Marine target fusion sensing method based on fuzzy set theory

Abstract

The invention belongs to the technical field of ocean information, and particularly relates to a sea target fusion sensing method based on a fuzzy set theory, which can make full use of complementary information among sensors through multi-sensor data fusion, make up for the deficiency of a single sensor, and can realize the fusion sensing of a sea target in the fusion process. The fuzzy set theory can carry out reasonable fuzzification processing on data provided by different sensors, the false drop rate and the omission rate are reduced, and then the accuracy and robustness of target sensing are improved.

Description

Technical Field

[0001] The present invention belongs to the field of marine information technology, and in particular relates to a marine target fusion perception method based on fuzzy set theory. Background Art

[0002] With the development of marine technology, unmanned vessels, as autonomous surface platforms, are playing an increasingly important role in marine environmental monitoring and ocean information acquisition. Unmanned vessels integrate multiple sensors, such as visual cameras and 3D lidar, to acquire information about obstacles and targets in their surroundings, enabling autonomous navigation and decision-making. However, in complex marine environments, unmanned vessels face a diverse array of obstacles, posing significant challenges to their safe and autonomous navigation. Object perception is a prerequisite and key core technology for autonomous navigation and collision avoidance at sea.

[0003] Currently, unmanned vessels rely heavily on data from radar and optical sensors to detect targets at sea. However, the maritime environment is complex and ever-changing, and sensor data often contains uncertainty and ambiguity. Furthermore, the heterogeneity and incompleteness of sensor information make it difficult for a single sensor to meet the requirements for maritime target detection. Summary of the Invention

[0004] In response to the various shortcomings of the existing technology, the inventors have researched and designed a maritime target fusion perception method based on fuzzy set theory in long-term practice, which is suitable for autonomous navigation and collision avoidance of unmanned ships against maritime obstacles.

[0005] The present invention's fusion perception method for maritime targets based on fuzzy set theory is specifically as follows:

[0006] Step 1: De-noise the maritime target sensor data, then map each sensor data to the corresponding grid of the grid map, and record the observation value of each grid;

[0007] Step 2: Design the membership function of each sensor according to the sensor's measurement range and the difference in detection performance in different ranges;

[0008] Step 3: Divide the sensing range of each sensor, use the membership function of each sensor, design fuzzy rules for grids belonging to obstacle areas by region, and build a fuzzy rule library for obstacle target fusion perception;

[0009] Step 4: Based on the input of the current target sensor, activate the corresponding rules and use the corresponding fuzzy rules to calculate the fusion membership value of the output grid as the obstacle area;

[0010] Step 5. Finally, set the membership threshold and determine whether each grid in the grid map is an obstacle area based on the threshold to complete the fusion perception of maritime targets.

[0011] Furthermore, in step one, the marine target sensor includes but is not limited to three-dimensional laser radar and stereo vision.

[0012] Furthermore, in step 1, the sensor data is denoised by removing invalid measurement data by setting a ranging range threshold.

[0013] Furthermore, in step one, the three-dimensional lidar point cloud data and the three-dimensional reconstructed stereoscopic point cloud data are two-dimensionally projected to obtain a bird's-eye view, and a mapping relationship is established with the grid map.

[0014] Furthermore, in step 2, the position information of the target at sea measured by each sensor is used as the input variable. According to the sensor measurement range and the difference in detection performance in different ranges, the membership function of each sensor is designed. The specific method is as follows:

[0015] If there is 3D laser radar detection data in the grid, the membership function for judging whether the grid is an obstacle target area is designed as follows, where d is the measured distance of the target at sea, .

[0016] If there is no 3D lidar detection data in the grid, the membership function for judging whether the grid is a passable area is designed as follows, where d is the distance between the grid and the center of the map, .

[0017] If there is stereo vision detection data in the grid, the membership function for judging whether the grid is an obstacle target area is designed as follows, where d is the measured distance of the target at sea, .

[0018] If there is no stereo vision detection data in the grid, the membership function for judging whether the grid is a passable area is designed as follows, where d is the distance between the grid and the center of the map, .

[0019] Furthermore, in step three, the perception range of the three-dimensional lidar and stereo vision is divided into three areas: R1, R2 and R3, where R1 is the area that can be perceived by both the three-dimensional lidar and stereo vision, R2 is the area that can be perceived only by stereo vision, and R3 is the area that can be perceived only by the three-dimensional lidar.

[0020] Furthermore, in step 3, the membership function of each sensor is used to design fuzzy rules for grid membership in obstacle areas by region, and a fuzzy rule base for obstacle target fusion perception is constructed. The specific design method is as follows:

[0021] In the R1 area, based on the current 3D lidar and stereo vision detection data, the fuzzy rules for determining whether the grid is an obstacle target area are designed as follows: Rule 1.1: If the grid has 3D lidar and stereo vision detection data, the grid is the comprehensive membership of the obstacle target area. ; Rule 1.2: If a grid has 3D LiDAR detection data but no stereo vision detection data, , then the grid is the comprehensive membership of the obstacle target area ;when , then the grid is the comprehensive membership of the obstacle target area ; Rule 1.3: If the grid has stereo vision detection data but no 3D LiDAR detection data, when , then the grid is the comprehensive membership of the obstacle target area ;when , then the grid is the comprehensive membership of the obstacle target area ; Rule 1.4: If the grid does not have 3D LiDAR and stereo vision detection data, the comprehensive membership of the grid as an obstacle target area is ;

[0022] In the R2 area, based on the current stereo vision detection data, the fuzzy rule for determining whether the grid is an obstacle target area is designed as follows: Rule 2.1: If there is stereo vision detection data for the grid, the grid is the comprehensive membership of the obstacle target area ; Rule 2.2: If there is no stereo vision detection data for the grid, the grid is the comprehensive membership of the obstacle target area ;

[0023] In the R3 area, based on the current 3D lidar detection data, the fuzzy rule for determining whether the grid is an obstacle target area is designed as follows: Rule 3.1: If there is 3D LiDAR detection data for the grid, the grid is the comprehensive membership of the obstacle target area ; Rule 3.2: If there is no 3D LiDAR detection data for the grid, the grid is the comprehensive membership of the obstacle target area .

[0024] Furthermore, in step 4, the current 3D laser radar and stereo vision detection data are used as input and converted into membership values; according to the current input, the corresponding fuzzy rules are activated, and the corresponding fuzzy rules are used to calculate the comprehensive membership value of the output grid as the obstacle target area. .

[0025] Furthermore, in step 5, the membership threshold is set to 0.5. If the comprehensive membership If the threshold is exceeded, the grid is marked as an obstacle target area, and the grid map is traversed to complete the fusion perception of the maritime target.

[0026] The beneficial effects of the present invention are:

[0027] 1) Through multi-sensor data fusion, we can make full use of the complementary information between sensors, make up for the shortcomings of a single sensor, reduce the false detection rate and missed detection rate, and thus improve the accuracy and robustness of target perception.

[0028] Fuzzy set theory is primarily used to address uncertainty and ambiguity, making it suitable for handling inconsistencies and ambiguity between sensor data. This makes it suitable for multi-sensor fusion perception of maritime targets. During the fusion process, fuzzy set theory can rationally fuzzify the data provided by different sensors, effectively improving the accuracy and robustness of the results. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a flowchart of the inversion method of the present invention.

[0030] Figure 2 It is the three-dimensional laser radar and stereo vision data under the typical marine target scene of the present invention.

[0031] Figure 3 It is a schematic diagram of the division of the three-dimensional laser radar and stereo vision detection area of the present invention.

[0032] Figure 4 This is a schematic diagram of the fusion perception results of maritime targets based on fuzzy set theory of the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0034] See also Figure 1, the present invention proposes a fusion perception method for maritime targets based on fuzzy set theory, the specific steps are as follows:

[0035] Step 1: De-noise the maritime target sensor data, then map each sensor data to the corresponding grid of the grid map, and record the observation value of each grid.

[0036] Step 2: Design the membership function of each sensor according to the sensor's measurement range and the difference in detection performance in different ranges.

[0037] Step 3: Divide the perception range of each sensor, use the membership function of each sensor, design fuzzy rules for grids belonging to obstacle areas by region, and build a fuzzy rule library for obstacle target fusion perception.

[0038] Step 4: According to the input of the current target sensor, activate the corresponding rules and use the corresponding fuzzy rules to calculate the fusion membership value of the output grid as the obstacle area.

[0039] Step 5. Finally, set the membership threshold and determine whether each grid in the grid map is an obstacle area based on the threshold to complete the fusion perception of maritime targets.

[0040] In step 1, maritime target sensors include, but are not limited to, 3D lidar and stereo vision. Sensor data denoising is performed by setting a range threshold to remove invalid measurement data. The 3D lidar point cloud data and the 3D reconstructed stereo vision point cloud data are then projected 2D to create a bird's-eye view, which is then mapped to the raster map.

[0041] In step 2, the target position information measured by each sensor is used as the input variable. According to the sensor measurement range and the difference in detection performance in different ranges, the membership function of each sensor is designed. The specific method is as follows:

[0042] If there is three-dimensional lidar detection data in the grid, the membership function for judging whether the grid is an obstacle target area is designed as follows, where d is the measured distance of the target at sea.

[0043]

[0044] If there is no three-dimensional lidar detection data in the grid, the membership function for judging whether the grid is a passable area is designed as follows, where d is the distance between the grid and the center of the map.

[0045]

[0046] If there is stereo vision detection data in the grid, the membership function for judging whether the grid is an obstacle target area is designed as follows, where d is the measured distance of the target at sea.

[0047]

[0048] If there is no stereo vision detection data in the grid, the membership function for judging whether the grid is a passable area is designed as follows, where d is the distance between the grid and the center of the map.

[0049]

[0050] In step three, the perception range of the 3D LiDAR and stereo vision is divided into three regions: R1, R2, and R3. R1 is the area perceived by both the 3D LiDAR and stereo vision, R2 is the area perceived by stereo vision only, and R3 is the area perceived by the 3D LiDAR only. Then, using the membership functions of each sensor, fuzzy rules are designed for grid membership in obstacle regions by region, and a fuzzy rule library for fusion perception of obstacles and targets is constructed. The specific design method is as follows:

[0051] In the R1 area, based on the current 3D lidar and stereo vision detection data, the fuzzy rules for determining whether the grid is an obstacle target area are designed as follows: Rule 1.1: If the grid has 3D lidar and stereo vision detection data, the grid is the comprehensive membership of the obstacle target area. .

[0052] Rule 1.2: If a grid has 3D LiDAR detection data but no stereo vision detection data, , then the grid is the comprehensive membership of the obstacle target area ;when , then the grid is the comprehensive membership of the obstacle target area .

[0053] Rule 1.3: If the grid has stereo vision detection data but no 3D LiDAR detection data, when , then the grid is the comprehensive membership of the obstacle target area ;when , then the grid is the comprehensive membership of the obstacle target area .

[0054] Rule 1.4: If the grid does not have 3D LiDAR and stereo vision detection data, the comprehensive membership of the grid as an obstacle target area is .

[0055] In the R2 area, based on the current stereo vision detection data, the fuzzy rule for determining whether the grid is an obstacle target area is designed as follows: Rule 2.1: If there is stereo vision detection data for the grid, the grid is the comprehensive membership of the obstacle target area .

[0056] Rule 2.2: If there is no stereo vision detection data for the grid, the grid is the comprehensive membership of the obstacle target area .

[0057] In the R3 area, based on the current 3D lidar detection data, the fuzzy rule for determining whether the grid is an obstacle target area is designed as follows: Rule 3.1: If there is 3D LiDAR detection data for the grid, the grid is the comprehensive membership of the obstacle target area .

[0058] Rule 3.2: If there is no 3D LiDAR detection data for the grid, the grid is the comprehensive membership of the obstacle target area .

[0059] In step 4, the current 3D laser radar and stereo vision detection data are used as input and converted into membership values. According to the current input, the corresponding fuzzy rules are activated and the output grid is calculated using the corresponding fuzzy rules as the comprehensive membership value of the obstacle target area. .

[0060] In step 5, the membership threshold is set to 0.5. If the comprehensive membership If the threshold is exceeded, the grid is marked as an obstacle target area, and the grid map is traversed to complete the fusion perception of the maritime target.

[0061] The following further introduces the maritime target fusion perception method of the present invention based on a specific embodiment.

[0062] In step (1) of this embodiment, the marine target sensor includes a three-dimensional laser radar and a stereo vision sensor, but is not limited to the above two sensors in other embodiments. The effective detection range of the three-dimensional laser radar used is 120m, and the effective detection range of the stereo vision sensor is 200m.

[0063] Figure 2 This is 3D lidar and stereo vision data from a typical maritime target scenario. Sensor data denoising is performed by setting a range threshold to remove invalid measurement data. The 3D lidar point cloud data and the 3D reconstructed stereo vision point cloud data are then projected 2D to create a bird's-eye view, which is then mapped to the raster map.

[0064] In step (2) of this embodiment, the position information of the target at sea measured by each sensor is used as an input variable, and the membership function of each sensor is designed according to the sensor measurement range and the difference in detection performance in different ranges.

[0065] If there is three-dimensional lidar detection data in the grid, the membership function for judging whether the grid is an obstacle target area is designed as follows, where d is the measured distance of the target at sea.

[0066]

[0067] If there is no three-dimensional lidar detection data in the grid, the membership function for judging whether the grid is a passable area is designed as follows, where d is the distance between the grid and the center of the map.

[0068]

[0069] If there is stereo vision detection data in the grid, the membership function for judging whether the grid is an obstacle target area is designed as follows, where d is the measured distance of the target at sea.

[0070]

[0071] If there is no stereo vision detection data in the grid, the membership function for judging whether the grid is a passable area is designed as follows, where d is the distance between the grid and the center of the map.

[0072]

[0073] In step (3) of this embodiment, the perception range of the three-dimensional laser radar and stereo vision is divided into three areas: R1, R2 and R3. Figure 3 The following diagram illustrates the division of detection areas for the 3D LiDAR and stereo vision systems used: R1 represents the area perceived by both the 3D LiDAR and stereo vision systems, R2 represents the area perceived by stereo vision systems only, and R3 represents the area perceived by the 3D LiDAR only. Fuzzy rules for determining grid membership in obstacle regions are then designed using the membership functions of each sensor, building a fuzzy rule library for fusion perception of obstacles and targets.

[0074] In the R1 area, based on the current 3D lidar and stereo vision detection data, the fuzzy rules for determining whether the grid is an obstacle target area are designed as follows: Rule 1.1: If the grid has 3D lidar and stereo vision detection data, the grid is the comprehensive membership of the obstacle target area. .

[0075] Rule 1.2: If a grid has 3D LiDAR detection data but no stereo vision detection data, , then the grid is the comprehensive membership of the obstacle target area ;when , then the grid is the comprehensive membership of the obstacle target area .

[0076] Rule 1.3: If the grid has stereo vision detection data but no 3D LiDAR detection data, when , then the grid is the comprehensive membership of the obstacle target area ;when , then the grid is the comprehensive membership of the obstacle target area .

[0077] Rule 1.4: If the grid does not have 3D LiDAR and stereo vision detection data, the comprehensive membership of the grid as an obstacle target area is .

[0078] In the R2 area, based on the current stereo vision detection data, the fuzzy rule for determining whether the grid is an obstacle target area is designed as follows: Rule 2.1: If there is stereo vision detection data for the grid, the grid is the comprehensive membership of the obstacle target area .

[0079] Rule 2.2: If there is no stereo vision detection data for the grid, the grid is the comprehensive membership of the obstacle target area .

[0080] In the R3 area, based on the current 3D lidar detection data, the fuzzy rule for determining whether the grid is an obstacle target area is designed as follows: Rule 3.1: If there is 3D LiDAR detection data for the grid, the grid is the comprehensive membership of the obstacle target area .

[0081] Rule 3.2: If there is no 3D LiDAR detection data for the grid, the grid is the comprehensive membership of the obstacle target area .

[0082] In step (4) of this embodiment, the current 3D laser radar and stereo vision detection data are used as input and converted into membership values; according to the current input, the corresponding fuzzy rules are activated, and the output grid is calculated using the corresponding fuzzy rules as the comprehensive membership value of the obstacle target area. .

[0083] In step (5) of this embodiment, the membership threshold is set to 0.5. If the comprehensive membership If the threshold is exceeded, the grid is marked as an obstacle target area, and the grid map is traversed to complete the fusion perception of the maritime target. Figure 4Fusion perception results for maritime targets based on fuzzy set theory and Figure 2 A comparison of the 3D lidar and stereo vision data for maritime targets shows that the maritime targets in the scene are generally beyond the ranging range of the 3D lidar. Multi-source sensor data fusion enables perception of maritime targets at greater distances (approximately 100 to 200 meters) ahead, while reducing false detections by stereo vision.

[0084] The above description is only an implementation method of the present application and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the description and drawings of this application, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A fusion perception method for maritime targets based on fuzzy set theory, characterized in that: The specific method is as follows: Step 1: De-noise the maritime target sensor data, then map each sensor data to the corresponding grid of the grid map, and record the observation value of each grid; Step 2: Design the membership function of each sensor according to the sensor's measurement range and the difference in detection performance in different ranges; Step 3: Divide the sensing range of each sensor, use the membership function of each sensor, design fuzzy rules for grids belonging to obstacle areas by region, and build a fuzzy rule library for obstacle target fusion perception; Step 4: Based on the input of the current target sensor, activate the corresponding rules and use the corresponding fuzzy rules to calculate the fusion membership value of the output grid as the obstacle area; Step 5. Finally, set the membership threshold and determine whether each grid in the grid map is an obstacle area based on the threshold to complete the fusion perception of maritime targets.

2. The method according to claim 1, characterized in that In step 1, the marine target sensor includes but is not limited to three-dimensional laser radar and stereo vision.

3. The method according to claim 2, characterized in that In step 1, the sensor data is denoised by setting a ranging range threshold to remove invalid measurement data.

4. The method according to claim 3, characterized in that In step 1, the 3D lidar point cloud data and the 3D reconstructed stereo point cloud data are projected in 2D to obtain a bird's-eye view, and a mapping relationship is established with the raster map.

5. The method according to claim 1, wherein In step 2, the target position information measured by each sensor is used as the input variable. According to the sensor measurement range and the difference in detection performance in different ranges, the membership function of each sensor is designed. The specific method is as follows: If there is 3D lidar detection data in the grid, the membership function for judging whether the grid is an obstacle target area is designed as follows, where d is the measured distance of the target at sea, ； If there is no 3D lidar detection data in the grid, the membership function for judging whether the grid is a passable area is designed as follows, where d is the distance between the grid and the center of the map, ； If there is stereo vision detection data in the grid, the membership function for judging whether the grid is an obstacle target area is designed as follows, where d is the measured distance of the target at sea, ； If there is no stereo vision detection data in the grid, the membership function for judging whether the grid is a passable area is designed as follows, where d is the distance between the grid and the center of the map, 。 6. The method according to claim 5, characterized in that In step three, the perception range of the 3D lidar and stereo vision is divided into three areas: R1, R2, and R3. R1 is the area that can be perceived by both the 3D lidar and stereo vision, R2 is the area that can be perceived only by stereo vision, and R3 is the area that can be perceived only by the 3D lidar.

7. The method according to claim 6, characterized in that In step 3, the membership function of each sensor is used to design fuzzy rules for grids belonging to obstacle areas by region, and a fuzzy rule base for obstacle target fusion perception is constructed. The specific design method is as follows: In the R1 area, based on the current 3D lidar and stereo vision detection data, the fuzzy rules for determining whether the grid is an obstacle target area are designed as follows: Rule 1.1: If the grid has 3D lidar and stereo vision detection data, the grid is the comprehensive membership of the obstacle target area. ; Rule 1.2: If a grid has 3D LiDAR detection data but no stereo vision detection data, , then the grid is the comprehensive membership of the obstacle target area ;when , then the grid is the comprehensive membership of the obstacle target area ; Rule 1.3: If the grid has stereo vision detection data but no 3D LiDAR detection data, when , then the grid is the comprehensive membership of the obstacle target area ;when , then the grid is the comprehensive membership of the obstacle target area ; Rule 1.4: If the grid does not have 3D LiDAR and stereo vision detection data, the comprehensive membership of the grid as an obstacle target area is ; In the R2 area, based on the current stereo vision detection data, the fuzzy rule for determining whether the grid is an obstacle target area is designed as follows: Rule 2.1: If there is stereo vision detection data for the grid, the grid is the comprehensive membership of the obstacle target area ; Rule 2.2: If there is no stereo vision detection data for the grid, the grid is the comprehensive membership of the obstacle target area ; In the R3 area, based on the current 3D lidar detection data, the fuzzy rule for determining whether the grid is an obstacle target area is designed as follows: Rule 3.1: If there is 3D LiDAR detection data for the grid, the grid is the comprehensive membership of the obstacle target area ; Rule 3.2: If there is no 3D LiDAR detection data for the grid, the grid is the comprehensive membership of the obstacle target area .

8. The method according to claim 7, characterized in that In step 4, the current 3D lidar and stereo vision detection data are used as input and converted into membership values; according to the current input, the corresponding fuzzy rules are activated, and the output grid is calculated using the corresponding fuzzy rules as the comprehensive membership value of the obstacle target area. .

9. The method according to claim 8, characterized in that In step 5, the membership threshold is set to 0.

5. If the comprehensive membership If the threshold is exceeded, the grid is marked as an obstacle target area, and the grid map is traversed to complete the fusion perception of the maritime target.

Images