operation trajectory planning of a trailing suction hopper dredger, the method comprising: Step S1: Simultaneously acquire sampling data from multiple source sensors and corresponding environmental images at multiple set monitoring times; process the sampling data from the multiple source sensors using a data processing algorithm to obtain an obstacle perception model; configure the environmental images to obtain a work trajectory model; associate the output results of the obstacle perception model and the work trajectory model according to the monitoring time; the multiple source sensors include a sonar device, a camera, and a positioning device; Step S2: Acquire multi-source sensor data of the current work area and simultaneously collect the current environmental image while acquiring the multi-source sensor data of the current work area; Step S3: Obtain the current obstacle location information by identifying the current multi-source sensor data through the obstacle perception model; obtain the current operation trajectory information by identifying the current environmental image through the operation trajectory model; Step S4: Obtain the hull status data of the dredger at its current position; Step S5: Obtain obstacle distribution information in the current work area based on the hull status data; Step S6: Obtain the set operation path based on the obstacle distribution information; Step S7: If the set operation path matches the current operation trajectory information, then obtain the set safe area information according to the correlation between the output results of the obstacle perception model and the operation trajectory model; determine whether the current obstacle position information conflicts with the set safe area information. If yes, generate feasible trajectory data; otherwise, generate obstacle avoidance trajectory data.

[0008] Preferably, step S2 includes: Step S21: Install a sensor receiving module on the hull of the dredger; the sensor receiving module is connected to the input and output terminals of the control unit; Step S22: Sensor arrays are installed at the front, middle and rear of the hull respectively; the sensor arrays are equipped with multi-source sensors facing outwards; Step S23: An image acquisition module is installed on the top of the hull, facing the work area; the image acquisition module is connected to the input / output terminal of the control unit and outputs the current environmental image to the control unit; Step S24: When the control unit receives a monitoring command, the output of the control unit drives the sensor receiving module to acquire multi-source sensor data; the output of the control unit drives the image acquisition module to acquire the current environmental image.

[0009] Preferably, step S5 includes: acquiring the water depth data of the current working area; acquiring the water depth information of the current working area based on the hull status data and the water depth data; acquiring the obstacle height information of the current working area based on the water depth information; and generating the monitoring command based on the obstacle height information.

[0010] Preferably, step S5 includes: obtaining the distribution of multiple obstacles in the current working area based on the hull status data; obtaining the location information of multiple obstacles and the spacing information of multiple obstacles in the current working area based on the distribution of multiple obstacles in the current working area; and generating the monitoring command based on the location information of multiple obstacles and the spacing information of multiple obstacles.

[0011] Preferably, step S2 includes: Acquire multi-source sensor data for the current work area; the multi-source sensor data includes sonar data, camera data, and positioning data; Determine whether the sonar data, camera data, and positioning data meet the set integrity requirements. If yes, determine whether the sonar data, camera data, and positioning data are within the sensor type and location range based on the sensor type and location range. If yes, acquire the current environmental image simultaneously with the multi-source sensor data of the current work area. If no, return to this step and reacquire the multi-source sensor data of the current work area. Determine whether the sonar data, camera data, and positioning data meet the set integrity requirements; if not, generate a data error message.

[0012] Preferably, step S7 includes: if the set operation path does not match the current operation trajectory information, then the current operation path is obtained through the shipborne navigation system or trajectory planning module, and the process returns to this step to re-determine whether the set operation path matches the current operation trajectory information.

[0013] Preferably, the present invention also includes a multi-source obstacle perception and operation trajectory planning system for a trailing suction hopper dredger, the system comprising: The model building unit is configured to simultaneously acquire sampling data from multiple source sensors and corresponding environmental images at multiple set monitoring times; process the sampling data from the multiple source sensors through a data processing algorithm to obtain an obstacle perception model; configure the environmental images to obtain a work trajectory model; and correlate the output results of the obstacle perception model and the work trajectory model according to the monitoring time; the multiple source sensors include a sonar device, a camera, and a positioning device; The data acquisition unit is configured to acquire multi-source sensor data of the current work area and, while acquiring the multi-source sensor data of the current work area, simultaneously acquire the current environmental image. The information recognition unit is configured to identify the current obstacle location information by using the obstacle perception model to identify the current multi-source sensor data; and to identify the current operation trajectory information by using the operation trajectory model to identify the current environmental image. The status acquisition unit is configured to acquire the hull status data of the dredger at its current position. The distribution acquisition unit is configured to acquire obstacle distribution information of the current working area based on the hull status data. The path acquisition unit is configured to acquire a set operation path based on the obstacle distribution information. The trajectory generation unit is configured to, if the set operation path matches the current operation trajectory information, obtain the set safe area information according to the correlation between the output results of the obstacle perception model and the operation trajectory model; determine whether the current obstacle position information conflicts with the set safe area information; if so, generate feasible trajectory data; if not, generate obstacle avoidance trajectory data.

[0014] Preferably, the data acquisition unit is further configured as follows: A sensor receiving module is installed on the hull of the dredger; the sensor receiving module is connected to the input and output terminals of the control unit; Sensor arrays are installed at the front, middle and rear of the hull; the sensor arrays are equipped with multi-source sensors facing outwards; An image acquisition module is installed on the top of the hull, facing the work area; the image acquisition module is connected to the input / output terminal of the control unit and outputs the current environmental image to the control unit; When the control unit receives a monitoring command, the output of the control unit drives the sensor receiving module to acquire multi-source sensor data; the output of the control unit drives the image acquisition module to acquire the current environmental image.

[0015] Preferably, the distribution acquisition unit is further configured to: acquire water depth data of the current working area; acquire water depth information of the current working area based on hull status data and the water depth data; acquire obstacle height information of the current working area based on the water depth information; generate the monitoring command based on the obstacle height information; or the distribution acquisition unit is further configured to: acquire the distribution of multiple obstacles in the current working area based on hull status data; acquire the location information of multiple obstacles and the spacing information of multiple obstacles in the current working area based on the distribution of multiple obstacles in the current working area; generate the monitoring command based on the location information of multiple obstacles and the spacing information of multiple obstacles.

[0016] Preferably, the data acquisition unit is further configured as follows: Acquire multi-source sensor data for the current work area; the multi-source sensor data includes sonar data, camera data, and positioning data; Determine whether the sonar data, camera data, and positioning data meet the set integrity requirements. If yes, determine whether the sonar data, camera data, and positioning data are within the sensor type and location range based on the sensor type and location range. If yes, acquire the current environmental image simultaneously with the multi-source sensor data of the current work area. If no, return to this step and reacquire the multi-source sensor data of the current work area. Determine whether the sonar data, camera data, and positioning data meet the set integrity requirements; if not, generate a data error message.

[0017] Compared with the prior art, the beneficial effects of the present invention are: In terms of obstacle perception capabilities, by simultaneously acquiring sampling data from multiple sensors, including sonar, cameras, and positioning devices, along with corresponding environmental images, and utilizing data processing algorithms to construct an obstacle perception model, multi-dimensional and three-dimensional perception of obstacles in the work area is achieved. Sonar effectively detects the location and outline of underwater obstacles, cameras provide intuitive visual image information, and positioning devices ensure accurate obstacle location. The fusion of these three data sources overcomes the limitations of a single sensor. For example, in turbid waters, sonar data may be inaccurate due to water quality, while cameras, by capturing abnormal images of the water surface or shallow water, can help correct sonar detection results and improve the accuracy of obstacle location identification. Simultaneously, by correlating the outputs of the obstacle perception model and the work trajectory model with the monitoring time, a dynamic correlation is formed between obstacle information at different times and the work trajectory. This allows for real-time tracking of obstacle changes, such as the movement trajectory of obstacles driven by water flow, thereby more accurately grasping the dynamic distribution of obstacles in the work area.

[0018] Regarding the scientific rigor and flexibility of the operational trajectory planning, a reasonable operational path is generated by acquiring hull status data (such as hull attitude, speed, and draft) and combining it with obstacle distribution information. The introduction of hull status data makes trajectory planning more closely aligned with the actual operating capabilities of the dredger. For example, based on draft and water depth information, it prevents the dredger from entering areas with insufficient water depth and running aground. When the set operational path matches the current operational trajectory information, it further combines safety zone information to determine whether obstacle positions conflict, thereby generating feasible trajectory data or obstacle avoidance trajectory data. This hierarchical judgment mechanism ensures the rigor of trajectory planning: if the obstacle is outside the safety zone, an obstacle avoidance trajectory is directly generated to avoid collision risks; if it is within the safety zone but does not constitute a conflict, a feasible trajectory is maintained to ensure operational continuity. When the set operational path does not match the current operational trajectory, the onboard navigation system or trajectory planning module re-acquires the operational path and iteratively judges it, achieving dynamic adjustment of trajectory planning. This allows the dredger to adapt to sudden changes in the environment during operation, such as suddenly appearing floating obstacles or temporarily adjusted operational area boundaries.

[0019] Regarding the reliability of data acquisition and processing, sensor arrays are installed at the front, middle, and rear of the hull, and an image acquisition module is installed on the top, achieving comprehensive coverage of the work area. The distributed layout of the sensor arrays ensures 360-degree detection of obstacles around the hull without blind spots, avoiding the detection blind spots of traditional single-point sensors. For example, the sensors at the front of the hull can detect obstacles ahead in advance, while the sensors at the middle and rear can monitor the dynamics of obstacles on both sides and behind the hull in real time, providing comprehensive data support for trajectory planning. Simultaneously, a completeness and validity verification mechanism is added during data acquisition. First, it checks whether the sonar data, camera data, and positioning data meet the set completeness requirements, and then verifies the rationality of the data based on the sensor type and location range, ensuring the accuracy and reliability of the data source input to the model. If the data is incomplete or exceeds a reasonable range, it is promptly re-acquired or error information is generated to avoid errors in obstacle perception and trajectory planning due to data quality issues.

[0020] In terms of system integration and intelligence, the system achieves fully automated processing from data acquisition and model building to trajectory generation through the collaborative work of multiple functional units, including model building, data acquisition, and information recognition. Data interaction and command-driven communication between units are achieved through control units, reducing manual intervention, mitigating the risk of human error, and improving operational response speed. For example, when the control unit receives a monitoring command, it automatically drives the sensor receiving module and image acquisition module to synchronously collect data, eliminating the need for operators to trigger each step individually, thus improving operational efficiency. Furthermore, the system can dynamically generate operational trajectories based on real-time data, giving the dredger a certain degree of autonomous decision-making capability. This adapts to the development trend of intelligent and unmanned marine operation equipment, laying the technological foundation for future automated operations of dredgers. Attached Figure Description

[0021] Figure 1 This is a schematic diagram illustrating the working principle of the multi-source obstacle perception and operation trajectory planning method for trailing suction hopper dredgers described in this invention. Figure 2 A schematic diagram for multi-source sensor data verification; Figure 3 This is a schematic diagram of the trajectory generation logic. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1 - Figure 3 The present invention relates to a method for multi-source obstacle perception and operation trajectory planning of a trailing suction hopper dredger, the specific implementation steps of which are as follows: Step S1: At multiple set monitoring times, sampling data is synchronously collected using multi-source sensors, and corresponding environmental images are acquired using image acquisition equipment. Specifically, the multi-source sensors include a sonar device, a camera, and a positioning device. The sonar device is used to acquire echo data of underwater obstacles, the camera is used to acquire visual images of the work area, and the positioning device (such as GPS) is used to acquire the position coordinates and motion parameters of the dredger. The sampling data from the multi-source sensors is processed using data processing algorithms (such as filtering, feature extraction, and target recognition algorithms) to construct an obstacle perception model. This model can output information such as the position, shape, and size of obstacles based on sensor data. Simultaneously, a work trajectory model is configured by analyzing the environmental images (such as image segmentation, edge detection, and path feature extraction). This model is used to identify feasible work paths in the images. Furthermore, based on the monitoring timestamps, a correlation is established between the output results of the obstacle perception model and the work trajectory model to achieve synchronous correspondence between obstacle information and work trajectory information at the same time point.

[0024] Step S2: During the dredging operation, acquire multi-source sensor data of the current working area in real time, including sonar data, camera data and positioning data, and at the same time as acquiring sensor data, acquire images of the current environment through image acquisition equipment.

[0025] Step S3: Input the current multi-source sensor data into the obstacle perception model, and use the model to calculate and identify the current obstacle location information, including the obstacle's coordinates, range, and type; at the same time, input the current environmental image into the operation trajectory model to identify the current operation trajectory information, including the location, direction, and boundary features of the preset operation path.

[0026] Step S4: Obtain the hull status data of the dredger at its current position, including hull attitude parameters (such as roll angle, pitch angle, and heading angle), sailing speed, propeller status, etc., and collect and transmit the data to the control system in real time through onboard sensors (such as inertial navigation system and speedometer).

[0027] Step S5: Based on the hull status data and water depth data collected by sensors, analyze the obstacle distribution information in the current working area. Specifically, by comparing the water depth data with the hull's draft, determine the height and distribution range of underwater obstacles, and combine this with the hull attitude parameters to correct the spatial position of the obstacles, forming a complete obstacle distribution map.

[0028] Step S6: Based on the obstacle distribution information, generate a set operation path that avoids obstacles using a path planning algorithm (such as A* algorithm or Dijkstra algorithm) to ensure that a safe distance is maintained between the path and the obstacles.

[0029] Step S7: Determine whether the set operation path matches the current operation trajectory information (e.g., whether the path direction and starting point are consistent). If they match, obtain the set safety zone information (i.e., the safe operation range associated with the obstacle perception model and the operation trajectory model) based on the association established in Step S1; further determine whether the current obstacle location information exceeds the safety zone range. If there is a conflict (i.e., the obstacle is within the safety zone), generate feasible trajectory data to bypass the obstacle through the trajectory optimization algorithm; if there is no conflict, generate obstacle avoidance trajectory data along the original path. If the set operation path does not match the current operation trajectory information, re-obtain the current operation path through the shipborne navigation system or trajectory planning module, and return to Step S7 to re-perform the matching judgment.

[0030] The present invention will be further described below with reference to Examples 1 to 5: Example 1: A sensor receiving module is installed on the hull of the dredger. This module serves as a data transmission hub between the multi-source sensors and the control unit. It connects to the input and output terminals of the control unit via wired communication lines (such as shielded cables) or wireless communication links (such as Wi-Fi and Bluetooth) to ensure stable transmission of sensor data and real-time issuance of control commands. The sensor receiving module has signal conditioning functions, which can preprocess the raw signals output by the multi-source sensors. For example, it can amplify the weak electrical signals output by the sonar device, perform preliminary compression of the video stream data transmitted by the camera, and convert the format of the coordinate data output by the positioning device to adapt to the input requirements of the control unit.

[0031] Sensor arrays are installed at the bow, midships, and stern of the dredger, with the installation positions optimized based on the dredger's structural characteristics and operational requirements. The bow sensor array is typically positioned on the bow deck or mast, facing directly forward and to the sides at a certain angle (e.g., 30 degrees to each side), used to detect obstacles in front of the work area, such as underwater reefs, shipwrecks, or other vessels. The midships sensor array is installed between the hulls on both sides (port and starboard), facing the open sea, primarily used to monitor obstacles on the sides of the hull, such as shoals near the channel edge, floating debris, or aquaculture facilities. The stern sensor array is located on the stern deck or aft structure, facing the stern, used to detect obstacles behind the stern to prevent collisions during reversing or turning. Each sensor array contains a multi-source sensor, namely an integrated arrangement of sonar devices, cameras, and positioning devices: the sonar devices use transceiver transducers and are installed at the bottom or front of the array to ensure that sound waves can effectively penetrate the water surface and enter the underwater environment; the cameras are waterproof and dustproof industrial-grade cameras equipped with automatic aperture and zoom lenses, and are installed in the middle or upper part of the array to obtain clear images of the work area; the antenna part of the positioning device is installed at the top of the array to ensure stable reception of satellite signals (such as GPS and Beidou signals) and obtain accurate position coordinates.

[0032] An image acquisition module is installed on the top of the hull (such as the top of the cockpit or the top of the mast). This module is fixedly installed facing the work area, and its field of view covers the main work areas in front of and to the sides of the dredger. The image acquisition module can select different types of cameras depending on the working environment. For example, a high-definition color camera is used in environments with sufficient visible light, while an infrared camera or a low-light camera is switched to in nighttime or low-light environments to ensure clear environmental images are captured under different lighting conditions. The image acquisition module is connected to the control unit via a video transmission cable (such as an HDMI cable or an SDI cable) or a network transmission protocol (such as RTSP or ONVIF) to transmit the real-time acquired image data to the control unit for storage and processing. To prevent damage to the image acquisition module in severe weather (such as heavy rain or strong winds), it is equipped with a protective outer shell. The shell is waterproof, shockproof, and resistant to salt spray corrosion. An internal temperature control system is also installed inside the shell to maintain the internal operating temperature of the module within a suitable range (such as -20℃ to 60℃) through heating or heat dissipation devices.

[0033] When the control unit receives a monitoring command, it triggers the data acquisition process. Monitoring commands can be categorized into two types: timed triggers and event triggers. Timed triggers are generated by a timer within the control unit, such as automatically initiating data acquisition every 10 or 30 seconds. Event triggers are generated by external signals (such as manual triggering by an operator through a human-machine interface, or abnormal swaying detected by the ship's attitude sensors) or system logic (such as a potential risk predicted by an obstacle perception model). The control unit's output sends a drive signal to the sensor receiving module, which includes parameters such as the acquisition time window and sensor operating mode. After receiving the drive signal, the sensor receiving module sequentially activates the multi-source sensors in each sensor array: First, it starts the sonar device, which emits sound waves and receives echo signals. The operating frequency of the sonar device is adjusted according to the detection depth and resolution requirements. For example, high-frequency sonar (e.g., 200kHz) is used in shallow water to obtain high-resolution images, while low-frequency sonar (e.g., 30kHz) is used in deep water to expand the detection range. Next, the camera is activated, automatically adjusting parameters such as exposure time and gain according to the current lighting conditions, and begins acquiring video streams or still images. Simultaneously, the positioning device is activated, entering a high-precision positioning mode and outputting positioning data containing information such as longitude, latitude, altitude, speed, and heading. During the sensor data acquisition process, the sensor receiving module caches the data in real time and adds timestamps accurate to the millisecond level to ensure the time synchronization between the multi-source sensor data and the environmental images.

[0034] Simultaneously, the control unit's output sends a capture command to the image acquisition module, containing parameters such as image resolution and capture frequency. Upon receiving the command, the image acquisition module adjusts the lens focal length and shooting angle to begin capturing images of the current environment. To avoid image data redundancy, dynamic frame rate technology can be used, i.e., reducing the capture frequency (e.g., 5 frames / second) when the operating environment is stable, and increasing the capture frequency (e.g., 30 frames / second) when obstacles or changes in the work trajectory are detected. The acquired image data is compressed in real time using compression algorithms (e.g., H.264, H.265) to reduce data transmission bandwidth and storage space usage.

[0035] After completing a data acquisition, the sensor receiving module and image acquisition module send a data transmission request to the control unit. Upon receiving the request, the control unit reads the multi-source sensor data cached by the sensor receiving module and the environmental image data transmitted by the image acquisition module through the input port, and stores the data in system memory or an external storage device (such as a solid-state drive). During data storage, the control unit performs preliminary data verification, such as checking whether the echo intensity of the sonar data is within a reasonable range, whether there is abnormal noise in the pixel values ​​of the camera image, and whether the coordinates of the positioning data exceed the working area. If data anomalies are found, the control unit will mark the batch of data and remove it or trigger a re-acquisition process in subsequent processing.

[0036] To ensure the long-term stable operation of the multi-source sensors, both the sensor array and the image acquisition module are equipped with self-testing mechanisms. For example, the sonar device performs periodic self-tests by emitting calibration sound waves and analyzing the amplitude and phase of the echo signals to determine whether the transducer is working properly; the camera performs periodic autofocus and white balance calibration by shooting standard test charts and adjusting lens parameters to ensure image clarity and color accuracy; the positioning device periodically synchronizes with satellite signals to verify whether the positioning accuracy meets system requirements. The self-test results are fed back to the control unit through the sensor receiving module. When a device malfunction or performance degradation is detected, the control unit alerts the operator to perform maintenance or replacement through an alarm system (such as audible and visual alarms, SMS notifications).

[0037] During dredging operations, the sensor receiving module and image acquisition module remain continuously on standby, responding to monitoring commands from the control unit at any time. By deploying sensor arrays at the front, middle, and rear, combined with the top image acquisition module, a comprehensive perception network covering the hull and the work area is formed, ensuring no blind spots. The synchronous acquisition and timestamping of multi-source sensor data and environmental images provide high-precision spatiotemporal aligned data for subsequent obstacle perception and work trajectory models, guaranteeing the accuracy and reliability of model analysis.

[0038] Example 2: During dredging operations, water depth measurement equipment (such as echo sounders) installed on the bottom of the hull or keel is used to acquire real-time water depth data for the current working area. This equipment calculates the vertical distance from the water surface to the seabed or the top of an obstacle by emitting sound waves and measuring the round-trip time of the sound waves; its measurement accuracy can reach the centimeter level. The acquired water depth data is transmitted to the control unit in the form of an electrical signal. The control unit filters the signal to eliminate noise caused by factors such as hull movement and water current disturbance, ensuring the accuracy of the data.

[0039] Ship status data is collected in real time via inertial navigation systems, attitude sensors, and other equipment, including parameters such as draft, roll angle, heel angle, and heading angle. Draft refers to the vertical distance from the bottom of the hull to the water surface, measured by a draft gauge or pressure sensor mounted on the hull side. Roll and heel angles reflect the degree of the hull's tilt, calculated using accelerometers and gyroscopes in the inertial navigation system. The control unit correlates the collected water depth values ​​with the draft from the ship status data to calculate the actual water depth information for the current operating area. Specifically, the actual water depth equals the collected water depth value minus the ship's draft; this value represents the clearance distance from the bottom of the hull to the seabed or the top of an obstacle. For example, if the collected water depth value is 15 meters and the ship's draft is 8 meters, the actual water depth is 7 meters, indicating that the seabed or the top of an obstacle is 7 meters below the bottom of the hull.

[0040] After acquiring the actual water depth information, the control unit compares and analyzes it with the safe operating water depth threshold of the dredger. The safe operating water depth threshold is preset based on factors such as the structural strength of the hull, the size of the propeller, and the working requirements of the dredging equipment. It is typically greater than or equal to the hull's draft plus a certain safety margin (e.g., 2 meters). If the actual water depth is less than the safe operating water depth threshold, it indicates the presence of an obstacle higher than the safe depth, potentially posing a collision risk to the hull or operating equipment. In this case, the control unit calculates the obstacle's height based on the difference between the actual water depth and the safe operating water depth threshold. For example, if the safe operating water depth threshold is 10 meters and the actual water depth is 7 meters, then the obstacle height is the safe operating water depth threshold minus the actual water depth, which is 3 meters, indicating the presence of an obstacle with a height of 3 meters in the area.

[0041] Obstacle height information is used to generate targeted monitoring commands to adjust the operating modes and parameters of the multi-source sensors. Specifically, when an obstacle height is detected to exceed a set threshold (e.g., 1 meter), the control unit sends commands to the sensor receiving module to adjust the scanning frequency and beam angle of the sonar device. For example, the sonar scanning frequency is increased from the usual 1 scan / second to 2 scans / second to increase the monitoring density of the obstacle area; simultaneously, the sonar beam angle is switched from a wide beam (e.g., 120 degrees) to a narrow beam (e.g., 30 degrees) to improve the accuracy of obstacle boundary and shape recognition. For the camera, the control unit commands the image acquisition module to adjust the lens focal length, perform local magnification of the obstacle area, and enhance the image contrast and sharpness to more clearly identify the visual features of the obstacle (e.g., color, texture, outline). The positioning device then enters a high-precision differential positioning mode, improving the positioning accuracy from meter-level to centimeter-level by receiving differential signals from ground-based augmentation stations or satellites, ensuring the accuracy of the obstacle's position coordinates.

[0042] Furthermore, the control unit marks high-risk areas in the operation trajectory model based on obstacle height information. The operation trajectory model generates the permissible trajectory range for dredging vessel operations by analyzing information such as water depth contour lines and channel boundaries in environmental images. When obstacle height information indicates the presence of a high-risk obstacle in a certain area, the operation trajectory model automatically removes that area from the feasible trajectory and enforces a safe distance when planning the path. For example, if the obstacle height is 3 meters and the set safe distance is twice the obstacle height (i.e., 6 meters), the operation trajectory model ensures that the distance between the vessel's center and the obstacle edge is no less than 6 meters when generating the path.

[0043] In processing water depth and obstacle height information, the control unit also needs to consider the correction of the measurement results by the ship's attitude parameters. For example, when the ship rolls, the water depth data collected on both sides of the ship will deviate due to the ship's tilt. The control unit calculates the actual vertical depth of the ship using the roll angle and corrects the water depth data accordingly. The specific method is as follows: based on the roll angle... Given the hull width B, calculate the actual water depth correction value on one side of the hull. ,in These are the original water depth values. Attitude correction ensures the accuracy of water depth and obstacle height information, avoiding misjudgments caused by ship swaying.

[0044] The control unit displays obstacle height information and corresponding monitoring commands to the operator in real time via a human-machine interface. The operator can manually adjust monitoring parameters or intervene in the trajectory planning process as needed. For example, if the operator finds a complex obstacle distribution in a certain area, they can use interface commands to further increase the sensor's scanning frequency or expand the image acquisition range. Simultaneously, the system logs the obstacle height information and the execution process of monitoring commands, including data acquisition time, obstacle location, and sensor parameter adjustment records, for later review and fault analysis.

[0045] To meet the needs of water depth measurement in different operating environments, the water depth measurement equipment can be configured with multiple operating modes, such as single-beam echo sounding and multi-beam echo sounding. In conventional operating areas, single-beam echo sounding mode is used to cover a large area at a high measurement speed. In areas with dense obstacles or complex terrain, the system switches to multi-beam echo sounding mode, which transmits multiple beams to form a fan-shaped coverage, acquiring more detailed seabed topographic data and thus more accurately calculating the height and distribution range of obstacles. After the multi-beam echo sounding data is transmitted to the control unit through the sensor receiving module, the control unit uses digital signal processing technology to synthesize the data and generate a three-dimensional point cloud model of the seabed topography, providing a more intuitive basis for obstacle height analysis.

[0046] Example 3: During dredging operations, the vessel's status data is collected in real time through equipment such as inertial navigation systems, speed sensors, and steering angle sensors, including the vessel's speed. Heading angle angular velocity of steering Location coordinates These parameters are transmitted to the control unit in time-series format. The control unit analyzes the changing trends of the ship's trajectory to determine whether there are multiple obstacles in the work area. For example, a sudden decrease in the ship's speed, frequent changes in the heading angle, or a turning angular velocity exceeding a set threshold may indicate the presence of a group of obstacles ahead that need to be avoided.

[0047] The control unit combines multi-source sensor data (such as sonar echo signals, camera image features, and positioning coordinates) to perform cluster analysis on obstacles in the work area, identifying multiple independent obstacle segments. Specifically, the sonar device generates point cloud data of obstacles by emitting sound waves and receiving echoes, with each point cloud data point containing distance information. Azimuth Pitch angle Information such as: The camera uses image segmentation algorithms (such as the watershed algorithm and the mean shift algorithm) to separate the obstacle target from the background in the image, extracting the outline features and pixel coordinates of the obstacle; The positioning device assigns geographic coordinates to each obstacle detection result. The control unit performs spatiotemporal registration of multi-source sensor data, unifying it to the same coordinate system. Then, it uses a density clustering algorithm (such as the DBSCAN algorithm) to cluster the obstacle detection results, dividing obstacles with similar spatial locations and consistent motion states into the same "segment", thus forming a multi-segment obstacle distribution.

[0048] For each obstacle segment, the control unit extracts its location information, including the starting coordinates of that obstacle segment. End point coordinates Center coordinates And spatial extent (such as length, width, and height). For linearly distributed obstacles (such as submarine pipelines and breakwaters), the coordinates of the starting and ending points are used to describe their direction of extension and length; for irregularly shaped groups of obstacles (such as reefs), the center coordinates and spatial extent are used to characterize their overall location and occupied area.

[0049] The distance between two adjacent obstacle segments is obtained by calculating the Euclidean distance between the center coordinates of the two obstacle segments, using the following formula:

[0050] in, Indicates the distance between two adjacent obstacles. and These are the center coordinates of the two obstacles. This formula calculates the straight-line distance using three-dimensional spatial coordinates, avoiding distance deviations caused by the ship's attitude or sensor perspective.

[0051] The control unit determines the distribution characteristics of the obstacle group based on the position and spacing information of multiple obstacle segments. If the spacing... Greater than the set safety distance threshold (e.g., 100 meters), then these are considered dispersed obstacles, and work paths can be planned to traverse through the gaps; if the gaps are... Less than or equal to If three obstacles are detected to be arranged linearly with a spacing of less than 50 meters, the control unit will mark the area as a "high-risk dense obstacle area" and generate a detour path in the work trajectory model.

[0052] Based on the distribution characteristics of multiple obstacle segments, the control unit generates corresponding monitoring commands to optimize the sensor operating mode. For scattered obstacles, the control unit instructs the sensor array to perform targeted scanning of each obstacle segment: the front sensor array focuses on the obstacle segment in front, adjusts the sonar beam direction to align with the center coordinates of the obstacle, and simultaneously controls the camera to zoom and capture detailed features of the obstacle (such as surface texture and signs of movement); the middle and rear sensor arrays maintain a conventional scanning mode to monitor environmental changes on both sides and behind the hull. For areas with dense obstacles, the control unit activates a multi-sensor collaborative monitoring mode: the sonar device switches to a narrow beam, high-frequency scanning mode to improve the resolution of obstacle boundaries; the camera activates the panoramic stitching function to obtain an overall image of the obstacle group through multi-frame image synthesis; the positioning device, in conjunction with the inertial navigation system, performs real-time dynamic tracking of the obstacle group's position to ensure coordinate accuracy better than 0.5 meters.

[0053] The control unit overlays virtual markers onto the environmental image based on the location information of multiple obstacle segments. For example, it uses rectangles of different colors to mark the extent of each obstacle segment and arrows to indicate the direction of their extension, assisting operators in visually judging the obstacle distribution. Operators can manually select an obstacle segment through the human-machine interface to trigger a detailed information query function, obtaining parameters such as the obstacle's type (e.g., fixed obstacle, floating obstacle), material (e.g., metal, rock), and size. These parameters are obtained through sonar echo characteristic analysis (e.g., echo intensity reflecting material density) and image recognition algorithms (e.g., convolutional neural network classification).

[0054] To adapt to potential dynamic changes in obstacles (such as the movement of floating objects), the control unit continuously monitors the position and spacing of multiple obstacle segments, analyzing their movement trajectories using data from adjacent timestamps. For example, if the position coordinates of an obstacle segment change more than a set threshold (e.g., 1 meter / minute) within three consecutive monitoring cycles, it is identified as a moving obstacle. The control unit automatically adjusts the monitoring frequency to once per second and initiates a prediction algorithm (such as Kalman filtering) to estimate its future position, providing dynamic data support for trajectory planning.

[0055] During the generation of monitoring commands, the control unit also needs to consider the sensor's detection range limitations. For example, if the maximum detection range of the forward sonar is 200 meters, and an obstacle is located 180 meters from the hull, the control unit will instruct the sonar device to enter high-gain mode in advance to enhance the sensitivity of the echo signal reception. If the obstacle is beyond the sensor's detection range, supplementary information will be obtained through shipborne radar or communication with surrounding ships and shore-based monitoring systems to form a cross-device collaborative sensing network.

[0056] The information on the distribution of multiple obstacle segments is also used to optimize the path planning logic of the operation trajectory model. Based on the location and spacing of obstacle segments, the operation trajectory model transforms the path planning problem into a shortest path problem in graph theory: starting from the current position of the dredger and ending at the target work point, each obstacle segment is considered a "node" in the graph, and the gaps or safe detour areas between obstacles are considered "edges." The shortest and safest path is searched using Dijkstra's algorithm or A* algorithm. During the search process, edges between obstacle segments with smaller gaps are assigned higher weights (i.e., detour costs), thus guiding the algorithm to prioritize gaps or outer paths with larger gaps.

[0057] The control unit records the analysis results of the multi-segment obstacle distribution and the execution of monitoring commands in real time, forming an obstacle monitoring log. The log includes: monitoring time, obstacle segment number, location coordinates, spacing values, sensor parameter adjustment records, trajectory planning results, etc. This log can be used for post-operation review and analysis, such as evaluating the effectiveness of sensor configuration strategies, optimizing obstacle clustering algorithm parameter settings, or providing historical reference data for trajectory planning in similar subsequent operating environments.

[0058] Example 4: This embodiment describes the specific process for verifying the integrity of multi-source sensor data on a dredger, including integrity judgment of sonar data, camera data, and positioning data, verification of sensor type and location range, and handling mechanisms for data anomalies, as detailed in the specific application scenario below: Before or during dredging operations, the data acquisition unit initiates a multi-source sensor data acquisition process. Taking a port dredging operation as an example, after the dredging vessel enters the designated work area, the control unit first triggers the sensor receiving module to collect sonar data, camera data, and positioning data for the current area. The sonar data includes information such as the echo intensity, timestamp, and beam angle of underwater obstacles; the camera data consists of real-time video streams or still images of the work area; and the positioning data includes parameters such as longitude, latitude, altitude, timestamp, and positioning accuracy factor (PDOP) based on the BeiDou satellite system.

[0059] Perform data integrity verification. The control unit checks the field integrity of three types of data: Sonar data must include at least five key fields: echo intensity (reflecting obstacle material and distance), timestamp (accurate to milliseconds), beam angle of incidence, measurement depth, and sensor ID (identifying front, middle, or rear sonar). If a sonar data entry is missing the "beam angle of incidence" field, or if the timestamp format does not conform to the ISO 8601 standard (e.g., lacking time zone information), it is considered incomplete data.

[0060] Camera data: must include image resolution (e.g., 1920×1080), frame rate (e.g., 30fps), exposure time, gain value, and the image data itself. If more than 5 consecutive frames in the video stream show "frame rate 0" or "image data empty", or if the pixel values ​​of the still image are all 255 (pure white noise), the data is considered incomplete.

[0061] Positioning data must include longitude (within ±180°), latitude (within ±90°), altitude (relative to mean sea level), positioning time, number of satellites (≥4), and PDOP value (≤6). If the number of satellites is less than 3 or the PDOP value is greater than 10 (indicating poor positioning accuracy), the data is considered incomplete.

[0062] Taking a specific data acquisition as an example: the sonar data lacked the "sensor ID" field, causing the control unit to be unable to identify which part of the ship's sonar array the data originated from; the camera data showed large areas of pixelation in 10 consecutive frames due to a transmission link failure; and the "altitude" field of the positioning data was negative and exceeded the geographical range of the work area (the average altitude of this area is 5 meters). All three types of data were deemed to fail to meet integrity requirements. The control unit generated data error messages, issued warnings to the operators through the user interface, and temporarily suspended the image acquisition module to avoid redundant storage of invalid image data.

[0063] If the data integrity check passes, the control unit further verifies the data's validity based on the sensor type and location range. Taking the physical layout of the sensor array as an example: Forward sonar: Installed at the bow, its detection range is 0°-120° (horizontal angle) forward of the bow, with a depth of 50 meters underwater. The azimuth angle in its data should be -60° to +60° (with the ship's centerline as 0°), and the depth value should be ≤50 meters. If the azimuth angle of a forward sonar data point is -80° (pointing to the port aft side), or the depth value is 60 meters, it exceeds the physical detection range of the sensor and is considered abnormal data.

[0064] Midships camera: Installed midway on the port side, with a field of view facing the open sea to the port side. The images it captures should include elements such as the water surface, distant coastline, or navigational aids. If the images show internal ship structures (such as deck machinery), it indicates that the camera's viewing angle is off, and the data location range is abnormal.

[0065] Positioning antenna: Installed on the top of the hull, it can theoretically receive satellite signals from any direction. However, when the number of satellites suddenly drops to 2 and the PDOP value soars to 15, the positioning may fail due to the antenna being blocked (such as temporary blockage by the crane arm), which is considered an abnormal position range.

[0066] In another example, the azimuth angle of the forward sonar data was +70° (exceeding the +60° range). Inspection revealed that this was due to a slight displacement of the sensor array caused by wave impact, resulting in a rightward deviation in the detection direction. Frequent appearances of ship railings in the mid-range camera image indicated that the camera's installation angle was excessively tilted downwards, requiring manual adjustment. The elevation value of the positioning data was -10 meters, significantly deviating from the actual elevation of the working area (5 meters). This was verified to be a misconfiguration of the sensor's output unit (mistakenly using "feet" instead of "meters"). All of the above data were flagged as abnormal by the control unit because they exceeded the reasonable thresholds for the sensor type or location range, triggering a data re-acquisition process.

[0067] The handling mechanism for data anomalies is as follows: Initial anomaly: The control unit sends a resampling command to the sensor receiving module to reacquire multi-source sensor data for the current working area. Taking sonar data anomaly as an example, during resampling, the control unit may adjust the sonar's operating frequency (e.g., switch from 200kHz to 150kHz) or extend the sampling time (from 1 second to 2 seconds) to obtain more stable data.

[0068] Secondary anomaly: If the data still does not meet the requirements after re-sampling, the control unit generates a detailed error report, including the abnormal data type, timestamp, sensor number and possible cause of failure (such as "front sonar azimuth angle out of range, suspected mechanical displacement"), and pushes it to maintenance personnel via email or SMS.

[0069] Continuous anomaly: If the same sensor outputs abnormal data in three consecutive resampling attempts, the control unit automatically isolates the sensor and switches to a redundant sensor (such as activating a backup sonar probe) or adjusts the detection range of other sensors to fill the gap. For example, if the forward port sonar fails, the control unit instructs the forward starboard sonar to expand its detection angle to -90° to +90°, temporarily covering the port area.

[0070] After data integrity and location range verifications are passed, the control unit triggers the image acquisition module to synchronously acquire images of the current environment. Taking a normal acquisition scenario as an example: sonar data clearly shows an underwater obstacle 30 meters ahead, with an azimuth of +30° and a depth of 20 meters; the camera image clearly captures the water reflection of the obstacle and surrounding navigation marks; positioning data shows the ship's coordinates as (120.5°E, 30.2°N), 8 satellites, and a PDOP value of 2.5. The control unit confirms that all three types of data meet integrity requirements and are within a reasonable range, and then instructs the image acquisition module to acquire a video stream at 1080P resolution and 25fps, and associates and stores the image data with the sensor data using the same timestamp (e.g., 2025-06-03T14:30:00.000Z).

[0071] The frequency of data integrity checks is related to the complexity of the operating environment. In nearshore areas with dense obstacles, the check cycle is set to once per second; in open water, the check cycle can be extended to once every 5 seconds. For positioning data, real-time checks are performed regardless of the environment to ensure the accuracy of the dredger's position coordinates and avoid trajectory planning deviations due to positioning errors.

[0072] Example 5: This embodiment relates to the specific module configuration and collaborative workflow of a multi-source obstacle perception and operation trajectory planning system for a trailing suction hopper dredger. Taking the system mounted on a certain type of trailing suction hopper dredger as an example, the functional implementation and interaction process of each unit are detailed as follows: The model building unit is responsible for establishing the basic model for obstacle perception and operation trajectory. During the system initialization phase, operators set multiple monitoring time points (e.g., every 10 seconds as a monitoring cycle) through the human-machine interface. At the set time points, the model building unit simultaneously acquires historical sampling data from multiple sources (e.g., sonar echo sequences from the past 24 hours, video clips of the operation area recorded by cameras, and ship trajectory coordinates recorded by positioning devices) and corresponding environmental images. For example, in a port dredging operation, the model building unit collects sonar data showing underwater obstacle echoes of different shapes, such as circles and elongated shapes; camera images can identify channel boundaries, buoys, and floating objects on the water surface; and positioning data records the S-shaped operation path of the ship within the operation area.

[0073] The model building unit preprocesses sonar data using data processing algorithms: first, median filtering removes impulse noise; then, Fourier transform extracts the frequency features of the echo signal; and finally, a support vector machine (SVM) algorithm is used to train an obstacle perception model, enabling it to distinguish different types of obstacles such as reefs, shipwrecks, and pipelines based on echo features, and output the obstacle's position coordinates and size parameters. For environmental images, the model building unit uses a convolutional neural network (CNN) for training, identifying features such as the channel centerline, no-navigation zone boundary, and water depth gauge in the image to generate a work trajectory model. This model can output parameters such as the radius of curvature and turning point coordinates of feasible work paths. Finally, based on the monitoring timestamp, the model building unit associates and stores the obstacle perception results at the same time (e.g., an obstacle is located at coordinates (100, 200) at 14:00:00) with the work trajectory information (e.g., the planned path at that time requires a 30° right turn), forming a spatiotemporally corresponding training dataset.

[0074] The data acquisition unit is responsible for acquiring multi-source data from the work area in real time. Sensor arrays located at the bow, midships, and stern of the vessel include a forward sonar unit that emits sound waves at a frequency of 200kHz and receives 10 sets of echo data per second, containing information such as distance (0-200 meters) and azimuth (-60° to +60°); a midships camera captures the work area on the left side at a resolution of 1920×1080 and a frame rate of 25 frames per second, capturing visual information such as water surface ripples and floating obstacles; and a positioning device that outputs longitude, latitude, altitude (accuracy down to the centimeter level), and heading angle (accuracy ±0.1°) in real time via the BeiDou satellite system. When the control unit triggers a monitoring command at 14:05:00, the data acquisition unit simultaneously acquires all three types of data, performs preliminary gain compensation on the sonar data (automatically amplifying the signal when echoes attenuate due to water depth changes), and performs automatic white balance correction on the camera images (eliminating color casts caused by cloud cover).

[0075] The information recognition unit uses a trained model to analyze real-time data. Taking the front sonar data as an example, after inputting it into the obstacle perception model, the model identifies a circular obstacle with a diameter of approximately 5 meters at an azimuth angle of +20°, located 150 meters ahead of the bow, which is determined to be an abandoned fishing net. Simultaneously, the image captured by the mid-mounted camera is input into the operation trajectory model. The model detects a yellow dashed line on the right side of the channel in the image (representing the boundary of the recommended operation path), 30 meters to the left of the hull. The ship must maintain its course within 5 meters to the left of the dashed line. The information recognition unit outputs the obstacle location information (coordinates (150, 200), type "fishing net") and operation trajectory information (path centerline coordinate sequence, boundary constraints) to the control unit.

[0076] The status acquisition unit collects hull status data in real time through hull sensors. The inertial navigation system (INS) outputs parameters such as roll angle (current value +3°), pitch angle (-1.5°), and speed (5 knots) at a frequency of 100Hz; the propeller monitoring system provides feedback on the port main engine speed (800 rpm) and the starboard main engine throttle opening (60%); the bow and stern draft sensors measure the draft as 8.2 meters (bow) and 8.5 meters (stern), respectively. This data is transmitted to the control unit via the CAN bus to assess the impact of hull attitude on obstacle detection (e.g., the roll angle causing a sonar detection angle offset requires correction).

[0077] The distribution acquisition unit integrates the ship's hull status and sensor data to generate obstacle distribution information. For example, combining the water depth acquisition value (12 meters) and the stern draft (8.5 meters), the actual water depth is calculated to be 3.5 meters, which is lower than the safe operating water depth threshold (4 meters), indicating the presence of a shallow obstacle with a height of 0.5 meters in the area. Simultaneously, based on obstacle location data from three consecutive monitoring cycles (located at (100,200) at 14:00:00, at (105,202) at 14:00:10, and at (110,205) at 14:00:20), it is analyzed that the obstacle is moving southeast at a speed of 0.5 meters per second, and is determined to be a floating object requiring dynamic tracking. The distribution acquisition unit generates monitoring instructions based on the above information, such as increasing the sonar scanning frequency to 2 times per second, and locking onto and tracking the area with the camera.

[0078] The path acquisition unit plans the operation path based on obstacle distribution information. When an abandoned fishing net (fixed obstacle) is detected 150 meters ahead and a shoal (continuous obstacle zone) is detected 30 meters to the right, the path acquisition unit uses the A* algorithm to search for a feasible path on the electronic nautical chart: the starting point is the current hull position (0,0), and the ending point is the target dredging point (300,300), avoiding the 20-meter radius around the fishing net and the shoal area (water depth <4 meters). The planned result is a rightward curve with a path length of 350 meters, including two turning points (100,150) and (200,250), with turning angles of +45° and -30° respectively, ensuring a safe distance from the obstacles.

[0079] The trajectory generation unit executes path matching and trajectory generation logic. First, it determines whether the set operation path (detour curve) matches the current operation trajectory information (originally planned straight travel). If the direction and path shape are inconsistent, the shipborne navigation system is triggered to re-acquire the current operation path (after confirmation by the operator, the detour plan is accepted). Then, based on the association established by the model building unit, it obtains the set safety zone information (20 meters around the fishing net is a no-navigation zone, and 10 meters outside the boundary of the shoal area is a safe distance). It then determines whether the current obstacle position (center of the fishing net (150,200)) conflicts with the safety zone (20-meter radius of the no-navigation zone). If the result is that the obstacle is located in the center of the no-navigation zone, an obstacle avoidance trajectory needs to be generated. The trajectory generation unit combines the hull turning radius (100 meters) and the thruster response delay (0.5 seconds) to calculate the segmented obstacle avoidance trajectory: first, it travels at a constant speed with a turning angle of +45° to the turning point (100, 150), then adjusts its heading to +30° to travel to (200, 250), and finally travels straight to the end point, maintaining a distance of more than 5 meters from the boundary of the fishing net prohibited navigation area throughout the entire process.

[0080] Each unit interacts with the others via an Ethernet bus. The control unit, acting as the core hub, uses a real-time operating system (RTOS) as its kernel to ensure data processing latency is less than 50 milliseconds. For example, obstacle location information output by the information recognition unit is transmitted in real time to the trajectory generation unit via UDP protocol, ship attitude data from the state acquisition unit is synchronized to the distributed acquisition unit via TCP protocol, and training results from the model building unit are stored in solid-state drives (SSDs) as files for real-time inference.

[0081] In actual operation, the system achieves full automation of the entire process from obstacle perception, data verification, status analysis, path planning to trajectory generation through multi-unit collaboration. For example, during nighttime operations, the camera automatically switches to infrared mode, and the operation trajectory model of the model building unit enhances the identification of thermal features (such as heat sources from ship engines) in infrared images. When strong winds cause the ship's roll angle to exceed 5°, the status acquisition unit provides real-time feedback to the distributed acquisition unit, which then performs attitude correction on the sonar data to avoid misjudging obstacle positions due to ship tilting. The entire system, through modular design, supports sensor type expansion and algorithm upgrades, possessing strong environmental adaptability and scalability.

[0082] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0083] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for multi-source obstacle perception and operation trajectory planning of a trailing suction hopper dredger, characterized in that, The methods include: Step S1: Simultaneously acquire sampling data from multiple source sensors and corresponding environmental images at multiple set monitoring times; process the sampling data from the multiple source sensors using a data processing algorithm to obtain an obstacle perception model; configure the environmental images to obtain a work trajectory model; associate the output results of the obstacle perception model and the work trajectory model according to the monitoring time; the multiple source sensors include a sonar device, a camera, and a positioning device; Step S2: Acquire multi-source sensor data of the current work area and simultaneously acquire the current environmental image while acquiring the multi-source sensor data of the current work area; Step S3: Obtain the current obstacle location information by identifying the current multi-source sensor data through the obstacle perception model; obtain the current operation trajectory information by identifying the current environmental image through the operation trajectory model; Step S4: Obtain the hull status data of the dredger at its current position; Step S5: Obtain obstacle distribution information in the current work area based on the hull status data; Step S6: Obtain the set operation path based on the obstacle distribution information; Step S7: If the set operation path matches the current operation trajectory information, then obtain the set safe area information according to the correlation between the output results of the obstacle perception model and the operation trajectory model; determine whether the current obstacle position information conflicts with the set safe area information. If yes, generate feasible trajectory data; otherwise, generate obstacle avoidance trajectory data. Step S2 includes: Step S21: Install a sensor receiving module on the hull of the dredger; the sensor receiving module is connected to the input and output terminals of the control unit; Step S22: Sensor arrays are installed at the front, middle and rear of the hull respectively; the sensor arrays are equipped with multi-source sensors facing outwards; Step S23: An image acquisition module is installed on the top of the hull, facing the work area; the image acquisition module is connected to the input / output terminal of the control unit and outputs the current environmental image to the control unit; Step S24: When the control unit receives a monitoring command, the output of the control unit drives the sensor receiving module to acquire multi-source sensor data; the output of the control unit drives the image acquisition module to acquire the current environmental image. The control unit combines multi-source sensor data to perform cluster analysis on obstacles in the work area, identifying multiple independent obstacle segments. The sonar device generates point cloud data of obstacles by emitting sound waves and receiving echoes. Each point cloud data point contains information such as distance, azimuth, and elevation angle. The camera separates the obstacle target from the background in the image through an image segmentation algorithm, extracting the outline features and pixel coordinates of the obstacle. The positioning device assigns geographical coordinates to each obstacle detection result. The control unit performs spatiotemporal registration of the multi-source sensor data, unifying it to the same coordinate system. Then, it uses a density clustering algorithm to cluster the obstacle detection results, dividing obstacles with similar spatial locations and consistent motion states into the same "segment," thus forming a multi-segment obstacle distribution.

2. The method for multi-source obstacle perception and operation trajectory planning of a trailing suction hopper dredger according to claim 1, characterized in that, Step S5 includes: acquiring the water depth data of the current working area; acquiring the water depth information of the current working area based on the hull status data and the water depth data; acquiring the obstacle height information of the current working area based on the water depth information; and generating the monitoring command based on the obstacle height information.

3. The method for multi-source obstacle perception and operation trajectory planning of a trailing suction hopper dredger according to claim 2, characterized in that, Step S5 includes: obtaining the distribution of multiple obstacles in the current working area based on the hull status data; obtaining the location information of multiple obstacles and the spacing information of multiple obstacles in the current working area based on the distribution of multiple obstacles in the current working area; and generating the monitoring command based on the location information of multiple obstacles and the spacing information of multiple obstacles.

4. The method for multi-source obstacle perception and operation trajectory planning of a trailing suction hopper dredger according to claim 1, characterized in that, Step S2 includes: Acquire multi-source sensor data for the current work area; the multi-source sensor data includes sonar data, camera data, and positioning data; Determine whether the sonar data, camera data, and positioning data meet the set integrity requirements. If yes, determine whether the sonar data, camera data, and positioning data are within the sensor type and location range based on the sensor type and location range. If yes, acquire the current environmental image simultaneously with the multi-source sensor data of the current work area. If no, return to this step and reacquire the multi-source sensor data of the current work area. Determine whether the sonar data, camera data, and positioning data meet the set integrity requirements; if not, generate a data error message.

5. The method for multi-source obstacle perception and operation trajectory planning of a trailing suction hopper dredger according to claim 1, characterized in that, Step S7 includes: if the set operation path does not match the current operation trajectory information, then the current operation path is obtained through the shipborne navigation system or trajectory planning module, and the process returns to this step to re-determine whether the set operation path matches the current operation trajectory information.

6. A multi-source obstacle perception and operation trajectory planning system for a trailing suction hopper dredger, characterized in that, Its system includes: The model building unit is configured to simultaneously acquire sampling data from multiple source sensors and corresponding environmental images at multiple set monitoring times; process the sampling data from the multiple source sensors through a data processing algorithm to obtain an obstacle perception model; configure the environmental images to obtain a work trajectory model; and correlate the output results of the obstacle perception model and the work trajectory model according to the monitoring time; the multiple source sensors include a sonar device, a camera, and a positioning device; The data acquisition unit is configured to acquire multi-source sensor data of the current work area and, while acquiring the multi-source sensor data of the current work area, simultaneously acquire the current environmental image. The information recognition unit is configured to identify the current obstacle location information by using the obstacle perception model to identify the current multi-source sensor data; and to identify the current operation trajectory information by using the operation trajectory model to identify the current environmental image. The status acquisition unit is configured to acquire the hull status data of the dredger at its current position. The distribution acquisition unit is configured to acquire obstacle distribution information of the current working area based on the hull status data. The path acquisition unit is configured to acquire a set operation path based on the obstacle distribution information. The trajectory generation unit is configured to, if the set operation path matches the current operation trajectory information, obtain the set safe area information according to the correlation between the output results of the obstacle perception model and the operation trajectory model; determine whether the current obstacle position information conflicts with the set safe area information; if so, generate feasible trajectory data; if not, generate obstacle avoidance trajectory data. The data acquisition unit is also configured to: A sensor receiving module is installed on the hull of the dredger; the sensor receiving module is connected to the input and output terminals of the control unit; Sensor arrays are installed at the front, middle and rear of the hull; the sensor arrays are equipped with multi-source sensors facing outwards; An image acquisition module is installed on the top of the hull, facing the work area; the image acquisition module is connected to the input / output terminal of the control unit and outputs the current environmental image to the control unit; When the control unit receives a monitoring command, its output drives the sensor receiving module to acquire multi-source sensor data; the output of the control unit also drives the image acquisition module to acquire the current environmental image. The control unit combines multi-source sensor data to perform cluster analysis on obstacles in the work area, identifying multiple independent obstacle segments. The sonar device generates point cloud data of obstacles by emitting sound waves and receiving echoes. Each point cloud data point contains information such as distance, azimuth, and elevation angle. The camera separates the obstacle target from the background in the image through an image segmentation algorithm, extracting the outline features and pixel coordinates of the obstacle. The positioning device assigns geographical coordinates to each obstacle detection result. The control unit performs spatiotemporal registration of the multi-source sensor data, unifying it to the same coordinate system. Then, it uses a density clustering algorithm to cluster the obstacle detection results, dividing obstacles with similar spatial locations and consistent motion states into the same "segment," thus forming a multi-segment obstacle distribution.

7. The multi-source obstacle perception and operation trajectory planning system for a trailing suction hopper dredger according to claim 6, characterized in that, The distribution acquisition unit is further configured to: acquire water depth data of the current working area; acquire water depth information of the current working area based on hull status data and the water depth data; acquire obstacle height information of the current working area based on the water depth information; generate the monitoring command based on the obstacle height information; or the distribution acquisition unit is further configured to: acquire the distribution of multiple obstacles in the current working area based on hull status data; acquire the location information of multiple obstacles and the spacing information of multiple obstacles in the current working area based on the distribution of multiple obstacles in the current working area. The monitoring command is generated based on the location information of the multiple obstacles and the spacing information of the multiple obstacles.

8. The multi-source obstacle perception and operation trajectory planning system for a trailing suction hopper dredger according to claim 7, characterized in that, The data acquisition unit is also configured to: Acquire multi-source sensor data for the current work area; the multi-source sensor data includes sonar data, camera data, and positioning data; Determine whether the sonar data, camera data, and positioning data meet the set integrity requirements. If yes, determine whether the sonar data, camera data, and positioning data are within the sensor type and location range based on the sensor type and location range. If yes, acquire the current environmental image simultaneously with the multi-source sensor data of the current work area. If no, return to this step and reacquire the multi-source sensor data of the current work area. Determine whether the sonar data, camera data, and positioning data meet the set integrity requirements; if not, generate a data error message.

Images