A radar-based method for detecting obstacles for ships entering and leaving a port in low visibility

Abstract

This invention relates to the field of ship collision avoidance technology, specifically a radar-based obstacle detection method for ships entering and leaving ports in low visibility conditions. The method includes: labeling the obstacle types of each obstacle based on radar signals transmitted by the ship; clustering the obstacles to form an obstacle distribution heatmap; calculating the safe distance for each obstacle based on its type to obtain the obstacle identification boundary; acquiring the ship's current navigation behavior during entry and exit; if the navigation behavior meets the safe distance requirement, retrieving the expected execution time of the corresponding navigation behavior to construct a behavior sequence; determining the entry and exit distances of the behavior sequence at corresponding timestamps; segmenting and correcting the behavior according to these distances to determine the control boundary for each segment; and using the minimum intersection of the control boundary and the obstacle identification boundary as the output data. This method improves the accuracy of obstacle identification and the ship's obstacle avoidance performance.

Description

Technical Field

[0001] This invention relates to the field of ship collision avoidance technology, specifically a radar-based method for detecting obstacles when a ship enters or leaves port in low visibility conditions. Background Technology

[0002] Marine radar is a water safety detection technology that monitors and tracks various obstacles such as ships, buoys, and reefs in the waterway. It can obtain information such as the size, speed, and elevation of obstacles around the ship during navigation. However, in low visibility environments such as fog, haze, heavy rain, and night, radar signals may be interfered with by atmospheric and marine factors, resulting in anomalies in some radar monitoring data and affecting the obstacle avoidance operations of ships when entering and leaving ports.

[0003] For example, Chinese Patent Publication No. CN118011426A discloses a ship close-range collision avoidance early warning method and system based on lidar, which relates to the technical field of ship collision avoidance. The method uses lidar to collect point cloud data, encapsulates the collected point cloud data into data packets, obtains several data frames through preprocessing, and combines the data frames into a single scan frame by setting an angle refresh mechanism, which serves as the obstacle dataset to be identified. Based on the point cloud data in the obstacle dataset, the final distance between the obstacle and the ship is calculated. If the final distance is less than a preset safe distance, an early warning command is generated.

[0004] For example, Chinese Patent Publication No. CN118050728A discloses a target acquisition method and system for waterway safety monitoring. This invention first acquires sea surface and obstacle measurement points in the radar elevation monitoring map at each sampling time during ship navigation; then, it acquires the dynamic expectation factor for each obstacle measurement point, further acquiring all obstacle avoidance paths at each sampling time; based on the differences between obstacle avoidance paths, it acquires the threat coefficient of each obstacle measurement point, and then acquires the anomaly degree of each obstacle measurement point and adjusts the corresponding local reachability density to obtain the final radar elevation monitoring map.

[0005] Existing technologies use horizontal angle scanning and refreshing to sort obstacle point clouds in ascending order; and use the distribution difference between obstacle measurement points and obstacle avoidance paths to set the threat coefficient of target obstacle measurement points based on the distance between the obstacle measurement points and the obstacle avoidance paths. However, existing technologies focus on the distribution of obstacle measurement points based on coordinate points, without considering the relationship between the distance of the obstacle itself and the distance of the ship itself. This results in the inability to identify the relative distance between the two during obstacle avoidance processing. Furthermore, by only emphasizing the reachability threat, the obstacle avoidance path lacks effective boundary constraints, leading to frequent alarms in port entry and exit scenarios, which prevents the ship from performing port entry and exit functions normally. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a radar-based method for detecting obstacles in low visibility conditions for ships entering and leaving ports, comprising: S1, based on the radar signals currently transmitted by the ship, using the signal characteristics of the radar signals to preliminarily identify obstacles in the current entry and exit channels, and marking the obstacle type of each obstacle preliminarily identified.

[0007] S2, based on the trajectory of the obstacle under continuous identification, calculates the direction angle and distance of the obstacle relative to the current ship, and after clustering the obstacles, forms an obstacle distribution heat map.

[0008] S3, based on the obstacle type corresponding to each obstacle, calculates the safe distance corresponding to each obstacle, and divides the obstacle identification boundary corresponding to each obstacle in the obstacle distribution heat map.

[0009] S4: Obtain the current vessel's navigation behavior when entering or leaving the port. If the navigation behavior meets the safe distance requirement, retrieve the estimated execution time of the corresponding navigation behavior and construct a behavior sequence.

[0010] S5, determine the inbound and outbound distances of the behavior sequence at the corresponding timestamp, perform segmentation correction according to the inbound and outbound distances, determine the control boundary for each segment of travel, and use the minimum intersection of the control boundary and the obstacle recognition boundary as the output data.

[0011] The beneficial effects of this invention are as follows: First, this invention uses radar signal characteristics to initially identify and label obstacles in and out of port channels; calculates the direction angle and distance relative to ships based on the continuous trajectory of obstacles, and forms an obstacle distribution heat map after clustering; calculates the corresponding safe distance according to the obstacle type, and delineates the obstacle identification boundary for each obstacle; acquires ship driving behavior, retrieves the estimated execution time for behaviors that do not meet the safe distance requirement, and constructs a behavior sequence; corrects the driving control boundary in segments according to the distance to and from port, and uses the minimum intersection of the control boundary and the obstacle identification boundary as the output data. This avoids the risk of ship collisions caused by misjudgment of obstacles or blurred boundaries, and improves the accuracy and efficiency of obstacle identification for ships entering and leaving ports in low visibility conditions.

[0012] Second, this invention calibrates the radar signal coordinate system by binding it to the ship coordinate system, clarifying the correspondence between the detection angle and the ship's rudder direction and vertical direction; it identifies the obstacle echo position using signal characteristics such as echo intensity, Doppler frequency shift, and echo shape; it sets four initial obstacle types: fixed, semi-fixed, moving, and floating, and completes type correction by verifying the echo position trajectory; it reduces the misjudgment of obstacle type caused by a single signal feature, ensures that the identified obstacle type is consistent with the actual one, and further improves the accuracy of obstacle identification.

[0013] Third, this invention converts the direction, angle, and distance of obstacles into three-dimensional spatial coordinates in the ship's coordinate system, generates a set of trajectory points through trajectory tracking, calculates the multi-plane cross-sectional area of ​​the minimum bounding box, performs feature matching in combination with Euclidean distance, and verifies the obstacle type twice by vertical water surface height to filter out usable obstacle coordinates; it realizes data filtering under various obstacle types and improves the clustering accuracy for obstacle distribution locations.

[0014] Fourth, this invention configures geometric contour boundaries based on the location of obstacles in a heat map; first, it sets a basic safety distance based on the length and width of the ship, then decomposes the longitudinal and lateral components of the ship's trajectory, and combines reaction time, wind speed, and water speed to perform vector superposition to obtain a corrected safety distance. Finally, it combines the available width of the waterway, the width occupied by the ship in the waterway to calculate the maximum allowable safety distance, and the upper limit of the safety distance, and takes the minimum value of the three as the final safety distance; this makes the obtained obstacle boundaries conform to the operational delay situation in low visibility scenarios, thereby improving the effect of dividing the boundaries of multiple obstacle distributions, and finally completing the processing of obstacle risk quantification.

[0015] Fifth, this invention calculates the real-time distance sequence between a ship and obstacles under any driving behavior; it then filters the nearest encounter distance and the nearest encounter time using the real-time distance sequence to quantify the distance relationship between the obstacle and the current ship, thereby identifying driving behaviors with risks. By using the control boundary and obstacle identification boundary corresponding to the driving behavior, the risk intersection is quantified. Finally, the path parameters of different driving behaviors in low visibility scenarios are quantified by the boundary intersection method, improving the output accuracy of obstacle avoidance driving behavior range and the obstacle avoidance effect under obstacle detection. Attached Figure Description

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] Figure 1 This is a flowchart illustrating a radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions.

[0018] Figure 2 This is a flowchart illustrating step S1 of a radar-based method for detecting obstacles when a ship enters or leaves port in low visibility conditions.

[0019] Figure 3 This is a flowchart illustrating step S2 of a radar-based method for detecting obstacles when a ship enters or leaves port in low visibility conditions.

[0020] Figure 4 This is a flowchart illustrating step S3 of a radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions.

[0021] Figure 5This is a flowchart illustrating step S4 of a radar-based method for detecting obstacles when a ship enters or leaves port in low visibility conditions.

[0022] Figure 6 This is a flowchart illustrating step S5 of a radar-based method for detecting obstacles when a ship enters or leaves port in low visibility conditions. Detailed Implementation

[0023] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0024] See Figure 1 A radar-based method for detecting obstacles in low visibility conditions for ships entering and leaving ports includes: S1, based on the radar signals currently transmitted by the ship, performing preliminary identification of obstacles in the current entry and exit channel using the signal characteristics of the radar signals, and labeling the preliminary identified obstacle types of each obstacle.

[0025] S2, based on the trajectory of the obstacle under continuous identification, calculates the direction angle and distance of the obstacle relative to the current ship, and after clustering the obstacles, forms an obstacle distribution heat map.

[0026] S3, based on the obstacle type corresponding to each obstacle, calculates the safe distance corresponding to each obstacle, and divides the obstacle identification boundary corresponding to each obstacle in the obstacle distribution heat map.

[0027] S4: Obtain the current vessel's navigation behavior when entering or leaving the port. If the navigation behavior meets the safe distance requirement, retrieve the estimated execution time of the corresponding navigation behavior and construct a behavior sequence.

[0028] S5, determine the inbound and outbound distances of the behavior sequence at the corresponding timestamp, perform segmentation correction according to the inbound and outbound distances, determine the control boundary for each segment of travel, and use the minimum intersection of the control boundary and the obstacle recognition boundary as the output data.

[0029] In the current scheme, multiple sets of point cloud data measured by radar signals are regarded as data sequences composed of multiple obstacles. Based on the location of their measurement points, they are labeled as various obstacles. The changes in the obstacles under multiple scans are used to gradually realize the point-line-surface processing method. The identified obstacles will be based on the distribution of fixed obstacles, semi-fixed obstacles, moving obstacles, and floating obstacles, thereby determining the current handling method for ships when entering and leaving the port.

[0030] When using a point-based processing approach, the radar signal characteristics of multiple obstacles are emphasized. It is necessary to perform preliminary obstacle identification based on the characteristic form presented by the radar signal at each point, and then complete the trajectory fitting of the point cloud data by considering positional deviation and periodic verification.

[0031] As for the processing method of lines, it is necessary to map the point cloud data onto a three-dimensional plane. The size presented on the three-dimensional plane is used to help verify the distribution of moving obstacles, thereby further verifying the type of obstacle corresponding to the point cloud data when they are aggregated into a surface.

[0032] As for the surface processing method, it involves dividing the area into boundaries corresponding to each obstacle, and then determining the intersection between each driving behavior and the boundary corresponding to the obstacle during navigation, ultimately obtaining a relatively safe waterway and set of driving behaviors.

[0033] Fixed obstacles are mostly channel structures and unmarked abnormal obstacles; semi-fixed obstacles are mostly anchored vessels and aquaculture areas / fishing nets, which are handled through boundary recognition; mobile obstacles are mostly other moving hulls, some of which communicate with each other and can be avoided by predicting their intentions, while others do not communicate and need to be actively avoided by predicting their trajectories; as for floating obstacles, they are small and large floating objects, which need to be avoided by selecting a channel based on their distribution density.

[0034] like Figure 2 As shown, the implementation of step S1 includes: S11, based on the radar signal currently transmitted by the ship, binding and calibrating the coordinate system of the radar signal with the coordinate system of the current ship, clarifying the correspondence between the detection angle of the radar signal and the left rudder, right rudder, and vertical direction of the ship; for example, 0° of the radar corresponds to the front of the ship, 90° corresponds to the right side of the ship, and -90° corresponds to the left side of the ship; ensuring that the current radar signal transmission situation corresponds to the current ship attitude.

[0035] S12 uses radar signal echo intensity, Doppler frequency shift, and echo shape as signal characteristics to identify the echo position of each obstacle. At this time, the echo intensity reflects the obstacle's ability to scatter radar waves and is related to the material, surface roughness, and size. The Doppler frequency shift can be used to calculate the relative velocity of the obstacle and distinguish between stationary (such as fixed obstacles) and moving targets (such as ships and floating objects). The echo shape is analyzed by pulse compression or synthetic aperture technology to infer the geometry of the obstacle (such as point targets and area targets). Based on these signal characteristics, the specific situation of the obstacle's position in and out of the port can be determined.

[0036] S13, based on the combination of signal characteristics, sets the initial obstacle type, and checks the initial obstacle type based on the change trajectory of the echo position, and uses the checked data as the output obstacle type; the initial obstacle type includes fixed obstacles, semi-fixed obstacles, moving obstacles and floating obstacles.

[0037] When signal features are combined, fixed obstacles often exhibit high echo intensity + no Doppler frequency shift + point / area targets. In this case, the echo intensity value of metal structures (such as shipwrecks and bridge piers) can be used to identify fixed obstacles at the corresponding locations. At the same time, the echo position of fixed obstacles remains unchanged in multiple radar scans.

[0038] Semi-fixed obstacles can be identified based on medium echo intensity + low-speed Doppler frequency shift + point / area targets. In this case, the semi-fixed obstacles will move slightly due to water flow or wind. The echo intensity can be identified based on the echo intensity values ​​of anchored ships, aquaculture areas / fishing nets. These echo intensity values ​​can be extracted from the database to identify the corresponding obstacles. At the same time, since the ships themselves are based on satellite AIS, their messages can be identified to determine whether they are ships.

[0039] As for moving obstacles, they are usually other moving hulls. In this case, the situation of the corresponding ship can be determined by combining high echo intensity, high-speed Doppler frequency shift, and point / area targets, and by introducing satellite AIS messages. The echo intensity can then be obtained directly from the echo intensity value of the ship.

[0040] The final floating obstacles will be based on low echo intensity + no / low speed Doppler frequency shift + point / linear target. At this time, the floating objects are mostly with small echo intensity and the identified targets are relatively small. According to the combination of their features, the obstacles that meet the current feature combination will be initially labeled.

[0041] At this point, the initial obstacle types will be marked using a feature combination method to complete the obstacle type configuration process. This involves initial settings based on radar signal feedback. This method offers high accuracy for fixed-type obstacles. In setting the obstacle distribution heatmap, the relevant trajectories of moving obstacles will be processed to obtain a relatively complete obstacle distribution.

[0042] When checking the initial obstacle type in step S13, the implementation method further includes: S131, for the initial obstacle type of each obstacle, based on the position of the obstacle in multiple scans, calculate the deviation between each scan and the initial position to obtain the echo position deviation sequence under the same initial obstacle type.

[0043] S132, based on the standard deviation and mean value corresponding to the echo position deviation sequence, periodic deviation verification is performed on fixed obstacles, semi-fixed obstacles and floating obstacles in sequence.

[0044] After initially labeling obstacles, the echo position deviations over multiple scans need to be processed. For example, if the position deviation of a fixed obstacle is less than 1 meter in multiple identifications, it is reclassified as semi-fixed or floating. For the verification of semi-fixed obstacles, it is often determined whether the echo position deviation exhibits periodic fluctuations, such as fishing nets which are easily affected by water currents and exhibit periodic movement behavior. For the verification of floating obstacles, it is determined whether the trajectory change of the echo position deviation is random. If it is random, it is a floating obstacle. In floating obstacles, the echo position deviation often exhibits an irregular, isotropic random distribution without significant periodicity or directionality.

[0045] The periodic deviation verification in step S132 is implemented as follows: For fixed obstacles, verification is performed based on the mean and standard deviation of the echo position deviation. At this time, historical data of fixed obstacles in the radar signal identification process can be introduced. The mean and standard deviation of the echo position deviation in the historical data are used as judgment conditions. If the mean and standard deviation of the current echo position deviation are both less than the mean and standard deviation of the historical data, it is confirmed as a fixed obstacle. Otherwise, it needs to be reclassified and marked as a semi-fixed obstacle or a floating obstacle.

[0046] For semi-fixed obstacles, if only one of the mean and standard deviation of the echo position deviation is greater than the mean and standard deviation of historical data, then the similarity between the echo position deviation sequence and the tidal-affected sequences in historical data is determined. If the similarity of the echo position deviation sequence is greater than the similarity threshold, it is confirmed as a semi-fixed obstacle, and its echo deviation sequence is considered to exhibit periodic changes. The periodic changes in the echo deviation sequence can be viewed using Fourier transform or Pearson correlation coefficient calculations. The similarity threshold can be set within a range of 0.6 to quantify the regularity of sequence similarity.

[0047] For floating obstacles, if both the mean and standard deviation of the echo position deviation are greater than the mean and standard deviation of historical data, the kurtosis and skewness of the echo position deviation are then calculated. If the skewness is close to 0 and the kurtosis is close to 3, it indicates that the echo position deviation conforms to a normal distribution. That is, the kurtosis and skewness of the echo position deviation are calculated for this part of the data. When the kurtosis and skewness conform to a normal distribution, the obstacle is divided into multiple sectors according to the direction corresponding to the echo position deviation. Each sector can be 30°. The number of echo position deviations calculated in each sector is counted, and a chi-square test is performed on these echo position deviations. If the chi-square test value is greater than 0.05, it indicates that the echo position within each sector is not significant and clearly shows a random distribution. Based on these clustering patterns, the distribution of the corresponding floating obstacles can be determined, thus completing the adjustment processing for obstacles in the initial identification scenario. The echo position deviation within each sector is tested separately with a chi-square test. Data whose echo position deviation does not meet the chi-square test is considered as floating obstacles. This completes the periodic deviation verification process, and yields verified fixed and floating obstacles.

[0048] S133, based on the data from the completed periodic deviation verification, and combined with the Doppler frequency shift values, corrects the abnormal motion state of the floating obstacle.

[0049] After completing the periodic verification data, the initial identification of fixed obstacles has been completed. The Doppler frequency domain will be introduced to check whether there are any parts of semi-fixed obstacles and floating obstacles with sudden increases in speed. For example, if the Doppler frequency shift corresponding to the echo position deviation suddenly increases, it indicates that the moving speed has suddenly increased, which does not conform to the normal fluctuation of water flow, and it needs to be marked as a moving obstacle.

[0050] The Doppler frequency shift reflects the relative radial velocity between the obstacle and the radar, and its value is... ;in, Indicates Doppler frequency shift; Represents the speed of light. Represents relative radial velocity, Indicates the radar frequency.

[0051] At this point, based on the timestamps corresponding to the echo position deviation sequence, the Doppler frequency shift at the corresponding time will be identified. The semi-fixed obstacles will be processed according to the Doppler frequency shift. For example, if the value of the Doppler frequency shift is greater than 1.5 times the historical average, it means that the corresponding obstacle has a significant acceleration. These floating obstacles need to be labeled and classified as moving obstacles.

[0052] S134, for moving obstacles, verify the consistency of the moving obstacle's trajectory, and use the verified data to complete the adjustment of the initial obstacle type.

[0053] When verifying moving obstacles, it is necessary to determine that the trajectory of the echo position deviation at the corresponding position is consistent with its moving path, thereby determining the specific trajectory of the moving obstacle.

[0054] The trajectory of a moving obstacle needs to be verified for consistency. This involves converting the echo position deviation into a continuous coordinate form, fitting the position of the moving obstacle using the least squares method or Kalman filter, and forming a trajectory curve. The trajectory curve is then compared with a known waterway, and the angle between the trajectory and the waterway is calculated. If the angle is less than 10°, it is confirmed as a moving obstacle, and it can be determined that the corresponding ship or other obstacle is traveling normally on the waterway. Otherwise, it is necessary to check whether it is a misclassified semi-fixed obstacle, thereby completing the adjustment of the initial obstacle type.

[0055] In one embodiment of the present invention, such as Figure 3 As shown, the implementation of step S2 includes: S21, converting the direction angle and distance of each identified obstacle into three-dimensional spatial coordinates in the ship coordinate system, where the x-axis is the ship's sailing direction, the y-axis is the ship's transverse direction (left / right rudder direction), and the z-axis is the vertical height above the water surface (used to distinguish between water surface obstacles and air interference).

[0056] S22, based on the three-dimensional spatial coordinates of each obstacle, the obstacles are filtered by the cross-sectional area of ​​the obstacle on each plane to determine the available obstacle coordinates.

[0057] At this point, the cross-sectional area of ​​each obstacle on the x, y, and z axes will be determined when the radar signal is reflected. Obstacle error tracking will be performed using the cross-sectional area of ​​each surface to determine the location of each obstacle at each signal reflection. This will enable the movement verification of obstacles on each plane, further verify the obstacle type, and make the subsequently generated obstacle recognition boundary more reliable.

[0058] The implementation of step S22 includes: S221, based on the three-dimensional spatial coordinates of the obstacle under continuous recognition, the obstacle is tracked to generate a set of trajectory points corresponding to each obstacle.

[0059] The location of each obstacle is identified by connecting radar signals. For example, radar signals are identified by continuous frames. The Kalman filter algorithm is used, with each frame interval of 0.1-0.2s, and the three-dimensional coordinates of 10-20 consecutive frames are tracked. The obstacle identified each time is determined, and the distance between obstacles in consecutive frames is compared. If the distance value is less than the average distance of continuous tracking under the corresponding obstacle type, it is marked as a valid point. These points are combined into a trajectory point set, and thus multiple consecutive obstacle coordinates are obtained.

[0060] S222, based on the generated trajectory point set, calculate the minimum bounding box containing all points, and extract the cross-sectional area of ​​each plane according to the method of horizontal plane, vertical plane and horizontal plane; the horizontal plane, vertical plane and horizontal plane are represented as xy plane, xz plane and yz plane respectively.

[0061] In the current scenario, it is required to use radar that supports elevation angle measurement, and to complete the cross-sectional identification of obstacles based on the distance, azimuth, and elevation angle measured by the radar. Feature matching calculation is performed based on the cross-sectional area of ​​each plane to identify large ships, small boats and rafts, single buoys, buoy groups, quay walls, birds in the air, and floating rocks that are likely to appear in port entry and exit scenarios. In this process, the focus is more on the processing and identification of moving obstacles to determine the available coordinate values ​​in each type of obstacle, and then obtain the coordinate center of the corresponding obstacle to assist in the clustering processing of obstacle distribution.

[0062] S223, based on the cross-sectional area of ​​each plane, perform feature matching on each obstacle, calculate the Euclidean distance between the cross-sectional area of ​​the current three planes and each obstacle type in the database, and select the obstacle type with the smallest Euclidean distance value as the current matched obstacle type. The feature matching value will be based on the cross-sectional area values ​​of various obstacles in the database to obtain the most similar obstacle type.

[0063] S224. If the obstacle type matched by the feature is consistent with the obstacle type initially identified, the verification of the moving obstacle is completed, and the trajectory point set of the obstacle is used as the output coordinate value.

[0064] S225, if the obstacle type matched by the feature is inconsistent with the obstacle type initially identified, a second verification is performed based on the vertical water surface height. When the obstacle type matched by the feature conforms to the range of the vertical water surface height, the obstacle type matched by the feature replaces the obstacle type initially identified, and the corresponding trajectory point set is used as the output coordinate value.

[0065] It should be noted that each type of obstacle has a specific vertical water level in the port entry and exit scenarios. For example, birds in the air generally satisfy z>5m, and buoys generally satisfy z∈[-0.5m,1m]. These values ​​can be used for further verification.

[0066] S226 If the value does not meet the range of vertical water surface height, the corresponding obstacle will be reprocessed until all obstacles are matched.

[0067] If the value does not conform to the range of vertical water surface height, it indicates that there is an abnormality in the type of obstacle matched by the current feature. It needs to be marked as an unknown obstacle and retained as a potential risk point. The corresponding obstacle needs to be re-judged and processed. After removing the noise points in the corresponding obstacle, all obstacles need to be checked.

[0068] S23. Based on the coordinate values ​​of each obstacle, perform spatial clustering with the coordinate center of the obstacle, and use the clustered obstacles as the output obstacle distribution heatmap.

[0069] When clustering obstacles using their coordinate centers, the coordinate centers are set using the center points corresponding to the trajectory point set. Multiple obstacles are then clustered using the DBSCAN algorithm to form multiple clusters. The clustering settings can be adjusted based on the current port conditions. For example, if the current vessels are large, the neighborhood radius needs to be increased appropriately, such as 5-8m. If the vessels are small, a neighborhood radius of 3-5m can be selected to improve the clustering processing for different types of obstacles. Then, spatial clustering is performed on all obstacles based on this neighborhood radius. The minimum number of points in each cluster can be set to 8-10 or 3-5. If there are many vessels and buoys in the current port, 8-10 points can be selected. If there are few buoys in the current port, 3-5 points can be selected as the minimum number of points. Then, the obstacles are clustered one by one according to their coordinate centers, and finally, the heat map configuration of the obstacle clustering is completed.

[0070] In one embodiment of the present invention, when delineating boundaries, obstacle identification boundaries are configured for each ship's travel path based on the obstacle type and the safe distance between the obstacle and the current waterway; for example, for the boundary position after obstacle clustering, the boundary of the obstacle body plus the space boundary required for ship avoidance are used to obtain the set obstacle identification boundary.

[0071] like Figure 4 As shown, the implementation of step S3 includes: S31, extracting the distribution location of each obstacle from the obstacle distribution heat map, and configuring the geometric contour boundary corresponding to each obstacle; the geometric contour boundary is set based on the cluster of the corresponding obstacle in the obstacle distribution heat map, taking the three-dimensional spatial coordinates of the cluster as the reference, extracting the maximum contour range of the cluster in the xy horizontal plane of the ship's navigation, marking the extreme points of the outer edge of the cluster and connecting them to form a geometric contour boundary that fits the physical size and shape of the obstacle, and this boundary is the boundary of the obstacle body.

[0072] In the current step of processing, specific boundary ranges will be defined for moving obstacles, and obstacles other than moving obstacles will be set using a combination of body boundary and avoidance boundary. The boundary of moving obstacles will be set with buffer boundary based on the obstacle's motion state, the direction angle and speed corresponding to its trajectory, and thus complete its boundary configuration.

[0073] S32, calculate the safe distance of the ship based on the current speed of the ship; based on the geometric contour boundary of the obstacle, extend the length of the safe distance outward to form the obstacle recognition boundary; at this time, the obstacle recognition boundary will take into account the safe distance of the ship at the corresponding speed, thereby realizing the boundary recognition of obstacle avoidance.

[0074] First, the length of the safe distance needs to be calculated based on the current speed and the scene. Then, based on the length of the safe distance and the geometric contour of the obstacle, an obstacle recognition boundary with an avoidance area is formed.

[0075] Therefore, the method for calculating the safe distance of the ship in step S32 includes: S321, setting a basic safe distance based on the length and width of the ship.

[0076] When calculating safety distances, a basic safety distance is generally set based on the length and width of the vessel, such as: basic safety distance = k × (sum of length and width of the vessel), where k is an empirical coefficient. Its value range is based on the ratio of the vessel's size to the size of a standard hull, and is selected according to a range of 0.5-1.5. The dimensions of the standard hull will be adjusted according to the average size of the corresponding vessel type to complete the configuration treatment for each vessel type. For example, for small vessels with a length <50 meters, whose length-to-width ratio is between 3.0 and 4.0, the k value can be set to 0.5-0.7 to cope with small vessels that are maneuverable and have short stopping distances. For medium-sized vessels (50-120 meters in length, with a length-to-beam ratio between 5.0 and 6.0), the k-value can be set to 0.7-0.9 to accommodate medium-sized vessels with standard maneuverability. For large vessels (120-200 meters in length, with a length-to-beam ratio between 6.0 and 7.5), the k-value can be set to 0.9-1.2 to accommodate large vessels with high inertia and slow response. For very large vessels (>200 meters in length, with a length-to-beam ratio between 7.0 and 9.0), the k-value can be set to 1.2-1.5 to accommodate small vessels with poor maneuverability and requiring a larger safety margin. This completes the calculation of the basic safety distance.

[0077] S322 decomposes the current ship trajectory into a longitudinal component in the bow-stern direction and a lateral component in the port-starboard direction.

[0078] Then, the safe distance is adjusted based on the speed of the ship, and environmental factors, such as water flow speed and wind direction, are also taken into account. The safe distance is then adjusted by vector superposition.

[0079] When using speed to correct the basic safety distance, the current ship motion is decomposed into two components: a longitudinal component in the bow-stern direction and a lateral component in the port-starboard direction. The lateral component mainly affects ship drift, while the longitudinal component mainly affects speed and stopping distance.

[0080] S323, based on the ship's average reaction time to avoid obstacles, constructs a lateral correction distance using the wind speed and water speed as the lateral components, and constructs a longitudinal correction distance using the current ship speed; based on the vector superposition of the lateral and longitudinal correction distances, the corrected safe distance is obtained.

[0081] The lateral component incorporates the wind and water speeds received by the ship in the lateral direction. It examines the reaction time when the corresponding obstacle is detected and multiplies the reaction time by the corresponding wind and water speeds to obtain the drift distance that can be generated in the lateral direction. The longitudinal component calculates the product of the ship's speed and the reaction time at the corresponding reaction time, thus obtaining the additional displacement generated in the two components. These two displacements are then vector-superimposed to form a distance value, and the corrected safety distance is obtained by adding the base safety distance to this distance value.

[0082] It should be noted that reaction time represents the time interval between when a ship identifies an obstacle and determines that there is a collision risk via radar, and when the ship's steering gear / propellers execute collision avoidance steering / deceleration actions. When an obstacle is directly identified by radar, the average reaction time from historical navigation behavior will be used as the current reaction time to quantify the current impact in the lateral and longitudinal directions.

[0083] S324: Based on the current port channel, determine the available channel width and the channel occupancy width for ships, and calculate the maximum permissible safe distance for current ships. Based on the maximum permissible safe distance for current ships, the upper limit of the safe distance, and the corrected safe distance, the minimum value of the three is taken as the output safe distance.

[0084] When outputting the corrected safety distance, it is necessary to determine the available channel width and the width occupied by the vessel in the channel. The available channel width is expressed as the difference between the nominal channel width and the safety margin at the channel edge (usually 10-20m). The width occupied by the vessel in the channel can be calculated based on the current width of the vessel, such as the vessel width + 2 × (lateral maneuvering coefficient × vessel length between perpendiculars × sin(maximum rudder angle)). The lateral maneuvering coefficient used in this case will be selected from 0.03 to 0.05 depending on the visibility conditions. 0.5 is selected when visibility is low to represent the lateral turning distance of the vessel. Lateral drift correction; the length between the ship's perpendiculars is the core ship dimension that determines lateral drift. As for the ship's maximum rudder angle, it quantifies the impact of the rudder angle on lateral drift, and thus explains the width that the ship needs to occupy when sailing. Then, the maximum permissible safe distance is calculated by combining the available channel width and the ship's channel occupancy width. The maximum permissible safe distance represents the current distance that the ship can travel. The maximum permissible safe distance is expressed as (available channel width - ship's channel occupancy width) / 2 × 0.8. Here, multiplying by 0.8 means reserving a 20% safety margin for emergency operations when the ship avoids obstacles.

[0085] The final output safety distance is determined by selecting the minimum value from the safety distance corrected for wind and water speed, the maximum permissible safety distance, and the upper limit of the absolute permissible safety distance of the system. This minimum value is then used to determine the obstacle recognition boundary. The upper limit of the safety distance is generally 300m. This upper limit represents the specific value calibrated by the current port, and its calibrated value can be directly viewed from the database based on the port being navigated.

[0086] In one embodiment of the present invention, in step S4, the coordinate changes of obstacles triggered by the driving behavior are associated, the specific driving actions of the current ship are enumerated, such as left rudder 15° / 30°, right rudder 15° / 30°, straight navigation, deceleration / acceleration, etc., and the distance and time between the obstacle and the ship in this part of the driving behavior are used as labels, the safe distance at any time is recorded, and thus a behavior sequence associated with the obstacle is formed.

[0087] like Figure 5 As shown, the implementation of step S4 also includes: S41, for any driving behavior, calculate the real-time distance between the obstacle and the ship according to the current driving direction of the ship, and construct a real-time distance sequence.

[0088] S42, the minimum value in the distance sequence is regarded as the nearest encounter distance, the time point corresponding to the nearest encounter distance is regarded as the nearest encounter time, and the nearest encounter time and the nearest encounter distance are associated with driving behavior.

[0089] S43. Compare the nearest encounter distance with the safe distance of the corresponding obstacle. If it is greater, the current driving behavior is considered to meet the safe distance requirement, and the corresponding driving behavior is sorted into a behavior sequence in ascending order of the nearest encounter time. If it is not greater, the corresponding driving behavior needs to be removed to avoid the risk of collision when entering or leaving the port.

[0090] The nearest encounter distance represents the minimum distance between the ship and the obstacle within the predicted time period. It is a quantitative value used to check whether a collision will occur within a specific time period for the current driving behavior. The nearest encounter time represents the time to reach the nearest encounter distance. It is a time value used to further assist in the enumeration of driving behaviors.

[0091] The estimated execution time describes the time required for the vessel to complete the corresponding driving behavior. The real-time distance between the vessel and the obstacle is determined, and the nearest encounter distance and the nearest encounter time are defined to explain whether there is a collision risk in the current driving behavior, and to record the data on the collision risk and the relevant specific behavior content.

[0092] The output behavior sequence is used for risky behaviors where the nearest encounter distance is less than the obstacle's safe distance. Based on the relevant driving conditions in the risky behavior, data with collision risk is removed. Then, the estimated execution time required for the driving behavior and the nearest encounter time are recorded separately. At the same time, the nearest encounter time represents the moment when the ship is closest to the obstacle after the driving behavior is executed. The smaller the value of this time, the stronger the timeliness of the ship's obstacle avoidance, which can make the current behavior sequence more reliable in avoiding obstacle collisions.

[0093] In one embodiment of the present invention, step S5 is set according to the distance between the control boundary and the obstacle identification boundary. The control boundary represents the channel width that the ship can freely use under the corresponding driving behavior, that is, it extends to both sides from the current coordinate center of the ship, and the length of the extension is the channel width occupied by the ship. The obstacle identification boundary represents the boundary range affected by the obstacle. When the intersection between the channel width occupied by the ship and the boundary range affected by the obstacle reaches the minimum, it means that the current driving behavior can effectively avoid the obstacle at the corresponding position, thereby completing the obstacle detection processing under low visibility.

[0094] It should also be noted that the channel occupancy width is adjusted based on the distance to and from the port and the maximum rudder angle of the vessel to obtain the channel occupancy width under different directional angles. Furthermore, since the port entry and exit are divided into multiple segments based on the distance between vessels, such as the near-port segment within 300m of the port, the mid-port segment within 300-700m of the port, and the far-port segment within 700-1000m of the port, the safe distance corresponding to obstacles will vary in each port segment, thus causing the obstacle identification boundary to change. At this time, the channel occupancy width and obstacle identification boundary will be determined by finding the channel corresponding to the minimum area under continuous conditions based on the area of ​​the boundary intersection, thereby achieving the optimal distribution processing for obstacle avoidance.

[0095] like Figure 6 As shown, the implementation of step S5 includes: S51, based on the current ship's channel occupancy width and direction angle, extending to both sides from the current ship's coordinate center to form the control boundary corresponding to the navigation behavior; the control boundary formed at this time will be extended to both sides according to the ship's direction angle at the corresponding position, thereby obtaining the corresponding control boundary.

[0096] When the control boundary is formed, a local coordinate system is established with the ship's coordinate center as the origin (0, 0), ensuring that the current direction angle always coincides perfectly with the ship's x-axis (ship's longitudinal axis) and points in real time towards the direction of the ship's forward movement. Then, the control boundary on the corresponding longitudinal axis is set based on the product of the current speed and the expected execution time of the corresponding driving behavior. The control boundary on the horizontal y-axis is represented by half the width of the ship's channel occupancy. Assuming the ship's channel occupancy width is... At this point, the control boundary marked in the local coordinate system can be 0 ≤ x ≤ speed × estimated execution time. ≤y≤ This completes the initial setting of the control boundary.

[0097] S52, based on the current port entry and exit distance of the vessels, is divided into near-port section, mid-port section and far-port section, and the boundary adjustment amount of the control boundary under each section is determined.

[0098] It is known that the near-port section is the core channel section for ships departing from or about to berth. The channel has the smallest available width and the densest distribution of obstacles, such as wharves, buoys, and reefs. In addition, the wind and water currents are complex and have a significant impact. It is the channel section with the highest risk of collision for ships entering and leaving the port, and also the section with the highest safety redundancy requirements. At this time, the control boundary needs to be strongly constrained and corrected. The size of the control boundary is adjusted according to the real-time direction and angle. For example, based on the current size of the control boundary, the lateral and longitudinal directions are adjusted respectively. The lateral direction is set to 1.3 times the original boundary, and the longitudinal direction is set to 0.5 times the original boundary to form a narrow and short control boundary. This maximizes the distance between the outer edge of the control boundary and the ship's hull, thereby improving the ship's safety redundancy. Ultimately, the intersection area calculated for the near-port section is minimized among the three segments.

[0099] As for the Zhonggang section, it is the core channel for ships entering and leaving the port. The channel has a moderate usable width, sparse and simple obstacles, and weak wind and water flow effects. It is a transitional section for ship navigation with a moderate collision risk. At this time, the configured control boundaries will not be modified.

[0100] Finally, the outer harbor section is the section where ships leave or are about to enter the main port channel. This section has the widest usable width, no dense obstacles, and minimal wind and current influence, making it the channel section with the lowest risk of collisions when ships enter or leave the port. At this point, the original control boundary will be reduced to decrease the constraints on ship movement. For example, the control boundary will be adjusted laterally to 0.8 times its original width, and longitudinally extended to 1.2 times its original width, thus completing the adjustment of the control boundary.

[0101] S53, for obstacle identification boundaries, add redundant warning boundaries at the obstacle identification boundaries of each segment. When the intersection area of ​​the control boundary and the redundant warning boundary is greater than a preset threshold, determine the boundary adjustment amount of the obstacle identification boundary by the safe distance between the ship and the obstacle under the corresponding intersection area.

[0102] When the intersection area is greater than a preset threshold, it means that the current ship has entered the warning area of ​​the obstacle recognition boundary and has a certain degree of overlap, which is prone to collision risk. The set obstacle recognition boundary will be partially contracted inward to determine whether the intersection area falls below the preset threshold after removing some redundant safety boundaries, thereby completing the redundancy warning in low visibility scenarios.

[0103] For example, redundant warning boundaries of 0.3 times the safety distance, 0.2 times the safety distance, and 0.1 times the safety distance are configured for near-port, mid-port, and far-port sections, respectively. These redundant warning boundaries are boundaries that extend uniformly outward from the obstacle recognition boundary. The preset threshold is the average value of the intersection area between the control boundary and the redundant warning boundary under normal driving conditions. When the value exceeds this average value, it indicates that the obstacle recognition boundary is extremely close, and a certain warning is required to prevent the risk of subsequent collision.

[0104] As for the boundary adjustment amount for obstacle recognition boundaries, it will be corrected by adjusting the extension distance corresponding to the redundant warning boundary, rather than directly adjusting the size of the obstacle recognition boundary. The obstacle recognition boundary is the boundary outline formed by the safe distance and the obstacle itself, which represents the safety margin under safe driving conditions. It is not possible to directly reduce its boundary size inward, thereby causing the problem of insufficient safety margin.

[0105] When adjusting the redundant warning boundary, the length of the original redundant warning boundary will be gradually adjusted by 20% until the current intersection area falls below the preset threshold. Then, the length of the current redundant warning boundary will be regarded as the boundary adjustment amount of the obstacle recognition boundary.

[0106] S54 performs an intersection operation on the adjusted control boundary and obstacle recognition boundary, and uses the data with the smallest intersection area as the minimum intersection to complete the parameter configuration for driving behavior in port entry and exit scenarios.

[0107] In step S54, the intersection area values ​​of each segment are finally counted throughout the entire port entry and exit process. Then, the driving behavior that satisfies the minimum intersection and the driving behavior under the corresponding boundary length are selected, so as to know the way to complete obstacle detection and avoidance in low visibility scenarios.

[0108] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered by the protection scope of the present invention.

Claims

1. A radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions, characterized in that, include: S1, based on the radar signals currently transmitted by the ship, uses the signal characteristics of the radar signals to make a preliminary identification of obstacles in the current port entry and exit channels, and marks the type of obstacle identified in the preliminary identification of each obstacle; S2, based on the trajectory of the obstacle under continuous recognition, calculate the direction angle and distance of the obstacle relative to the current ship, and after clustering the obstacles, form an obstacle distribution heat map; S3, based on the obstacle type corresponding to each obstacle, calculate the safe distance corresponding to each obstacle, and divide the obstacle identification boundary corresponding to each obstacle in the obstacle distribution heat map; S4: Obtain the current vessel's navigation behavior when entering or leaving the port. If the navigation behavior meets the safe distance requirement, retrieve the estimated execution time of the corresponding navigation behavior and construct a behavior sequence. S5, determine the inbound and outbound distances of the behavior sequence at the corresponding timestamp, perform segmentation correction according to the inbound and outbound distances, determine the control boundary for each segment of travel, and use the minimum intersection of the control boundary and the obstacle recognition boundary as the output data.

2. The radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions, as described in claim 1, is characterized in that... The implementation methods for step S1 include: S11, based on the radar signals currently transmitted by the ship, bind and calibrate the coordinate system of the radar signals with the coordinate system of the current ship, and clarify the correspondence between the detection angle of the radar signals and the ship's left rudder, right rudder, and vertical direction. S12 uses the echo intensity, Doppler frequency shift, and echo shape of radar signals as signal characteristics to identify the echo location of each obstacle; S13, based on the combination of signal characteristics, sets the initial obstacle type, and checks the initial obstacle type based on the change trajectory of the echo position, and uses the checked data as the output obstacle type; the initial obstacle type includes fixed obstacles, semi-fixed obstacles, moving obstacles and floating obstacles.

3. The radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions according to claim 2, characterized in that, When checking the initial obstacle type in step S13, the implementation method also includes: S131, For each obstacle's initial obstacle type, based on the obstacle's position in multiple scans, calculate the deviation between each scan and the initial position to obtain the echo position deviation sequence under the same initial obstacle type; S132, based on the standard deviation and mean value corresponding to the echo position deviation sequence, periodic deviation verification is performed on fixed obstacles, semi-fixed obstacles and floating obstacles in sequence; S133, based on the data from the periodic deviation verification, and combined with the value of the Doppler frequency shift, corrects the abnormal motion state of the floating obstacle; S134, for moving obstacles, verify the consistency of the moving obstacle's trajectory, and use the verified data to complete the adjustment of the initial obstacle type.

4. The radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions according to claim 3, characterized in that, The periodic deviation verification in step S132 is implemented in the following ways: For fixed obstacles, the mean and standard deviation of the echo position deviation are used for verification. If the mean and standard deviation of the current echo position deviation are both less than the mean and standard deviation of the historical data, then it is confirmed as a fixed obstacle. For semi-fixed obstacles, if only one of the mean and standard deviation of the echo position deviation is greater than the mean and standard deviation of the historical data, then the similarity between the echo position deviation sequence and the tidal-affected sequence in the historical data is judged. If the similarity of the echo position deviation sequence is greater than the similarity threshold, then it is confirmed as a semi-fixed obstacle. For floating obstacles, if both the mean and standard deviation of the echo position deviation are greater than the mean and standard deviation of historical data, calculate the kurtosis and skewness of the echo position deviation. If the kurtosis and skewness conform to a normal distribution, data whose echo position deviation does not meet the chi-square test are considered as floating obstacles.

5. The radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions according to claim 1, characterized in that, Step S2 can be implemented in the following ways: S21, convert the direction, angle and distance of each identified obstacle into three-dimensional spatial coordinates in the ship coordinate system; S22, based on the three-dimensional spatial coordinates of each obstacle, the obstacles are filtered by the cross-sectional area of ​​the obstacle on each plane to determine the coordinates of the available obstacles; S23. Based on the coordinate values ​​of each obstacle, perform spatial clustering with the coordinate center of the obstacle, and use the clustered obstacles as the output obstacle distribution heatmap.

6. The radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions according to claim 5, characterized in that, The implementation methods of step S22 include: S221, Based on the three-dimensional spatial coordinates of the obstacle under continuous recognition, the obstacle is tracked to generate a set of trajectory points corresponding to each obstacle; S222, based on the generated trajectory point set, calculate the minimum bounding box containing all points, and extract the cross-sectional area of ​​each plane according to the method of horizontal plane, vertical plane and horizontal plane; S223, based on the cross-sectional area of ​​each plane, perform feature matching on each obstacle, calculate the Euclidean distance between the cross-sectional area of ​​the current three planes and each obstacle type in the database, and select the obstacle type with the smallest Euclidean distance value as the current matched obstacle type; S224, If the obstacle type matched by the feature is consistent with the obstacle type initially identified, then the verification of the moving obstacle is completed, and the trajectory point set of the obstacle is used as the output coordinate value. S225, If the obstacle type matched by the feature is inconsistent with the obstacle type initially identified, a second verification is performed based on the vertical water surface height. When the obstacle type matched by the feature conforms to the range of the vertical water surface height, the obstacle type matched by the feature replaces the obstacle type initially identified, and the corresponding trajectory point set is used as the output coordinate value. S226 If the value does not meet the range of vertical water surface height, the corresponding obstacle will be reprocessed until all obstacles are matched.

7. The radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions according to claim 1, characterized in that, Step S3 can be implemented in the following ways: S31, extract the distribution location of each obstacle from the obstacle distribution heat map, and configure the corresponding geometric contour boundary of each obstacle; S32, calculate the safe distance of the ship based on the current speed of the ship; based on the geometric contour boundary of the obstacle, extend the length of the safe distance outward to form the obstacle recognition boundary.

8. The radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions according to claim 7, characterized in that, The methods for calculating the safe distance of the vessel in step S32 include: S321, establishes basic safety distances based on the length and width of the vessel; S322 decomposes the current ship trajectory into a longitudinal component in the bow-stern direction and a lateral component in the port-starboard direction; S323, based on the ship's average reaction time to avoid obstacles, constructs the lateral correction distance using the wind speed and water speed as the lateral components, and constructs the longitudinal correction distance using the current ship speed; the corrected safe distance is obtained by vector superposition of the lateral and longitudinal correction distances. S324: Based on the current port channel, determine the available channel width and the channel occupancy width for ships, and calculate the maximum permissible safe distance for current ships. Based on the maximum permissible safe distance for current ships, the upper limit of the safe distance, and the corrected safe distance, the minimum value of the three is taken as the output safe distance.

9. The radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions according to claim 1, characterized in that, The implementation of step S4 also includes: S41, for any driving behavior, calculate the real-time distance between the obstacle and the ship based on the current direction of the ship's travel, and construct a real-time distance sequence; S42, take the minimum value in the distance sequence as the nearest meeting distance, take the time point corresponding to the nearest meeting distance as the nearest meeting time, and associate the nearest meeting time and the nearest meeting distance with driving behavior; S43. Compare whether the nearest encounter distance is greater than the safe distance of the corresponding obstacle. If it is greater, the current driving behavior is considered to meet the safe distance. Sort the corresponding driving behaviors in ascending order of the nearest encounter time into a behavior sequence.

10. The radar-based method for detecting obstacles to ships entering and leaving port in low visibility conditions according to claim 1, characterized in that, Step S5 can be implemented in the following ways: S51, based on the current ship's channel occupancy width and direction angle, extends to both sides from the current ship's coordinate center to form the control boundary corresponding to the navigation behavior; S52, based on the current port entry and exit distance of the vessels, is divided into near port section, mid port section and far port section in sequence, and the boundary adjustment amount of the control boundary under each section is determined; S53, For obstacle identification boundaries, a redundant warning boundary is added at the obstacle identification boundary of each segment. When the intersection area of ​​the control boundary and the redundant warning boundary is greater than a preset threshold, the boundary adjustment amount of the obstacle identification boundary is determined by the safe distance between the ship and the obstacle under the corresponding intersection area. S54 performs an intersection operation on the adjusted control boundary and obstacle recognition boundary, and uses the data with the smallest intersection area as the minimum intersection.

Images