Intelligent anti-collision early warning method and system for ship loader

Abstract

The invention relates to the field of data processing, in particular to an intelligent anti-collision early warning method and system for a ship loader, and the method comprises the steps: obtaining the preprocessed continuous multi-frame point cloud data of the ship loader and the surrounding area of the ship loader; calculating the local density of each point cloud data, and setting an initial clustering radius according to the local density; selecting any point cloud data as a target point cloud, acquiring similarity values of all neighborhood point clouds in an initial clustering radius range of the target point cloud, and adjusting the initial clustering radius of the target point cloud by using the similarity values to obtain an optimal clustering radius; and clustering and segmenting the single-frame point cloud data into different clusters by using the optimal clustering radius, each cluster representing a target, calculating the distance between the target and the chute tube, and triggering anti-collision early warning according to the distance. According to the invention, accurate identification and distance measurement of targets around the ship loader are realized, and the accuracy of anti-collision early warning and the safety of ship loading operation are improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing. More specifically, the present invention relates to an intelligent anti-collision warning method and system for a ship unloader. Background Art

[0002] A ship unloader is a large-scale mechanical device used in a bulk cargo terminal. Its main function is to load bulk materials (such as coal, ore, grain, etc.) from the storage area of the terminal into the cargo hold of a cargo ship. The chute is a key component of the ship unloader and is located at the terminal position of the ship unloader. Its role is to guide the materials from the conveying system of the ship unloader into the cargo hold. Since the chute needs to move continuously to achieve uniform stowage in the hold, and the draft of the ship changes dynamically with the change of the loading volume, the relative position between the chute and the cargo hold is also constantly changing. In order to avoid collision accidents between the chute and the cargo hold and ensure the safety and efficiency of the ship loading operation, an anti-collision warning system becomes essential.

[0003] Currently, the anti-collision warning system usually uses radar to obtain three-dimensional point cloud data in the working environment of the ship unloader. The system will segment these point cloud data to identify target obstacles (such as the cargo hold wall, etc.) and calculate the distance between these target obstacles and the chute, and issue an anti-collision warning based on the distance.

[0004] However, since the point cloud data obtained by the radar often contains a large amount of noise and interference information, such as changes in ambient light, differences in the reflection characteristics of objects, errors of radar equipment, etc., these noise and interference information will interfere with the accuracy of the point cloud data, making it difficult for the segmentation algorithm to accurately identify obstacles, and then causing the anti-collision system to issue unnecessary warning signals, thus interfering with the normal operation of the ship unloader. Summary of the Invention

[0005] To solve the above technical problems of being unable to accurately identify obstacles, resulting in the anti-collision system issuing unnecessary warning signals and interfering with the normal operation of the ship unloader, the present invention provides solutions in the following aspects.

[0006] In a first aspect, an intelligent anti-collision warning method for a ship unloader includes: Obtaining continuous multi-frame point cloud data of the ship unloader and a certain area around it after preprocessing, where the certain area around includes the chute, the cargo hold, and the goods; Calculating the local density of each point cloud data and setting an initial clustering radius according to the local density; Select any point cloud data as the target point cloud, obtain all the neighborhood point clouds within the initial clustering radius of the target point cloud, calculate the similarity values between the target point cloud and each of the neighborhood point clouds, adjust the initial clustering radius of the target point cloud using the similarity values to obtain the optimal clustering radius of the target point cloud, and further obtain the optimal clustering radii of all the point cloud data; the similarity value is positively correlated with the velocity consistency between the target point cloud and its neighborhood point cloud and negatively correlated with the distance between the target point cloud and the neighborhood point cloud. Use the optimal clustering radius to cluster and segment the point cloud data of a single frame into different clusters, where each cluster represents a target, calculate the distance between the target and the chute, compare this distance with a preset threshold, and if the distance is less than the threshold, trigger a collision warning.

[0007] The present invention first calculates the initial clustering radius based on a preset clustering radius reference value and local density, enabling the clustering to adapt to regions with different densities. Secondly, it tracks the velocity consistency and distance of consecutive multi-frame point cloud data and dynamically adjusts the initial clustering radius. Even if there are temporarily docked goods near the chute (with a short spatial distance but a large velocity difference), it can segment them into different clusters to avoid misjudgment. Finally, using the optimal clustering radius for clustering can accurately separate key targets such as ship loaders, chutes, and goods, avoiding deviations in distance calculation caused by incorrect clustering.

[0008] In summary, the present invention improves the accuracy of segmentation and distance detection by dynamically adjusting clustering parameters and combining multi-modal information (space and velocity), thereby effectively triggering a warning and enhancing operation safety.

[0009] Preferably, the point cloud data includes the three-dimensional coordinates and reflection intensity of the point cloud data.

[0010] Preferably, the process of obtaining the local density includes: Normalize the three-dimensional coordinates and reflection intensity of the point cloud data and use kernel density estimation to calculate the local density of each point cloud data. Among them, a Gaussian kernel function is selected and the bandwidth of the kernel density estimation is calculated.

[0011] Kernel density estimation is a non-parametric method that does not depend on the prior distribution assumption of the data and is applicable to various complex and unknown distribution forms. For point cloud data, its distribution may be very complex, and kernel density estimation can flexibly estimate its local density.

[0012] Preferably, the initial clustering radius satisfies the relational expression: ; in the formula, is the initial clustering radius of the th point cloud data, is the clustering radius reference value, is the global density of all point cloud data, is the local density of the th point cloud data; among them, the global density of all point cloud data is the mean value of the local densities of all point cloud data.

[0013] In point cloud data, noise and outliers may cause abnormal fluctuations in local density. By introducing the constraint of global density, the tolerance of the algorithm to these local anomalies will be enhanced, thereby improving the robustness of the clustering results.

[0014] Preferably, the similarity satisfies the relational expression: ; in the formula, is the similarity value between the th target point cloud and the th neighborhood point cloud, is the velocity consistency between the th target point cloud and the th neighborhood point cloud, is the Euclidean distance between the th target point cloud and the th neighborhood point cloud, is the exponential function with the natural constant e as the base; among them, is obtained by calculating the cosine similarity between the velocity vectors of the th target point cloud and the th neighborhood point cloud.

[0015] By combining velocity consistency and distance information, it is possible to more accurately determine whether two point clouds belong to the same object or are related. Especially in dynamic scenes, velocity information can help distinguish moving targets and avoid mis-matching.

[0016] Preferably, the optimal clustering radius satisfies the relational expression: ; in the formula, is the optimal clustering radius of the th target point cloud, is the initial clustering radius of the th target point cloud, is the similarity value between the th target point cloud and the th neighborhood point cloud, is the total number of neighborhood points of the target point cloud, is the exponential function with the natural constant e as the base.

[0017] By introducing the weighted average of similarity values, the formula can dynamically adjust the initial clustering radius according to the local features of the point cloud. In areas with high similarity, the clustering radius is larger and can include more points; in areas with low similarity, the clustering radius is smaller to avoid including too many dissimilar points. This can improve the accuracy and efficiency of clustering.

[0018] Preferably, obtaining the optimal clustering radius of all point cloud data further includes: Using the optimal clustering radius to segment the point cloud data of a single frame to obtain multiple clustering clusters; selecting any two consecutive frames of point cloud data and aligning the two consecutive frames of point cloud data; For the point cloud data within each clustering cluster in the first frame, calculate the information entropy of the clustering labels of the corresponding point cloud data in the second frame; Establish an objective function, input the sum of all information entropies, and output the optimal clustering radius reference value; obtain the adjusted optimal clustering radius based on the optimal clustering radius reference value.

[0019] The adjustment of the clustering radius is an iterative optimization process. The clustering radius reference value provides an initial reference, and the introduction of local features makes the initial clustering radius closer to the actual distribution of the data. However, to further improve the clustering accuracy and stability, the optimized initial radius result needs to be fed back to the adjustment of the reference value. This iterative optimization can continuously approach the optimal clustering radius, thereby obtaining more accurate clustering results.

[0020] Preferably, the expression of the objective function is: ; where is the objective function, is the information entropy of the clustering labels corresponding to the point cloud data in the second frame for the th clustering cluster in the first frame, is the number of clustering clusters in the first frame, is the optimal clustering radius reference value.

[0021] Preferably, the clustering is the DBSCAN clustering algorithm.

[0022] In a second aspect, a ship unloader intelligent anti-collision warning system includes: a processor and a memory, and the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the ship unloader intelligent anti-collision warning method according to any one of the above is implemented.

[0023] The beneficial effects of the present invention are: Through point cloud data processing and clustering analysis, the present invention continuously optimizes clustering parameters to achieve real-time monitoring of the ship loader and its surrounding environment (such as the chute, the cabin, and the cargo), thereby effectively preventing the ship loader from colliding with other objects during operation and improving the safety and efficiency of the ship loading operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 FIG. is a flowchart of the method from step S1 to step S4 in an intelligent anti-collision early warning method for a ship loader according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.

[0026] The following will describe in detail the specific embodiments of the present invention in conjunction with the accompanying drawings.

[0027] Referring to Figure 1 , an intelligent anti-collision early warning method for a ship loader includes steps S1 to S4, which are specifically as follows: S1: Obtain continuous multiple frames of point cloud data of the ship loader and a certain area around it, where the certain area around it includes the chute, the cabin, and the cargo.

[0028] In one embodiment, a lidar device is installed on the chute maintenance platform of the ship loader, and the lidar is used to collect point cloud data of the area around the ship loader in real time. These areas include the chute, the cabin, and the cargo in the cabin. The collected point cloud data includes the three-dimensional coordinates (i.e., x, y, z position information) of each point and the reflection intensity (i.e., the energy intensity returned after the laser beam irradiates the object surface, which can be used to judge the material, color, and other characteristics of the object).

[0029] In order to capture the dynamic information of the ship loader and the surrounding chute, cabin, and cargo more comprehensively and accurately, the lidar device needs to continuously collect data. Exemplarily, a frame of point cloud data is obtained at a certain time interval (for example, every 0.1 second), and multiple frames (such as 100 frames) of data are continuously collected.

[0030] Specifically, for each frame of data, according to the three-dimensional coordinate information of each point it contains, all points are organized together according to a certain rule (such as spatial position relationship or scanning order) to form an ordered point set.

[0031] Since the data obtained by the lidar may contain noise, methods such as mean filtering can be used to remove the noise data.

[0032] S2: Calculate the local density of each point cloud data, and set the initial clustering radius according to the local density.

[0033] In port operations, the ship loader needs to maintain a safe distance from various surrounding equipment and objects to avoid collision accidents. After clustering and segmenting the point cloud data, the distance and positional relationship between the ship loader and surrounding objects can be monitored in real time, potential collision risks can be detected in a timely manner, and corresponding obstacle avoidance measures can be taken to ensure the safe progress of operations.

[0034] The local density of a point is a very important factor because it can reflect the spatial distribution characteristics of the point cloud data. In areas with high density, the distance between points is relatively close, so a smaller clustering radius is sufficient to contain enough points to form a meaningful cluster. On the contrary, in areas with low density, the distance between points is relatively far, and a larger clustering radius is required to contain enough points.

[0035] Therefore, when calculating the clustering radius of point cloud data, the local density of the point cloud data needs to be considered.

[0036] In one embodiment, the general steps for calculating the local density of point cloud data are as follows: First, obtain the four-dimensional data (x, y, z, reflection intensity) of each point cloud data, and further normalize to scale the data of these dimensions to the same scale.

[0037] Then, calculate the local density of each point cloud data through kernel density estimation. Among them, a Gaussian kernel function is selected, and the Silverman's rule is used to calculate the bandwidth of the kernel density estimation.

[0038] It should be noted that kernel density estimation is a prior art and will not be elaborated here too much.

[0039] Next, according to the local density of each point cloud data, set an initial clustering radius for it, that is, the relational expression is satisfied as:

[0040] In the formula, is the initial clustering radius of the th point cloud data, is the clustering radius reference value, is the global density of all point cloud data, is the local density of the th point cloud data. Among them, the global density of all point cloud data is the mean value of the local densities of all point cloud data, and the clustering radius reference value is used to ensure that the initial clustering radius will not be too large or too small. Here, the clustering radius reference value is set to 0.01.

[0041] Among them, if the If the local density of a point cloud data is high, it means there are many points around it. Therefore, the initial clustering radius should be reduced to ensure that the high-density area is over-segmented into multiple small clusters. Conversely, if the local density of a point cloud data is low, it means there are few points around it. Therefore, the initial clustering radius should be increased to ensure that the clusters in the low-density area can be detected.

[0042] In another embodiment, the initial clustering radius also satisfies the relational expression:

[0043] In the formula, is the initial clustering radius of the th point cloud data, is the clustering radius reference value, is the global density of all point cloud data, is the local density of the th point cloud data, is the th point cloud data's weight (higher weights are assigned to the data that needs special attention during the clustering process according to the actual situation). Among them, the global density of all point cloud data is the mean value of the local densities of all point cloud data, and the clustering radius reference value is used to ensure that the initial clustering radius is neither too large nor too small. Here, the clustering radius reference value is set to 0.01.

[0044] S3: Select any one point cloud data as the target point cloud, obtain all the neighboring point clouds within the range of its initial clustering radius, calculate the similarity values between the target point cloud and each of the neighboring point clouds, and use the similarity values to adjust the initial clustering radius of the target point cloud to obtain the optimal clustering radius of the target point cloud, and further obtain the optimal clustering radii of all point cloud data.

[0045] When the distance between two point cloud data is slightly greater than the initial clustering radius, but in fact they are very similar in other features. If the initial clustering radius is not adjusted, these two point cloud data may be assigned to different clusters. Therefore, based on the initial clustering radius, through continuous adjustment and optimization, the accuracy and adaptability of clustering need to be improved.

[0046] In one embodiment, the radial velocity of the point cloud data is measured by a lidar, that is, the velocity component of the point cloud along the radar beam direction. The radial velocity of each point cloud data needs to be decomposed into three-dimensional coordinates, and further multiply the radial velocity by the unit vector of the three-dimensional coordinates of the corresponding point cloud data to obtain the three-dimensional velocity sequence of each point cloud data ( , , ).

[0047] Then, select any point cloud data as the target point cloud. For all neighboring points within the neighborhood centered on the target point cloud with a radius equal to the initial clustering radius of the target point cloud, calculate the cosine similarity value between the three-dimensional velocity sequence of the target point cloud and the three-dimensional velocity sequences of each neighboring point, which serves as the velocity consistency between the target point cloud and each neighboring point, and calculate the Euclidean distance between the target point cloud and each neighboring point.

[0048] Further, comprehensively evaluate whether the target point cloud and each neighboring point belong to the same object by combining the cosine similarity and the Euclidean distance. That is, both the velocity consistency and the spatial distance are considered, and then calculate the similarity value between the target point cloud and the neighboring points. The satisfaction relationship is as follows:

[0049] In the formula, is the similarity value between the th target point cloud and the th neighboring point cloud, is the th target point cloud and the th neighboring point cloud, is the th target point cloud and the th neighboring point cloud, is the exponential function with the natural constant e as the base.

[0050] It should be noted that the calculation methods of the cosine similarity and the Euclidean distance are prior arts and will not be elaborated here.

[0051] Furthermore, adjust the initial clustering radius of the target point cloud according to the similarity value between the target point cloud and the neighboring points to obtain the optimal clustering radius. The satisfaction relationship is as follows:

[0052] In the formula, is the th optimal clustering radius of the target point cloud, is the th initial clustering radius of the target point cloud, is the th target point cloud and the th neighboring point cloud, is the total number of neighboring points of the target point cloud, is the exponential function with the natural constant e as the base.

[0053] When the average similarity between the target point cloud and all neighboring point clouds in its neighborhood is negative, it indicates that there are significant differences between the target point cloud and the neighboring point clouds. If the initial clustering radius is not adjusted, these point cloud data with large differences may be wrongly clustered together, resulting in missegmentation. By reducing the initial clustering radius, the clustering range can be reduced, thereby avoiding merging dissimilar point clouds into the same cluster and improving the clustering accuracy. When the average similarity between the target point cloud and all neighboring point clouds in its neighborhood is positive, it indicates that the similarity between the target point cloud and the neighboring point clouds is high, and these points are likely to belong to the same cluster. By increasing the initial clustering radius, these similar points can be completely clustered together, avoiding the situation where the cluster is split into multiple small clusters due to too small a radius, thereby enhancing the integrity of the clustering.

[0054] According to the above operation of dynamically adjusting the initial distance class radius, the clustering algorithm can better adapt to the local structure changes of the data, and can find more reasonable clustering boundaries in both high-density regions and low-density regions.

[0055] Further, according to the above calculation steps of the optimal clustering radius of the target point cloud, the optimal clustering radii corresponding to all other point clouds can be obtained in the same way.

[0056] In another embodiment, the similarity value between the th target point cloud and the th neighboring point cloud also satisfies the relational expression:

[0057] In the formula, is the similarity value between the th target point cloud and the th neighboring point cloud, is the velocity consistency between the th target point cloud and the th neighboring point cloud, is the Euclidean distance between the th target point cloud and the th neighboring point cloud, is the exponential function with the natural constant e as the base, is the th target point cloud and the th neighboring point cloud, and

[0058] The above shape similarity is obtained by acquiring and analyzing the shape histograms of the th target point cloud and the th neighboring point cloud.

[0059] S4: Using the optimal clustering radius, the point cloud data of a single frame is clustered and segmented into different clusters, each cluster representing a target. Calculate the distance between the target and the chute, and compare this distance with a preset threshold. If the distance is less than the threshold, an anti-collision warning is triggered.

[0060] In one embodiment, it should be noted that the benchmark value of the clustering radius set in the above S2 is a fixed starting point. It may be able to better adapt to the point cloud data in some cases, but in more cases, it may not fully reflect the complexity and dynamic changes of the point cloud data. Even after the local adjustment in the above S3, this adjustment based on a fixed benchmark value may still have deviations.

[0061] For example, the point cloud data of all frames has similar characteristics on the whole. However, there may be significant differences in the point cloud data between different frames, such as density distribution, noise level, target shape, etc. Therefore, relying solely on the initial benchmark value may not globally optimize to adapt to the characteristics of all frames.

[0062] By dynamically adjusting the benchmark value of the clustering radius, the results after local adjustment can be globally corrected. This global correction can ensure the consistency and comparability of the clustering results between different frames. In other words, local adjustment focuses on the "characteristics within the point cloud", while global adjustment focuses on the "relationships between point clouds".

[0063] Specifically, using the optimal clustering radius, the point cloud data of a single frame is segmented by the DBSCAN clustering algorithm to obtain multiple clustering clusters, each clustering cluster representing a target (such as an obstacle like a cabin wall). Here, the minimum number of points is set to 5, which means that a clustering cluster needs to contain at least 5 points.

[0064] The ICP (Iterative Closest Point) algorithm is used to align (register) the point cloud data of two consecutive frames so as to compare their clustering results.

[0065] For each clustering cluster in the first frame, calculate the degree of aggregation of its corresponding point cloud data in the second frame, that is, calculate the information entropy of the clustering labels of the point cloud data corresponding to the point cloud data within each clustering cluster in the first frame in the second frame. The higher the information entropy, the more dispersed the distribution of these points in the point cloud data of the second frame, that is, the worse the consistency of the clustering results between the two frames; on the contrary, the lower the information entropy, the more consistent the clustering results between the two frames.

[0066] Furthermore, a target function is established, input the sum of all information entropies, and solve the minimum value of this target function through an optimization algorithm to obtain the optimal benchmark value of the clustering radius.

[0067] Further, apply the optimal clustering radius reference value to the relevant formulas in S2 and S3 above, and finally obtain the adjusted optimal clustering radius.

[0068] Among them, the expression of the above objective function is:

[0069] In the formula, is the objective function, is the information entropy of the cluster label corresponding to the point cloud data in the th cluster in the first frame in the second frame, is the number of clusters in the first frame, is the optimal clustering radius reference value.

[0070] After obtaining the adjusted optimal clustering radius, use the adjusted clustering radius to cluster and segment the point cloud data of each frame, identify the target objects (ships, cabins, goods, etc.), further calculate the Euclidean distance between the target and the chute, and compare this Euclidean distance with a pre-set safety distance threshold. If this Euclidean distance is less than the safety distance threshold, then trigger an anti-collision warning.

[0071] The system includes a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the ship loader intelligent anti-collision warning method according to the first aspect of the present invention is implemented.

[0072] The system also includes other components well-known to those skilled in the art such as a communication bus and a communication interface, and their settings and functions are known in the art, so they will not be described in detail here.

[0073] It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of this invention patent shall be subject to the appended claims.

Claims

1. An intelligent anti-collision warning method for a ship unloader, characterized in that Including: Obtain consecutive multi-frame point cloud data of the ship loader and a certain area around it, where the certain area around it includes a chute, a cabin, and goods; Calculate the local density of each point cloud data, and set an initial clustering radius according to the local density; Select any point cloud data as the target point cloud, obtain all neighboring point clouds within the initial clustering radius of the target point cloud, calculate the similarity values between the target point cloud and each of the neighboring point clouds, and use the similarity values to adjust the initial clustering radius of the target point cloud to obtain the optimal clustering radius of the target point cloud, and further obtain the optimal clustering radii of all point cloud data; the similarity value is positively correlated with the velocity consistency between the target point cloud and its neighboring point clouds and negatively correlated with the distance between the target point cloud and the neighboring point clouds; Use the optimal clustering radius to cluster and segment the single-frame point cloud data into different clusters, each cluster representing a target, calculate the distance between the target and the chute, and compare this distance with a preset threshold. If the distance is less than the threshold, trigger an anti-collision warning.

2. The intelligent anti-collision warning method for a ship unloader according to claim 1, wherein The point cloud data includes the three-dimensional coordinates and reflection intensity of the point cloud data.

3. The intelligent anti-collision warning method for a ship unloader according to claim 2, wherein The process of obtaining the local density includes: Normalize the three-dimensional coordinates and reflection intensity of the point cloud data, and use kernel density estimation to calculate the local density of each point cloud data; Among them, a Gaussian kernel function is selected, and the bandwidth of the kernel density estimation is calculated.

4. The intelligent anti-collision warning method for a ship unloader according to claim 3, wherein, The initial clustering radius satisfies the relationship: ; wherein, is the initial clustering radius of the -th point cloud data, is the clustering radius reference value, is the global density of all point cloud data, is the local density of the -th point cloud data; wherein, the global density of all point cloud data is the mean value of the local densities of all point cloud data.

5. The intelligent anti-collision warning method for a ship unloader according to claim 4, characterized in that, The similarity satisfies the relationship: ; Wherein, is the similarity value between the th target point cloud and the th neighborhood point cloud, is the velocity consistency between the th target point cloud and the th neighborhood point cloud, is the Euclidean distance between the th target point cloud and the th neighborhood point cloud, is the exponential function with the natural constant e as the base; wherein, is obtained by calculating the cosine similarity between the velocity vectors of the th target point cloud and the th neighborhood point cloud.

6. The intelligent anti-collision warning method for a ship unloader according to claim 5, characterized in that, The optimal clustering radius satisfies the relationship: ; where, is the optimal clustering radius of the th target point cloud, is the initial clustering radius of the th target point cloud, is the similarity value between the th target point cloud and the th neighboring point cloud, is the total number of neighboring points of the target point cloud, is the exponential function with the natural constant e as the base.

7. The intelligent anti-collision warning method for a ship unloader according to claim 6, characterized in that, The obtaining of the optimal clustering radii of all point cloud data further includes: Use the optimal clustering radius to segment the single-frame point cloud data to obtain multiple clustering clusters; select any two consecutive frames of point cloud data and align the two consecutive frames of point cloud data; For the point cloud data within each clustering cluster in the first frame, calculate the information entropy of the clustering labels of the corresponding point cloud data in the second frame; Establish an objective function, input the sum of all the information entropies, and output the optimal clustering radius reference value; obtain the adjusted optimal clustering radius based on the optimal clustering radius reference value.

8. The intelligent anti-collision warning method for a ship unloader according to claim 7, wherein The expression of the objective function is: ; In the formula, is the objective function, is the information entropy of the cluster label corresponding to the point cloud data in the -th cluster in the first frame in the second frame, is the number of clusters in the first frame, is the optimal clustering radius reference value.

9. The intelligent anti-collision warning method for a ship unloader according to claim 8, characterized in that The clustering is the DBSCAN clustering algorithm.

10. An intelligent anti-collision warning system for a ship unloader, characterized in that, Including: A processor and a memory, where the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the intelligent anti-collision warning method for the ship loader according to any one of claims 1-9 is implemented.

Images