Debris flow detection method and related device
By collecting point cloud sequences with radar and extracting the motion pattern features of plant clusters, the problem of all-weather debris flow detection has been solved, and efficient and accurate debris flow event detection has been achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG DAHUA TECH CO LTD
- Filing Date
- 2026-01-05
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technologies are insufficient for all-weather, all-time debris flow detection, and their accuracy is inadequate.
By using radar to collect point cloud sequences of the area to be detected, the actual movement pattern characteristics of plant clusters are extracted, and based on these characteristics, it is determined whether a debris flow event has occurred.
It enables all-weather, all-time debris flow detection, improving the accuracy and reliability of detection and reducing installation requirements.
Smart Images
Figure CN122017773A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of point cloud data processing technology, and in particular to a debris flow detection method and related apparatus. Background Technology
[0002] Debris flow is a special type of flood that carries large amounts of solid materials such as mud, sand, rocks, and boulders, which is caused by sudden and massive rainfall in valleys or on mountain slopes. It is a hazardous geological phenomenon.
[0003] Currently, there is an urgent need for a debris flow detection method to detect debris flow events in order to respond promptly in the event of a debris flow. Summary of the Invention
[0004] This application provides a debris flow detection method and related apparatus, which can detect whether a debris flow event has occurred.
[0005] This application provides a debris flow detection method, comprising: sequentially acquiring multiple point cloud frames to be detected at the site to be detected using radar to form a point cloud sequence to be detected; extracting actual motion pattern features of plant clusters at the site to be detected from the point cloud sequence to be detected, wherein a plant cluster includes at least one plant; and determining whether a debris flow event has occurred at the site to be detected based on the actual motion pattern features of the plant clusters.
[0006] This application provides a debris flow detection device, comprising: a data acquisition module, an extraction module, and a determination module. The data acquisition module is used to sequentially acquire multiple point cloud frames of the site to be detected using radar, forming a point cloud sequence. The extraction module is used to extract the actual motion pattern features of plant clusters at the site from the point cloud sequence, where each plant cluster includes at least one plant. The determination module is used to determine whether a debris flow event has occurred at the site based on the actual motion pattern features of the plant clusters.
[0007] This application provides an electronic device, including a memory and a processor, wherein the processor is used to execute program instructions stored in the memory to implement the above-described method.
[0008] This application provides a computer-readable storage medium having program instructions stored thereon, which, when executed by a processor, implement the above-described method.
[0009] This application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.
[0010] The above scheme, on the one hand, utilizes radar to collect point cloud sequences of the area to be detected, and then determines in real time whether a debris flow event has occurred at that location. Compared to image acquisition equipment, radar is unaffected by changes in ambient light and weather, thus enabling all-weather, all-time real-time debris flow detection with low installation requirements. On the other hand, it extracts the actual movement pattern features of plant clusters at the area to be detected from the point cloud sequence, and determines whether a debris flow event has occurred based on these features. These actual movement pattern features characterize the actual movement patterns of the plant clusters, thus enabling the determination of whether a debris flow event has occurred based on the actual movement patterns of the plant clusters at the area to be detected.
[0011] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this application. Attached Figure Description
[0012] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the technical solutions of this application.
[0013] Figure 1 This is a flowchart illustrating an embodiment of the debris flow detection method provided in this application; Figure 2 This is a schematic diagram of the point cloud frame to be detected in this application; Figure 3 This is a flowchart illustrating Embodiment 2 of the debris flow detection method provided in this application; Figure 4 This is a flowchart illustrating Embodiment 3 of the debris flow detection method provided in this application; Figure 5 This is a flowchart illustrating Embodiment 4 of the debris flow detection method provided in this application; Figure 6 This is a flowchart illustrating Embodiment 5 of the debris flow detection method provided in this application; Figure 7 This is a schematic diagram of the business process for debris flow detection in this application; Figure 8 This is a flowchart illustrating a specific example of debris flow detection in this application; Figure 9 This is a schematic diagram of the three-dimensional region of the plant cluster in this application; Figure 10 This is a schematic diagram of the structure of an embodiment of the electronic device of this application; Figure 11 This is a schematic diagram of the structure of an embodiment of the computer-readable storage medium of this application. Detailed Implementation
[0014] The embodiments of this application will now be described in detail with reference to the accompanying drawings.
[0015] In the following description, specific details such as particular system architectures, interfaces, and technologies are presented for illustrative purposes rather than for limiting purposes, in order to provide a thorough understanding of this application.
[0016] In this document, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, "many" in this document means two or more. The term "at least one" in this document means any combination of at least two of any one or more of a plurality of objects. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C. Finally, the term "several" in this document means any integer greater than 0, such as 1, 2, 3, 4, 5, ...
[0017] Figure 1 This is a flowchart illustrating an embodiment of the debris flow detection method provided in this application. Figure 1 As shown, in this embodiment, the debris flow detection method may include the following steps: S110: Multiple point cloud frames to be detected are collected sequentially by radar to form a point cloud sequence to be detected.
[0018] The execution subject of this embodiment is a debris flow detection device, which can be any electronic device with debris flow detection capability.
[0019] The detection site is a location where mudslides may occur, such as valleys and hillsides. The radar can acquire point cloud frames to be detected according to a preset measurement period / frequency. For example, the measurement period is 0.1 seconds, and the measurement frequency is 10 Hz. Two adjacent point cloud frames in the detection sequence can be acquired continuously or not. For example, if two adjacent point cloud frames are acquired continuously, the acquisition interval is the measurement period. Alternatively, if two adjacent point cloud frames are not acquired continuously, the acquisition interval is 20 seconds. Generally, a larger acquisition interval results in higher detection accuracy. A smaller acquisition interval results in faster detection speed and higher real-time performance.
[0020] Radar can be, but is not limited to, lidar. The principle of radar acquiring point cloud frames to be detected is as follows: Figure 2 This is a schematic diagram of the cloud frame of the points to be detected in this application. Figure 2The system includes three point cloud frames to be detected: frame 1, frame 2, and frame 3. The lidar simultaneously emits laser scanning lines 1, 2, ..., m at different angles, resulting in three-dimensional point sequences 1, 2, ..., m, which together form the point cloud frame 1 to be detected. The three-dimensional point sequence 1 sequentially includes three-dimensional point 1, 2, 3, ..., n. Three-dimensional point 1 includes three-dimensional coordinates (x, y, z) and laser reflection intensity D. The three-dimensional coordinates (x, y, z) are relative to a coordinate system established with the lidar as the origin, and their calculation formula is as follows: ; Where d represents the distance from the three-dimensional point to the radar. This represents the yaw angle of laser scan line 1 around the Z-axis. The elevation angle represents the laser scanning line 1 being emitted.
[0021] The other 3D points in 3D point sequence 1 are similar to those in 3D point sequence 1. The other 3D point sequences are similar to those in 3D point sequence 1. The other point cloud frames to be detected are similar to those in point cloud frame 1. Further details are omitted here.
[0022] S120: Extract the actual motion pattern features of plant clusters in the detection site from the point cloud sequence to be detected.
[0023] A plant cluster includes at least one plant.
[0024] Plants can be trees, flowers, grass, etc. Multiple plants that are close together or overlap will form a plant cluster in the radar field of view.
[0025] The actual movement pattern characteristics characterize the actual movement pattern of the plant cluster, which may include the actual movement direction changes and the actual movement amplitude of the plant cluster.
[0026] S130: Based on the actual movement pattern characteristics of plant clusters, determine whether a debris flow event has occurred at the site to be detected.
[0027] Through long-term research, the inventors of this application have discovered that mudslides can cause plant clusters to move according to a specific movement pattern.
[0028] The above scheme, on the one hand, utilizes radar to collect point cloud sequences of the area to be detected, and then determines in real time whether a debris flow event has occurred at that location. Compared to image acquisition equipment, radar is unaffected by changes in ambient light and weather, thus enabling all-weather, all-time real-time debris flow detection with low installation requirements. On the other hand, it extracts the actual movement pattern features of plant clusters at the area to be detected from the point cloud sequence, and determines whether a debris flow event has occurred based on these features. These actual movement pattern features characterize the actual movement patterns of the plant clusters, thus enabling the determination of whether a debris flow event has occurred based on the actual movement patterns of the plant clusters at the area to be detected.
[0029] It is understandable that the movement pattern of plant clusters may fall into three modes: Mode 1, Mode 2, and Mode 3. Mode 1 is a state of motion, characterized by vigorous, swaying movement. Vigorous motion is characterized by large amplitude, while swaying movement is characterized by periodic changes in direction, such as repeatedly shifting between two directions. Mode 2 is a state of motion, characterized by slight, directional movement. Slight movement is characterized by small amplitude, while directional movement is characterized by a fixed direction, such as consistently moving in the first or second direction. Mode 3 is a static state, characterized by almost zero amplitude. Generally, Mode 1 is caused by debris flow events, while Mode 2 is caused by wind. Therefore, if it is Mode 1, it means a debris flow event has occurred. If it is Mode 2 or Mode 3, it means no debris flow event has occurred.
[0030] Furthermore, S120 is further expanded as follows: Figure 3 This is a schematic flowchart of Embodiment 2 of the debris flow detection method provided in this application. This embodiment is a further extension of S120. Figure 3 As shown, in this embodiment, S120 may include the following steps: S121: Obtain the sub-motion pattern features of the plant cluster between each pair of adjacent point cloud frames to be detected.
[0031] For example, the point cloud sequence to be detected is {point cloud frame 1, point cloud frame 2, ..., point cloud frame n}. Point cloud frame i and point cloud frame i-1 are two adjacent point cloud frames to be detected. The sub-motion pattern features between point cloud frame i and point cloud frame i-1 are obtained, where i∈1,2,...,n.
[0032] Sub-motion pattern features can characterize the actual direction and amplitude of movement of plant clusters between two adjacent cloud frames to be detected.
[0033] S122: Statistically analyze the characteristics of each sub-movement mode of the plant cluster to obtain the actual movement mode characteristics of the plant cluster.
[0034] Statistical methods can include sequential combination, multiplication, addition, dispersion statistics, and central tendency statistics.
[0035] In some embodiments, the actual movement pattern characteristics include movement pattern central tendency characteristics and movement pattern dispersion characteristics. S122 includes: performing central tendency statistics and dispersion statistics on the characteristics of each sub-movement pattern of the plant cluster respectively, to obtain the movement pattern central tendency characteristics and movement pattern dispersion characteristics.
[0036] Among them, the central tendency feature of the motion pattern is the statistical value of the central tendency of each sub-motion pattern feature, which can be the mean, mode, median, etc. The dispersion feature of the motion pattern is the statistical value of the dispersion of each sub-motion pattern feature, which can be the variance, standard deviation, range, etc.
[0037] Figure 4 This is a flowchart illustrating Embodiment 3 of the debris flow detection method provided in this application. This embodiment is a further extension of S121, and the sub-motion mode features include sub-motion amplitude features and sub-motion direction features. For example... Figure 4 As shown, in this embodiment, S121 may include the following steps: S1211: Obtain the three-dimensional region of the plant cluster in each cloud frame of the point to be detected.
[0038] The three-dimensional region of the plant cluster represents the outer contour of the three-dimensional points belonging to the plant cluster in the cloud frame of the point to be detected.
[0039] In some embodiments, S1211 includes: for each point cloud frame to be detected, acquiring three-dimensional points belonging to plant clusters in the point cloud frame to be detected; and extracting the outer contours of the three-dimensional points belonging to plant clusters in the point cloud frame to be detected.
[0040] The 3D points belonging to plant clusters can be obtained using plant cluster detection algorithms, which can be implemented using, but are not limited to, plant cluster detection networks such as PointNet, PointNet++, and PointGNN. The outer contour can be extracted using, but is not limited to, convex hull detection.
[0041] S1212: For each pair of adjacent point cloud frames to be detected, obtain the degree of overlap of the three-dimensional regions of the plant clusters in the two adjacent point cloud frames to be detected as the sub-motion amplitude feature; and obtain the sub-motion direction feature of the three-dimensional regions of the plant clusters in the two adjacent point cloud frames to be detected.
[0042] When the previous distance is greater than the next distance, the sub-motion direction feature is the first motion direction feature; when the previous distance is less than the next distance, the sub-motion direction feature is the second motion direction feature. The first motion direction feature and the second motion direction feature represent opposite motion directions. The previous distance is the distance from the three-dimensional region of the plant cluster in the previous point cloud frame to the radar, and the next distance is the distance from the three-dimensional region of the plant cluster in the next point cloud frame to the radar.
[0043] In this model, the frame acquired earlier among two adjacent point cloud frames to be detected is designated as the preceding point cloud frame, and the frame acquired later is designated as the following point cloud frame. The degree of overlap of the three-dimensional regions of the plant cluster can be obtained, but is not limited to, through intersection-union ratio (IU). The sub-motion amplitude feature / overlap degree characterizes the motion amplitude of the plant cluster between two adjacent point cloud frames to be detected. The higher the overlap degree, the greater the motion amplitude.
[0044] Understandably, the relationship between the preceding and following distances reflects the direction of movement of the plant cluster between the previous and subsequent detection point cloud frames. If the preceding distance is greater than the following distance, it means the plant cluster is in motion, moving from a position far from the radar to a position closer to the radar, i.e., moving in the direction closer to the radar, denoted as the first direction of motion. If the preceding distance is equal to the following distance, it means the plant cluster is stationary. If the preceding distance is less than the following distance, it means the plant cluster is in motion, moving from a position close to the radar to a position far from the radar, i.e., moving in the direction far away from the radar, denoted as the second direction of motion.
[0045] The characteristics of each sub-movement direction can collectively characterize the actual movement direction change of the plant cluster, and the characteristics of each sub-movement amplitude can collectively characterize the actual movement amplitude of the plant cluster.
[0046] In some embodiments, the sub-motion direction feature is a symbolic variable, where the first motion direction feature is a first symbolic variable and the second motion direction feature is a second symbolic variable. In this case, the sub-motion pattern feature can be the result of multiplying the sub-motion amplitude feature and the corresponding sub-motion direction feature. For example, if the sub-motion direction feature is 'sign', the sub-motion pattern feature = 'sign' * sub-motion amplitude feature. If the sub-motion direction feature is the first motion direction feature, sign = 1; if the sub-motion direction feature is the second motion direction feature, sign = -1.
[0047] Furthermore, S130 is further expanded as follows: In some embodiments, S130 includes: acquiring the expected movement pattern characteristics of the plant cluster; determining the degree of feature difference between the actual movement pattern characteristics and the expected movement pattern characteristics, and determining whether a debris flow event has occurred based on the degree of feature difference.
[0048] In some embodiments, the desired motion pattern characteristics are motion pattern characteristics under conditions of a debris flow event. Based on this, if the degree of feature difference is less than a first feature difference threshold, a debris flow event is determined to have occurred; if it is not less than the first feature difference threshold, a debris flow event is determined not to have occurred.
[0049] In some embodiments, the desired motion pattern characteristics are those characteristic of motion patterns in the absence of a debris flow event. Based on this, if the degree of feature difference is less than a second feature difference threshold, it is determined that no debris flow event has occurred; if it is not less than the second feature difference threshold, it is determined that a debris flow event has occurred.
[0050] It is further understood that short-term directional winds or strong winds may also cause plant clusters to move according to the first motion pattern. Therefore, judging whether a debris flow event has occurred solely based on the actual motion pattern characteristics may lead to misjudgments, resulting in inaccurate event determination results. Therefore, to avoid misjudgments and improve the accuracy of event determination results, some embodiments, in S130, can combine the actual wind speed and actual motion pattern characteristics of the plant clusters during the acquisition time period of the point cloud sequence to be detected to determine whether a debris flow event has occurred. Specifically: Figure 5 This is a schematic flowchart of Embodiment 4 of the debris flow detection method provided in this application. Figure 5 As shown, in this embodiment, S130 may include the following steps: S131: Obtain the actual wind speed of the plant clusters during the collection time period of the point cloud sequence to be detected.
[0051] The actual wind speed of the plant cluster can be the wind speed at the acquisition time of any point cloud frame in the point cloud sequence to be detected, or it can be the statistical result of the wind speed at the acquisition time of each point cloud frame to be detected. For example, the average wind speed at each acquisition time can be used as the actual wind speed.
[0052] In some embodiments, a single anemometer is installed at the location to be tested, and the wind speed at the time of acquisition of the cloud frame of the point to be tested can be measured by the single anemometer.
[0053] In some embodiments, multiple anemometers are installed at the site to be tested, with different anemometers installed at different locations within the site to measure the wind speed at each location. This allows for the acquisition of multiple actual wind speeds for the plant cluster based on these multiple anemometers. The final actual wind speed for the plant cluster is then determined by the anemometer whose installation location is closest to the cluster, thereby improving the accuracy of the actual wind speed measurement.
[0054] S132: Obtain the actual correlation between the actual wind speed and the actual motion pattern characteristics of the plant cluster, and obtain the reference correlation between the reference wind speed and the reference motion pattern characteristics of the plant cluster.
[0055] In some embodiments, the reference wind speed and reference motion pattern features are obtained based on a reference point cloud sequence in which a debris flow event occurred.
[0056] In some embodiments, the reference wind speed and reference motion pattern features are obtained based on reference point cloud sequences where no debris flow event has occurred. It is understood that since debris flow events are relatively rare, reference wind speed and reference motion pattern features for situations involving debris flow events are scarce and difficult to simulate, while obtaining reference wind speed and reference motion pattern features for situations where no debris flow event has occurred is relatively easy. In some embodiments, the historical actual wind speed and historical actual motion pattern features corresponding to historical point cloud sequences where the event determination result is that no debris flow event has occurred can be used as the reference wind speed and reference motion pattern features, respectively.
[0057] In some embodiments, there are one or more sets of reference wind speed and reference motion pattern features. When there are multiple sets, the final reference correlation can be obtained by fitting multiple sets of reference wind speed and reference motion pattern features. The methods for obtaining the reference wind speed and actual wind speed are similar, as are the methods for obtaining the reference motion pattern features and actual motion pattern features.
[0058] In some embodiments, a set of reference wind speeds and reference motion pattern features is referred to as a set of reference data. Multiple sets of reference data belonging to the same reference wind speed constitute a subset of reference data, and multiple subsets of reference data for different reference wind speeds constitute a reference dataset. The number of reference data sets in the subsets of reference data for different reference wind speeds in the reference dataset may be different. Reference data sets with the same number of sets can be selected from the subsets of reference data for different reference wind speeds to fit the reference correlation and improve the accuracy of the reference correlation.
[0059] S133: Determine whether a debris flow event has occurred at the site to be monitored based on the degree of difference between the actual correlation and the reference correlation.
[0060] In some embodiments, the reference correlation is obtained when no debris flow event has occurred. If the degree of difference is less than the first correlation difference threshold, it is determined that no debris flow event has occurred; if it is not less than the first correlation difference threshold, it is determined that a debris flow event has occurred.
[0061] In some embodiments, the reference correlation is obtained in the event of a debris flow event. If the difference is less than the second correlation difference threshold, it is determined that a debris flow event has occurred; if it is not less than the second correlation difference threshold, it is determined that no debris flow event has occurred.
[0062] In some embodiments, the F-test can be used to measure the degree of difference between the actual correlation and the reference correlation, thereby determining whether a debris flow event has occurred. Specifically, the null hypothesis of the F-test is that the actual correlation and the reference correlation are the same, and the F-estimate is calculated; using the rejection threshold of the null hypothesis, the numerator degrees of freedom and the denominator degrees of freedom of the F-estimate, the theoretical F-value is looked up in the F-distribution table; if the F-estimate is greater than the theoretical F-value, it is determined that a debris flow event has occurred at the site to be tested; if the F-estimate is not greater than the theoretical F-value, it is determined that no debris flow event has occurred at the site to be tested.
[0063] In some embodiments, prior to S130, the method further includes: acquiring the radar attitude angle during the acquisition time period of the point cloud sequence to be detected; and combining the radar attitude angle with the actual motion pattern features of the plant cluster to update the actual motion pattern features of the plant cluster.
[0064] The radar's attitude angle can be obtained, but is not limited to, through gyroscope measurement. It is understandable that a debris flow event would cause a change in the attitude angle of the radar installed at the detection site. Therefore, incorporating the attitude angle as part of the actual motion pattern characteristics can improve the accuracy of determining whether a debris flow event has occurred based on these characteristics.
[0065] Furthermore, it can be understood that in the event of a debris flow event, the debris flows along the surface of the area to be detected, causing significant changes to the surface. Therefore, it is possible to determine whether a debris flow event has occurred based on the three-dimensional region of the surface in the point cloud frame to be detected. However, because the radar's field of view does not cover the surface of the area to be detected, or the surface of the area to be detected is obscured by vegetation, it may be impossible to extract the surface from at least one point cloud frame to be detected, or the extracted three-dimensional region of the surface may be incomplete, making it impossible to determine whether a debris flow event has occurred based on the three-dimensional region of the surface.
[0066] Therefore, the occurrence of a debris flow event can be determined by combining the three-dimensional area of the land surface and the movement patterns of plant clusters. Specifically: Figure 6 This is a flowchart illustrating Embodiment 5 of the debris flow detection method provided in this application. This embodiment is a further extension of the foregoing embodiments, such as... Figure 6 As shown, in this embodiment, the debris flow detection method may include the following steps: S210: The cloud frame of the current point to be detected has been acquired.
[0067] S220: Determine whether the current point cloud frame to be detected meets the surface extraction conditions.
[0068] The conditions for surface extraction include at least one of the following: the location to be detected includes the surface; the current three-dimensional region of the surface can be extracted from the cloud frame of the current point to be detected; and the area of the current three-dimensional region of the surface is greater than an area threshold.
[0069] Among them, an area greater than the area threshold means that the extracted current 3D region is sufficiently complete.
[0070] In response to the failure to meet the surface extraction conditions, a point cloud sequence to be detected is formed using the current point cloud frame to be detected and its adjacent preset number of historical point cloud frames to be detected, and step S230 is executed. In response to the fulfillment of the surface extraction conditions, step S240 is executed. The adjacent preset number of historical point cloud frames to be detected and the current point cloud frame to be detected are sequentially acquired using radar.
[0071] S230: Execute S120-S130.
[0072] S240: Determine whether a debris flow event has occurred based on the current three-dimensional area of the surface.
[0073] In some embodiments, S240 includes: determining that a significant change has occurred in the surface and determining that a debris flow event has occurred based on the degree of surface difference between the current three-dimensional region of the surface and the historical three-dimensional region of the surface in the historical point cloud frame to be detected; otherwise, determining that no debris flow event has occurred.
[0074] Among them, the acquisition time of the historical point cloud frame to be detected is earlier than the acquisition time of the current point cloud frame to be detected, and the event determination result corresponding to the historical point cloud frame to be detected is that no debris flow event has occurred.
[0075] In some embodiments, S240 further includes: acquiring the current attitude angle of the radar at the acquisition time of the current point cloud frame to be detected; acquiring the attitude angle difference degree between the current attitude angle and the historical attitude angle; in response to the attitude angle difference degree being greater than the attitude angle difference degree threshold and the surface difference degree being greater than the surface difference degree threshold, determining whether the surface has changed significantly and determining that a debris flow event has occurred; otherwise, determining that no debris flow event has occurred.
[0076] The above scheme prioritizes obtaining event determination results from the surface, given that surface extraction conditions are met. If these conditions are not met, the actual movement patterns of plant clusters are then used to determine whether a debris flow event has occurred. This approach improves the accuracy of event determination results while ensuring the normal acquisition of such results.
[0077] Furthermore, in some embodiments, the site to be detected includes multiple plant clusters. Based on the actual movement pattern characteristics of the plant clusters, it is determined whether a debris flow event has occurred at the site to be detected. This includes: for each plant cluster, determining whether a debris flow event has occurred based on the actual movement pattern characteristics of the plant cluster, and obtaining the event determination result corresponding to the plant cluster; if there is a preset proportion of plant clusters whose event determination results indicate that a debris flow event has occurred, it is finally determined that a debris flow event has occurred at the site to be detected.
[0078] The preset ratio can be set according to actual needs. The preset ratio is negatively correlated with the detection sensitivity.
[0079] Furthermore, in some embodiments, after S110, the method further includes: determining whether the acquisition time period of the point cloud sequence to be detected belongs to a first type of time period or a second type of time period; in response to belonging to the first type of time period, determining that no debris flow event has occurred at the location to be detected; in response to belonging to the second type of time period, executing S120-S130. The first type of time period can be spring, autumn, winter, etc., and the second type of time period can be summer, etc.
[0080] In some embodiments, S110 includes: during a second time period, using radar to sequentially acquire multiple point cloud frames to be detected at the location to be detected, so as to form a point cloud sequence to be detected.
[0081] Understandably, debris flow events generally occur in the second type of time period, so conducting debris flow detection only in the second type of time period can save the resources required for debris flow detection.
[0082] In some embodiments, after S130, the method further includes: in response to a debris flow event, notifying the user of the debris flow event. The notification to the user may be made via a network or other means.
[0083] In some embodiments, after notifying users of a debris flow event, the event can be manually verified to confirm its authenticity, and a response can be initiated if it is confirmed. Users can verify the event's authenticity through methods such as video recordings or on-site investigations.
[0084] In some embodiments, the point cloud sequence to be detected, the results of debris flow event determination, and the results of manual verification can also be uploaded to the cloud for sharing by relevant equipment or for subsequent traceability.
[0085] In some embodiments, after S130, the method further includes: in response to a debris flow event, labeling the point cloud sequence to be detected as a reference point cloud sequence in which a debris flow event has occurred; and in response to no debris flow event, labeling the point cloud sequence to be detected as a reference point cloud sequence in which no debris flow event has occurred, to enrich the reference point cloud sequence for subsequent acquisition of reference associations. It is understood that if the event determination result obtained from debris flow detection of the point cloud sequence to be detected is that a debris flow event has occurred, but the manual review result is that no debris flow event has occurred, it means that the point cloud sequence to be detected has caused a misjudgment. Labeling it as a reference point cloud sequence in which no debris flow event has occurred, and having it participate in the subsequent acquisition of reference associations, can reduce the probability of subsequent misjudgments and improve the accuracy of the event determination result.
[0086] To facilitate understanding, the debris flow detection method provided in this application is illustrated below with a specific example: Figure 7 This is a schematic diagram of the business process for debris flow detection in this application. For example... Figure 7 As shown, the business process includes: A1. Radar Installation: Radar installation is simple. Simply mount the radar on a support frame so that its field of view covers the area to be detected.
[0087] A2. Parameter adjustment: Set the working time (e.g., based on Beijing time), detection sensitivity, radar measurement cycle, etc.
[0088] A3. Data Acquisition: Acquire the cloud frame of the current point to be detected.
[0089] A4. Data Acquisition Time Period Classification: Determine whether the current data acquisition time period belongs to Category I or Category II. If it is Category I, confirm that no debris flow event has occurred and return to A3 to continue data acquisition; if it is Category II, proceed to A5.
[0090] A5. Debris Flow Detection: Performs debris flow detection on the current point cloud frame or point cloud sequence to determine whether a debris flow event has occurred. The point cloud sequence to be detected consists of the current point cloud frame and a preset number of adjacent historical point cloud frames. If a debris flow event occurs, proceed to A6-A7; if no debris flow event occurs, return to A3 to continue data acquisition.
[0091] A6. Notify users: Notify users of the mudslide event via the internet.
[0092] A7. Manual Verification: Manual verification determines whether a debris flow event has occurred. If the debris flow event is confirmed as true, the user can respond. If the debris flow event is false, no response is required.
[0093] A8. Upload to the cloud: Upload the point cloud sequence to be detected, the event determination results of debris flow, and the results of manual verification to the cloud for sharing by other devices.
[0094] A9. If a debris flow event is confirmed, the point cloud sequence to be detected is labeled as a reference point cloud sequence indicating that a debris flow event has occurred. If no debris flow event is confirmed, the point cloud sequence to be detected is labeled as a reference point cloud sequence indicating that no debris flow event has occurred.
[0095] Theoretical basis for debris flow detection: Basis 1: When a debris flow event occurs, the debris flow moves along the surface of the site to be tested. Therefore, the surface changes significantly before and after a debris flow event.
[0096] Basis 2: The radar attitude angle changed before and after the mudslide event.
[0097] Basis 3: During a debris flow event, the movement pattern of the plant clusters is the first movement pattern.
[0098] Based on the above theoretical basis, debris flow detection is performed on the point cloud sequence to be detected. Figure 8 This is a flowchart illustrating a specific example of debris flow detection in this application. For example... Figure 8 As shown, debris flow detection includes: B1. Initialization. Initialization includes: loading the debris flow detection model, accumulating the number of point cloud frames to be detected in the point cloud sequence, and setting the rejection threshold.
[0099] B2. Use radar to collect cloud frames of the points to be detected at the current time.
[0100] B3. Obtain the current attitude angle (P, T, Z) of the radar at the current moment.
[0101] B4. Determine whether the current cloud frame of the point to be detected meets the surface extraction conditions.
[0102] If the surface extraction conditions are met, proceed to B10; if the surface extraction conditions are not met, proceed to B5-B8.
[0103] B5. For the current point cloud frame to be detected and its adjacent preset number of historical point cloud frames to be detected, form a point cloud sequence to be detected according to the order of acquisition. For example, there are k point cloud frames to be detected in the point cloud sequence, and the acquisition time interval between two adjacent point cloud frames to be detected is T.
[0104] B6. Extract the actual motion pattern features of plant clusters from the point cloud sequence to be detected.
[0105] B61. Plant cluster detection algorithms are used to detect plant clusters in each point cloud frame to obtain the 3D points belonging to plant clusters in each frame. If the detected 3D points belonging to plant clusters are sparse in the point cloud frame, interpolation or smoothing can be performed on these points to construct more densely packed 3D points belonging to plant clusters.
[0106] B62. Using the convex hull detection algorithm, the outer contours of the three-dimensional points belonging to the plant clusters in each detection point cloud frame are extracted, and the three-dimensional regions of the plant clusters in each detection point cloud frame are obtained. Figure 9 This is a schematic diagram of the three-dimensional region of the plant cluster in this application. Because three-dimensional regions are not easily represented visually, therefore... Figure 9 The text uses two-dimensional space to represent a three-dimensional region. For example... Figure 9 As shown, the set of three-dimensional points belonging to the plant cluster in the point cloud frame to be detected constitutes the convex hull / outer contour of the plant cluster, i.e., the three-dimensional region.
[0107] B63. For each pair of adjacent point cloud frames to be detected, obtain the cross-union ratio (CUB) of the 3D regions of the plant clusters in the two adjacent point cloud frames, as a sub-motion amplitude feature. The formula for calculating the CUB of the 3D regions of the plant clusters in two adjacent point cloud frames is as follows:
[0108] in, , These represent the three-dimensional regions of the plant clusters in the point cloud frame i to be detected and the three-dimensional regions of the plant clusters in the point cloud frame i+T to be detected, respectively. This indicates intersection, union, and ratio.
[0109] B64. For each pair of adjacent point cloud frames to be detected, obtain the sub-motion direction features of the three-dimensional region of the plant cluster in the two adjacent point cloud frames to be detected.
[0110] In two adjacent point cloud frames to be detected, the one acquired earlier is designated as the previous point cloud frame, and the one acquired later is designated as the next point cloud frame to be detected. Let the distance from the 3D region of the plant cluster in the previous point cloud frame to the radar be the previous distance, and let the distance from the 3D region of the plant cluster in the next point cloud frame to the radar be the next distance.
[0111] When the preceding distance is greater than the following distance, the sub-motion direction feature is determined as the first symbolic variable. The first symbolic variable represents the first motion direction. When the preceding distance is less than the following distance, the sub-motion direction feature is determined as the second symbolic variable. The second symbolic variable represents the second motion direction. The first motion direction is the direction closer to the radar, and the second motion direction is the direction farther away from the radar.
[0112] For example, The distance to the radar is greater than In the case of the distance to the radar, the sub-motion direction feature sign=1, which is less than... When the distance to the radar is less than a certain value, the sub-motion direction characteristic is sigh=-1.
[0113] B64. Multiply the sub-motion direction features and sub-motion amplitude features of the three-dimensional region of the plant cluster in two adjacent point cloud frames to be detected to obtain the sub-motion pattern features.
[0114] For example, combining the sub-motion direction feature (sign) with the sub-motion amplitude feature Multiply to update the sub-motion amplitude features =sign* The updated sub-motion amplitude features are used as sub-motion pattern features.
[0115] B65. Statistical analysis of the movement pattern characteristics of each sub-movement pattern of the plant cluster yields the actual movement pattern characteristics of the plant cluster.
[0116]
[0117] in, , These represent the mean and variance of the features of the n sub-motion patterns, respectively.
[0118] B66. Combine the current attitude angles (P, T, Z) with the actual motion pattern features of the plant cluster to update the actual motion pattern features of the plant cluster.
[0119]
[0120] in, This represents the actual movement pattern characteristics of the updated plant clusters.
[0121] B7. Measure the actual wind speed of the plant cluster at the current moment and obtain the actual correlation between the actual wind speed and the actual motion pattern characteristics of the plant cluster. Similarly, obtain the reference correlation between the reference wind speed and the reference motion pattern characteristics of the plant cluster.
[0122] The relationship between wind speed and motion pattern characteristics can be represented by the following regression relationship:
[0123] in, Indicates a relationship. This represents the 6-dimensional parameters in the association relationship.
[0124]
[0125] in, Indicates the actual wind speed. Indicates the characteristics of the actual motion pattern. Indicates the actual relationship.
[0126]
[0127] in, This represents a vector composed of multiple reference wind speeds. This represents a vector composed of features from multiple corresponding reference motion patterns. This indicates a reference relationship.
[0128] B8. The actual correlation and the reference correlation are compared by using the F-test to determine whether a debris flow event has occurred.
[0129] Due to factors such as measurement errors and wind speed differences, the following may occur:
[0130] According to the significance test theory of regression relationship, if they are the same distribution or the same correlation, it means that (actual wind speed, actual motion pattern characteristics) and multiple sets (reference wind speed, reference motion pattern characteristics) correspond to the same event determination result. Since the event determination result corresponding to multiple sets (reference wind speed, reference motion pattern characteristics) is that no debris flow event has occurred, it means that the event determination result corresponding to (actual wind speed, actual motion pattern characteristics) is that no debris flow event has occurred.
[0131] Based on linear estimation theory, the following linear regression model can be constructed:
[0132] in, , These represent the actual measurement error and the reference measurement error, respectively. They are known and follow a Gaussian distribution.
[0133] Can make
[0134] B81: The null hypothesis for the F-test is that the actual association is the same as the reference association.
[0135] : ; Here, represents the null hypothesis.
[0136] B82: Calculate the F estimate.
[0137] If the null hypothesis holds, the following constraints must be satisfied:
[0138] According to the theory of least squares estimation with constraints, we can obtain:
[0139] in, , These represent the actual correlation and the reference correlation obtained through estimation, respectively. , These represent the actual sum of squares of residuals and the reference sum of squares of residuals obtained through estimation, respectively.
[0140] The calculated F-estimate is:
[0141] in, The number of groups representing (actual wind speed, actual motion pattern characteristics), The number of groups representing (actual wind speed, actual motion pattern characteristics). The F estimate follows an F-distribution. 6 represents the degrees of freedom of molecules, which is also the dimension of correlation. This represents the degrees of freedom in the denominator.
[0142] B83. According to 6, With a rejection threshold of 0.95, the theoretical value of F is queried in the F distribution table, and the event determination result corresponding to the plant cluster is determined based on the relationship between the estimated value of F and the theoretical value of F.
[0143] If the F-statistic is greater than the F-estimate, the event is determined to be a debris flow event. If it is not greater than the F-estimate, the event is determined to be a non-debris flow event.
[0144] If the event determination result corresponding to a preset proportion of plant clusters is a debris flow event, proceed to B10; if the event determination result corresponding to a preset proportion of plant clusters is a debris flow event, proceed to B11.
[0145] B9: Determine whether the degree of surface difference between the three-dimensional regions of the ground surface in the current point cloud frame to be detected and the historical point cloud frames to be detected is greater than the surface difference threshold, and whether the degree of attitude angle difference between the radar attitude angles is greater than the attitude angle difference threshold, i.e., whether the difference is significant.
[0146] If the difference is greater than the ground surface difference threshold or the attitude angle difference threshold, proceed to B10; if the difference is not greater than the ground surface difference threshold or the attitude angle difference threshold, proceed to B11.
[0147] B10. A mudslide event has been confirmed.
[0148] B11. It has been determined that no mudslide event has occurred.
[0149] This application also provides a debris flow detection device. In some embodiments, the debris flow detection device includes a data acquisition module, an extraction module, and a determination module.
[0150] The system comprises: an acquisition module, which uses radar to sequentially acquire multiple point cloud frames of the site to be detected to form a point cloud sequence; an extraction module, which extracts the actual motion pattern features of plant clusters of the site to be detected from the point cloud sequence, wherein a plant cluster includes at least one plant; and a determination module, which determines whether a debris flow event has occurred at the site to be detected based on the actual motion pattern features of the plant clusters.
[0151] For further detailed descriptions of this embodiment, please refer to the preceding descriptions of the embodiments, which will not be repeated here.
[0152] Figure 10 This is a schematic diagram of the structure of an embodiment of the electronic device of this application. Figure 10 As shown, the electronic device 50 includes a memory 51 and a processor 52. The processor 52 is used to execute program instructions stored in the memory 51 to implement the steps in any of the above method embodiments. In a specific implementation scenario, the electronic device 50 may include, but is not limited to, a microcomputer or a server. In addition, the electronic device 50 may also include a laptop computer, a tablet computer, or other carrier device, which is not limited here.
[0153] Specifically, processor 52 controls itself and memory 51 to implement the steps in any of the above method embodiments. Processor 52 may also be referred to as a CPU (Central Processing Unit). Processor 52 may be an integrated circuit chip with signal processing capabilities. Processor 52 may also be a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component. A general-purpose processor may be a microprocessor or any conventional processor. Furthermore, processor 52 may be implemented using integrated circuit chips.
[0154] Please see Figure 11 , Figure 11This is a schematic diagram of a computer-readable storage medium according to an embodiment of the present application. The computer-readable storage medium 60 stores program instructions 601 thereon, which, when executed by a processor, implement the steps in any of the above method embodiments.
[0155] This application also provides a computer program product comprising a computer program that, when executed by a processor, can implement the steps of the methods described in any of the foregoing embodiments. Specifically, the computer program product can be a software or program product containing a computer program, capable of running on a computing device or stored on any available medium.
[0156] In some embodiments, the functions or modules of the apparatus provided in this disclosure can be used to perform the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.
[0157] The description of the various embodiments above tends to emphasize the differences between the various embodiments. The similarities or similarities between them can be referred to, and for the sake of brevity, they will not be repeated here.
[0158] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus implementations described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, units or components may be combined or integrated into another system, or some features may be ignored or not executed. In another image location, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.
[0159] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
Claims
1. A debris flow detection method, characterized in that, include: Multiple point cloud frames to be detected are collected sequentially by radar to form a point cloud sequence to be detected. The actual motion pattern features of the plant clusters in the detection site are extracted from the point cloud sequence to be detected, wherein a plant cluster includes at least one plant. Based on the actual movement pattern characteristics of the plant clusters, it is determined whether a debris flow event has occurred at the site to be detected.
2. The method according to claim 1, characterized in that, The step of extracting the actual motion pattern features of the plant clusters at the site to be detected from the point cloud sequence to be detected includes: The sub-motion pattern features of the plant cluster between each two adjacent points cloud frames to be detected are obtained respectively; The actual movement pattern characteristics of the plant cluster are obtained by statistically analyzing the characteristics of each sub-movement pattern of the plant cluster.
3. The method according to claim 2, characterized in that, The sub-motion pattern features include sub-motion amplitude features and sub-motion direction features. The step of acquiring the sub-motion pattern features of the plant cluster between each pair of adjacent point cloud frames to be detected includes: The three-dimensional regions of the plant clusters in each of the point cloud frames to be detected are obtained respectively. The three-dimensional regions of the plant clusters represent the outer contours of the three-dimensional points belonging to the plant clusters in the point cloud frames to be detected. For each pair of adjacent point cloud frames to be detected, the degree of overlap of the three-dimensional regions of the plant cluster in the two adjacent point cloud frames to be detected is obtained as the sub-motion amplitude feature; and, The sub-motion direction features of the three-dimensional regions of the plant cluster in two adjacent point cloud frames to be detected are obtained. When the distance between the two preceding and following points is greater than the distance between the preceding and following points, the sub-motion direction feature is designated as a first motion direction feature. When the distance between the preceding and following points is less than the distance between the preceding and following points, the sub-motion direction feature is designated as a second motion direction feature. The first and second motion direction features represent opposite motion directions. The preceding distance is the distance from the three-dimensional region of the plant cluster in the preceding point cloud frame to the radar, and the following distance is the distance from the three-dimensional region of the plant cluster in the following point cloud frame to the radar.
4. The method according to claim 2, characterized in that, The actual movement pattern characteristics include movement pattern central tendency characteristics and movement pattern dispersion characteristics. The statistical analysis of each sub-movement pattern characteristic of the plant cluster to obtain the actual movement pattern characteristics of the plant cluster includes: The central tendency and dispersion characteristics of each sub-movement pattern of the plant cluster are statistically analyzed to obtain the central tendency characteristics and dispersion characteristics of the movement patterns.
5. The method according to claim 1, characterized in that, The determination of whether a debris flow event has occurred at the site to be detected based on the actual movement pattern characteristics of the plant clusters includes: Obtain the actual wind speed of the plant cluster during the acquisition time period of the point cloud sequence to be detected; The actual correlation between the actual wind speed of the plant cluster and the actual motion pattern characteristics is obtained, and the reference correlation between the reference wind speed and the reference motion pattern characteristics of the plant cluster is obtained under the condition that the debris flow event has not occurred. Based on the degree of difference between the actual correlation and the reference correlation, it is determined whether the debris flow event has occurred at the site to be detected.
6. The method according to claim 5, characterized in that, Determining whether the debris flow event has occurred at the site to be detected based on the degree of difference between the actual correlation and the reference correlation includes: The null hypothesis of the F-test is that the actual correlation is the same as the reference correlation, and the F-estimate is calculated. Using the rejection threshold of the null hypothesis, the numerator degrees of freedom and the denominator degrees of freedom of the F estimator, the theoretical value of F is looked up in the F distribution table; In response to the estimated value of F being greater than the theoretical value of F, it is determined that the debris flow event has occurred at the site to be detected; If the estimated value of F is not greater than the theoretical value of F, it is determined that the debris flow event has not occurred at the site to be tested.
7. The method according to claim 1, characterized in that, Before determining whether a debris flow event has occurred at the site to be detected based on the actual movement pattern characteristics of the plant clusters, the process includes: Obtain the attitude angle of the radar during the acquisition time period of the point cloud sequence to be detected; The radar's attitude angle is combined with the actual motion pattern characteristics of the plant cluster to update the actual motion pattern characteristics of the plant cluster.
8. The method according to claim 1, characterized in that, The method further includes: The cloud frame of the current point to be detected has been acquired; Determine whether the current point cloud frame to be detected meets the surface extraction conditions. The surface extraction conditions include at least one of the following: the location to be detected includes the surface, the current three-dimensional region of the surface can be extracted from the current point cloud frame to be detected, and the area of the current three-dimensional region of the surface is greater than an area threshold. In response to the satisfaction of the surface extraction conditions, it is determined whether the debris flow event has occurred based on the current three-dimensional region of the surface; In response to the failure to meet the surface extraction conditions, the current point cloud frame to be detected and its adjacent preset number of historical point cloud frames to be detected are used to form the point cloud sequence to be detected, and the steps of extracting the actual motion pattern features of the plant clusters in the site to be detected from the point cloud sequence to be detected and subsequent steps are performed.
9. The method according to claim 1, characterized in that, The site to be detected includes multiple plant clusters. Based on the actual movement pattern characteristics of the plant clusters, determining whether a debris flow event has occurred at the site to be detected includes: For each plant cluster, the occurrence of the debris flow event is determined based on the actual movement pattern characteristics of the plant cluster, and the event determination result corresponding to the plant cluster is obtained. If the event determination result corresponding to a preset proportion of the plant clusters is that the debris flow event has occurred, then the debris flow event is finally determined to have occurred at the site to be detected.
10. An electronic device, characterized in that, It includes a memory and a processor, the processor being configured to execute program instructions stored in the memory to implement the method of any one of claims 1-9.
11. A computer-readable storage medium, characterized in that, It stores program instructions that, when executed by a processor, implement the method of any one of claims 1-9.
12. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1-9.