Laser Scanning-based Depth Tracking and Recognition Method

Abstract

The present application provides a depth tracking and recognition method based on laser scanning, including: limiting the key scanning area based on the historical data of canal emergencies, specifically including: obtaining and processing the historical data of canal emergencies, and dividing the scanning area of the canal cross-section according to the historical data; using a three-dimensional laser scanner to obtain the point cloud data of the limited area of the canal cross-section; scanning for abnormal data in the obtained point cloud data; identifying the type of emergency according to the abnormal data in the scanning; performing dynamic tracking based on the abnormal data and the type of emergency; and analyzing the abnormal point cloud data to generate emergency alarm information. The beneficial effects of this system include: eliminating the need for on-site monitoring by inspection personnel, improving the safety of monitoring, and dynamic monitoring ensuring that when emergencies occur in sewers or canals, they can be detected in a timely manner, and the type of emergency and the specific location of the emergency can be identified. Finally, selective scanning and storage of data save a large amount of memory of the system and ensure the high efficiency of system operation.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent laser scanning, and particularly to a depth tracking and recognition method based on laser scanning.

Background Art

[0002] In environments such as water channels and sewer pipes that are difficult to access, portable scanning and small-scene scanning play very important roles.

[0003] Existing highways and urban roads are mainly divided into three types: cement concrete roads, asphalt concrete roads, and recycled asphalt concrete roads. At intervals of several hundred meters on these roads, a sewer or water channel can be seen, so that there will be no water accumulation on the road in bad weather such as heavy rain days, thus ensuring the smooth driving of the road. However, with the long-term use of these sewers, some sundries and sludge will accumulate inside the sewers. As these sundries and sludge increase, the sewers will be blocked. Once the sewers are blocked, rainwater and sewage cannot be discharged and will overflow onto the road, which will seriously affect the driving of traffic vehicles. At present, the urban sewer pipe network is intricate. During the process of detecting the smoothness of the sewer pipe network, generally, infrared detection equipment and laser detection equipment are used to detect the liquid level situation in the sewer pipe network, and then the liquid level signals detected by the infrared detection equipment and the laser detection equipment are analyzed and processed, and finally the smoothness of the sewer pipe network is obtained. The existing liquid level situation in the sewer pipe network is detected manually and cannot be detected in real time. When the liquid level rises due to the blockage of the sewer pipe network, maintenance personnel cannot discover and handle it in time, resulting in a low maintenance efficiency of the sewer pipe network. Therefore, it is necessary to monitor the danger situation of the sewer and give early warnings in time. The said danger situations specifically include: blockage, deformation, water flow rate, floating objects, rodent damage, etc. Some of these danger situations need to be monitored dynamically for a long time to ensure the normal function of the sewer or water channel.

Summary of the Invention

[0004] A depth tracking and recognition method based on laser scanning provided by an embodiment of the present invention includes:

[0005] Defining a key scanning area based on historical data of water channel danger situations: obtaining and processing historical data of water channel danger situations, and dividing the scanning area of the water channel cross-section according to the historical data;

[0006] Using a three-dimensional laser scanner to obtain point cloud data of the limited area of the water channel cross-section;

[0007] Scanning abnormal data in the obtained point cloud data: correcting the scanning data above and below the water, and obtaining an abnormal point cloud data set;

[0008] Identifying the type of danger situation according to the abnormal data in the scanning: detecting and identifying the type of foreign objects in the water channel, monitoring the change in the shape of the water channel itself, and monitoring the abnormal water flow rate of the water channel;

[0009] Perform dynamic tracking based on abnormal data and danger types: dynamic monitoring of foreign objects, dynamic monitoring of abnormal water flow;

[0010] Analyze the abnormal point cloud data to generate danger alarm information.

[0011] Preferably, the historical data of the canal danger is used to define the key scanning area, including:

[0012] Obtain and process the historical data of the canal danger;

[0013] Divide the scanning area of the canal cross-section according to the historical data.

[0014] Preferably, the obtaining and processing of the historical data of the canal danger includes:

[0015] Obtain the historical data of the canal cross-section danger; wherein, the historical data includes the danger type and its point cloud data, the number of occurrences of various dangers, the location partition where the danger occurs, and the specific location coordinates within the partition; the location partition includes the top, left and right sides, bottom and water surface of the canal cross-section;

[0016] Extract the characteristics of different dangers in the historical data, establish a standard danger matrix, the row vector is each location partition, the column vector is the danger characteristic, and count the types and numbers of dangers occurring in each location partition in the historical data table.

[0017] Preferably, the dividing of the scanning area of the canal cross-section according to the historical data includes:

[0018] Divide the canal cross-section into a preset number of regions:

[0019] Divide the location partition into a preset number of sub-regions of equal size;

[0020] Perform clustering processing on the historical data in the above sub-regions;

[0021] The historical data is divided into the corresponding sub-regions according to its specific location coordinates within the partition;

[0022] Obtain a risk model, and the historical data is used as the training data of the risk model and is divided into high-risk, medium-risk and risk-free regions;

[0023] For the point cloud data to be processed, preprocess it into data with the same characteristics as the training data of the risk model;

[0024] According to the risk model, determine whether the preprocessed data conforms to the characteristics of the high-risk region. If so, determine that the sub-region belongs to the high-risk region;

[0025] Perform the above processing on all sub - regions of all position partitions and merge adjacent same - risk regions within the same position partition; among them, high - risk and medium - risk regions are areas to be scanned, and risk - free regions are non - scanned areas.

[0026] Preferably, obtaining the point cloud data of the defined area of the water channel cross - section by using a 3D laser scanner includes:

[0027] Set the scanning range of the 3D laser scanner according to the scanning division and scan the water channel cross - section by using the 3D laser scanner; among them, the 3D laser scanner includes:

[0028] A support assembly, a rotation drive assembly arranged above the support assembly, a laser sensor rotation assembly, a GPS positioning assembly arranged above the support assembly, a laser sensor arranged on the laser sensor rotation assembly, and a control box assembly arranged below the support assembly;

[0029] The rotation drive assembly is used to drive the laser sensor rotation assembly to rotate;

[0030] The control box assembly is connected to the rotation drive assembly and the laser sensor through cables;

[0031] Obtain the required point cloud data: scan the high - risk and medium - risk regions, do not scan the risk - free regions, and set different scanning periods for the high - risk and medium - risk regions;

[0032] Dynamically adjust the scanning period according to the update of historical data: set the scanning period according to the actual situation, set a shorter scanning period for high - risk regions and a longer scanning period for medium - risk regions;

[0033] The point cloud data obtained each time of scanning is included in the historical data to update the historical data;

[0034] If the determination result of the risk region changes after the historical data is updated, adjust the scanning period of this region according to the changed determination result.

[0035] Preferably, the abnormal data in the scanned point cloud data includes:

[0036] Scan a standard water channel without danger to obtain a standard point cloud data set; the standard point cloud data set includes data scanned above water and data scanned underwater;

[0037] Correct the scanned data above and below water, including:

[0038] Pre - process the obtained point cloud data to obtain the pre - processed point cloud data;

[0039] Performing edge recognition on the preprocessed point cloud data to obtain a number of edge points;

[0040] Connecting each of the edge points once to obtain the water - above - water and water - below - water boundary line and the canal boundary line;

[0041] Dividing the point cloud data into two parts: above water and below water according to the water - above - water and water - below - water boundary line;

[0042] Performing the above - mentioned processing on the standard point cloud data set to obtain the canal boundary lines scanned above water and below water;

[0043] Comparing the canal boundary line obtained from the standard point cloud data set with the point cloud data to be corrected. If there is an offset, obtain the offset amount of each corresponding position coordinate;

[0044] Performing summation and averaging processing on the offset amounts to obtain the final correction amount and correcting the position coordinates;

[0045] Obtaining an abnormal point cloud data set, including:

[0046] Taking the standard point cloud data set as a sample set, extracting the reflection intensity information and coordinate data of the data as feature values;

[0047] Training a data point prediction model through a decision tree algorithm, inputting the point cloud data to be processed into the above - mentioned prediction model, and outputting a normal point cloud data set;

[0048] Removing the normal point cloud data set from the original point cloud data to obtain the abnormal data in the point cloud data.

[0049] Preferably, the recognition of the danger type according to the abnormal data in the scan includes:

[0050] Foreign object detection and type recognition in the canal:

[0051] The foreign object detection mainly includes the monitoring of floating objects and rodent damage. Mapping the abnormal point cloud data to generate a distance image and a reflection intensity image;

[0052] Performing point cloud segmentation and clustering on the abnormal point cloud data according to the distance image and the reflection intensity image to obtain a plurality of point cloud regions;

[0053] Performing feature extraction on each point cloud region, and classifying by the extracted feature vectors to identify the target, including: training a target detection model for rats; inputting the extracted feature vectors into the target detection model, and outputting the recognition result as rodent damage; extracting the feature values of the point cloud data within a preset range near the water - above - water and water - below - water boundary line and comparing with the standard point cloud data set to determine whether there is abnormal data. If so, output the recognition result as a floating object;

[0054] Monitoring of the morphological changes of the canal itself:

[0055] After the above foreign object detection, floating objects and rodent damage data in the abnormal data are removed to obtain the remaining point cloud data;

[0056] Edge data is extracted from the point cloud data through an edge extraction algorithm;

[0057] Similarly, edge data is extracted from the standard point cloud data through the above edge extraction algorithm. The edge extraction algorithm can adopt the Robert operator to extract edges in the image according to the formula:

[0058]

[0059] Compare the above two groups of data to obtain the similarity;

[0060] Set a threshold. When the similarity is higher than the threshold, it is considered that the water channel has not deformed;

[0061] Monitor the abnormal water flow of the water channel:

[0062] Based on the pre-obtained water surface and underwater demarcation line, the horizontal height of the demarcation line is obtained with the lowest point of the water channel as the reference, and different height thresholds are set corresponding to different water flow situations.

[0063] Preferably, the dynamic tracking based on abnormal data and danger types includes:

[0064] Dynamic monitoring of foreign objects:

[0065] A three-dimensional laser scanner is installed at a preset site of the water channel. After abnormal data is obtained at a certain site and its type is identified, the abnormal data is uploaded to the server;

[0066] The server dynamically tracks the floating object and rodent damage danger types in the abnormal data, including: extracting the characteristic values of the floating object or rodent damage danger; the server determines the sites that need to be called for dynamic tracking according to the water flow direction and the position information of the abnormal data, and sends a scanning instruction to the three-dimensional laser scanner at the site. The point cloud data obtained by scanning at this site is then transmitted back to the server; the server identifies whether the uploaded data is the danger to be tracked according to the characteristic values; if so, the instruction is sent to the next site, and if not, the site returns to the normal working state; obtain the original scanned point cloud data of the water channel, and perform three-dimensional surface reconstruction after preprocessing to construct a three-dimensional model of the water channel; obtain the abnormal data and its type, mark the danger on the three-dimensional model according to its position information, and connect the same danger monitored at different sites to obtain the danger movement trajectory;

[0067] Dynamic monitoring of abnormal water flow:

[0068] Obtain the water flow conditions at each site and compare the water flow differences between adjacent sites one by one;

[0069] Set a threshold for the difference. When the threshold is exceeded, it indicates that there is a blockage between the two sites being compared;

[0070] Adjust the scan period of the two sites to a preset value and set the maximum number of scans;

[0071] If the water flow difference between the two sites drops below the threshold before reaching the maximum number of scans, no processing is done. Otherwise, upload the position information and water flow difference of the two sites to the server.

[0072] Preferably, the analysis of abnormal point cloud data to generate danger warning information includes:

[0073] Obtain the abnormal point cloud data of dynamic tracking monitoring, and count the number of floating objects and the sites where floating objects are detected within a certain period of time;

[0074] Establish a prediction model of the blockage probability corresponding to the number of floating objects through a linear regression algorithm to predict the magnitude of the blockage probability;

[0075] Set a threshold for the magnitude of the probability. When the prediction result is higher than the threshold, the server will generate blockage warning information, where the blockage warning information includes the number of floating objects, the detected sites, and the predicted blockage probability;

[0076] Obtain the monitoring data of the morphological changes of the water channel itself;

[0077] Set a minimum value for the edge similarity according to the construction requirements of the water channel. When the monitoring data is lower than the set minimum value, the server will generate water channel deformation warning information, where the water channel deformation warning information includes the edge data and similarity of the water channel;

[0078] Count the number of rodent damage danger situations occurring within a certain period of time in the abnormal point cloud data;

[0079] Establish a prediction model of the rodent damage risk value corresponding to the number of rodent damage danger situations through a linear regression algorithm to predict the level of rodent damage risk;

[0080] Set a threshold for the level of risk. When the prediction result is higher than the threshold, the server will generate rodent damage warning information, and the rodent damage warning information includes the number of rodent damage within a certain period of time and the predicted rodent damage risk value.

[0081] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0082] First, it is not necessary for inspectors to conduct on-site monitoring, which improves the safety of monitoring. Secondly, dynamic monitoring ensures that when a danger occurs in a sewer or water channel, it can be detected in a timely manner and the type of danger and its specific location can be identified, facilitating maintenance personnel to eliminate the danger as soon as possible. Finally, selectively scanning and storing data saves a large amount of memory in the system, reduces redundant data, and ensures the high efficiency of system operation.

Description of the Drawings

[0083] Figure 1 It is a flowchart of the depth tracking and recognition method based on laser scanning of the present invention.

[0084] Figure 2 It is a system structure diagram of the three-dimensional laser scanner of the present invention.

Detailed Embodiment

[0085] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be described in detail below with reference to the drawings and specific embodiments.

[0086] Figure 1 It is a flowchart of a small integrated laser depth tracking and recognition method of the present invention. As Figure 1 shown, the depth tracking and recognition system based on laser scanning in this embodiment may specifically include:

[0087] Step 101, defining a key scanning area based on historical data of water channel dangers.

[0088] Obtain and process historical data of water channel dangers.

[0089] Obtain historical data of water channel cross-section dangers. The historical data includes: danger types and their point cloud data, the number of occurrences of various dangers, the location partitions where dangers occur, and the specific location coordinates within the partitions. The location partitions specifically include: the top, left and right sides, bottom, and water surface of the water channel cross-section. Extract the characteristics of different dangers in the historical data, establish a standard danger matrix, with row vectors being each location partition and column vectors being danger characteristics, and count the types and numbers of dangers occurring in each location partition in the historical data table.

[0090] Divide the scanning area of the water channel cross-section according to historical data.

[0091] Divide the cross-section of the water channel into a number of preset regions, specifically including: divide the position partition into a preset number of sub-regions of equal size, perform clustering processing on the historical data in the above sub-regions, and divide the historical data into corresponding sub-regions according to their specific position coordinates within the partition. Obtain a risk model, and divide the historical data, which serves as the training data for the risk model, into high-risk, medium-risk, and risk-free regions. For the point cloud data to be processed, preprocess it into data with the same characteristics as the training data of the risk model. According to the risk model, determine whether the preprocessed data conforms to the characteristics of the high-risk region. If so, determine that the sub-region belongs to the high-risk region. Perform the above processing on all sub-regions of all position partitions and merge adjacent same-risk regions within the same position partition. High-risk and medium-risk regions are used as the areas to be scanned, and risk-free regions are non-scanned areas. For example, divide the bottom of the cross-section of the water channel into 10 preset sub-regions, process the historical data of these 10 sub-regions through the risk model to obtain the risk types of the sub-regions, and merge adjacent same-risk regions to obtain the areas to be scanned and non-scanned areas.

[0092] Step 102: Use a three-dimensional laser scanner to obtain the point cloud data of the limited area of the cross-section of the water channel.

[0093] Set the scanning range of the three-dimensional laser scanner according to the scanning division. Use the three-dimensional laser scanner to scan the cross-section of the water channel. The three-dimensional laser scanner specifically includes: a support assembly, a rotary drive assembly arranged above the support assembly, a laser sensor rotary assembly, a GPS positioning assembly arranged above the support assembly, a laser sensor arranged on the laser sensor rotary assembly, and a control box assembly arranged below the support assembly. The rotary drive assembly is used to drive the laser sensor rotary assembly to rotate. The control box assembly is connected to the rotary drive assembly and the laser sensor through cables; obtain the required point cloud data, specifically referring to scanning the high-risk and medium-risk regions and not scanning the risk-free regions. Set different scanning cycles for the high-risk and medium-risk regions and dynamically adjust the scanning cycle according to the update of the historical data. Specifically include: set the scanning cycle according to the actual situation, set a shorter scanning cycle for the high-risk region, and set a longer scanning cycle for the medium-risk region. The point cloud data obtained each time is included in the historical data to update the historical data. If the determination result of the risk region changes after the historical data is updated, adjust the scanning cycle of the region according to the changed determination result. For example: the scanning cycle set for a certain high-risk region is 2 days. After scanning for a month without any danger, after the scanning results of this month are included in the historical data, the region is determined to be a medium-risk region, and then the scanning cycle of this region is changed to 5 days.

[0094] Step 103: Scan to obtain abnormal data in the point cloud data.

[0095] Scan a standard and danger-free water channel to obtain a standard point cloud data set, which includes data scanned above water and data scanned underwater.

[0096] Correction of above-water and underwater scan data.

[0097] Preprocess the obtained point cloud data to obtain preprocessed point cloud data. Then, perform edge recognition on the preprocessed point cloud data to obtain several edge points, connect each of the edge points in sequence to obtain the water surface - underwater boundary line and the water channel boundary line. Divide the point cloud data into two parts, above water and underwater, according to the water surface - underwater boundary line. Perform the above processing on the standard point cloud data set as well to obtain the water channel boundary lines obtained from above-water and underwater scans. Compare the water channel boundary line obtained from the standard point cloud data set with the point cloud data to be corrected. If there is an offset, obtain the offset amount of each corresponding position coordinate, perform a sum-average process on the offset amounts to obtain the final correction amount, and correct the position coordinates. For example, if it is found through comparison that most data points have an offset to the right, then add these offset amounts and calculate the average value to obtain the correction amount, and adjust the data points to the left according to the correction amount.

[0098] Obtain an abnormal point cloud data set.

[0099] Use the standard point cloud data set as a sample set, extract the reflection intensity information and coordinate data of the data as feature values, train a data point prediction model through a decision tree algorithm, input the point cloud data to be processed into the above prediction model, output a normal point cloud data set, and remove the normal point cloud data set from the original point cloud data to obtain the abnormal data in the point cloud data. For example, if the original point cloud data set is A, and a normal point cloud data set B is obtained after being processed by the data point prediction model, remove the same data in data set A and data set B to obtain data set C, then data set C is the required abnormal point cloud data set.

[0100] Step 104: Identify the type of danger according to the abnormal data in the scan.

[0101] Foreign object detection and type identification in the water channel.

[0102] The foreign object detection mainly includes floating object detection and rodent damage monitoring. The abnormal point cloud data is mapped to generate a distance image and a reflection intensity image. Based on the distance image and the reflection intensity image, the abnormal point cloud data is segmented and clustered to obtain multiple point cloud regions. Feature extraction is performed on each point cloud region, and the extracted feature vectors are classified to identify the target. Specifically, it includes: training a target detection model for mice, inputting the extracted feature vectors into the target detection model, and the output recognition result is rodent damage; detecting whether there is abnormal data within a preset range near the water surface - underwater boundary. If so, the recognition result is a floating object. For example, detecting whether there is abnormal data within a banded area 5 cm above and 5 cm below the boundary. If so, the abnormal data comes from floating objects on the water surface.

[0103] Monitoring the morphological changes of the water channel itself.

[0104] After the above - mentioned foreign object detection, the floating object and rodent damage data in the abnormal data are removed to obtain the remaining point cloud data. Edge data is extracted from the point cloud data through an edge extraction algorithm. Similarly, edge data is extracted from the standard point cloud data through the above - mentioned edge extraction algorithm. The edge extraction algorithm can use the Robert operator, and the edges in the image are extracted according to the formula:

[0105]

[0106] Compare the above two groups of data to obtain the similarity, and set a threshold. When the similarity is higher than the threshold, it is considered that the water channel has not deformed.

[0107] Monitoring the abnormal water flow of the water channel.

[0108] Based on the pre - obtained water surface - underwater boundary, the horizontal height of the boundary is obtained with the lowest point of the water channel as the reference. Different height thresholds are set corresponding to different water flow situations. For example, two height thresholds are set, which are 50 cm and 80 cm respectively. When the horizontal height of the water surface - underwater boundary is lower than 50 cm, it is a low water level; when it is higher than 50 cm and lower than 80 cm, it is a normal water level; when it is higher than 80 cm, it is an abnormal water level.

[0109] Step 105, perform dynamic tracking based on the abnormal data and the type of danger situation.

[0110] Dynamic monitoring of foreign objects.

[0111] A 3D laser scanner is installed at a preset site of the water channel. After abnormal data is obtained at a certain site and its type is identified, the abnormal data is uploaded to the server. The server dynamically tracks the types of floating objects and rodent damage risks in the abnormal data, specifically including: extracting the characteristic values of the floating objects or rodent damage risks. The server determines the sites that need to be called for dynamic tracking according to the water flow direction and the position information of the abnormal data, and sends a scanning instruction to the 3D laser scanner at the site. The point cloud data obtained by scanning at this site is then transmitted back to the server. The server identifies the risk of the uploaded data according to the characteristic values, whether it is the risk to be tracked. If so, the instruction is sent to the next site. If not, the site returns to the normal working state; obtaining the original water channel scanned point cloud data, performing 3D surface reconstruction after preprocessing to construct a 3D model of the water channel, obtaining the abnormal data and its type, and marking the risk on the 3D model according to its position information. Connecting the same risks monitored at different sites to obtain the risk movement trajectory. For example: a floating object is monitored at site A, and sites B and C are in sequence along the water flow direction. The server sends a scanning instruction to site B. When the scanning result transmitted back by B shows that there is a floating object again, the server sends a scanning instruction to site C.

[0112] Dynamic monitoring of abnormal water flow.

[0113] Obtain the water flow conditions at each site, compare the water flow differences between adjacent sites one by one, set the threshold of the difference. When the threshold is exceeded, it indicates that there is a blockage between the two sites being compared. The scanning periods of the two sites are adjusted to the preset value, and the maximum number of scans is set. If the water flow difference between the two sites drops below the threshold before the maximum number of scans is reached, no processing is performed. Otherwise, the position information and water flow difference of the two sites are uploaded to the server.

[0114] Step 106, analyze the abnormal point cloud data to generate a risk alarm message.

[0115] Obtain abnormal point cloud data for dynamic tracking and monitoring, count the number of floating objects and the locations where floating objects are detected within a certain period of time, establish a prediction model for the corresponding blockage probability based on the number of floating objects through a linear regression algorithm, predict the magnitude of the blockage probability, set a threshold for the magnitude of the probability, and when the prediction result is higher than the threshold, the server will generate a blockage alarm message. The blockage alarm message specifically includes: the number of floating objects, the detected locations, and the predicted probability of blockage; obtain monitoring data on the morphological changes of the water channel itself, set a minimum value for the edge similarity according to the construction requirements of the water channel, and when the monitoring data is lower than the set minimum value, the server will generate a water channel deformation alarm message. The water channel deformation alarm message specifically includes: the edge data of the water channel and the similarity; count the number of rodent damage emergencies occurring within a certain period of time in the abnormal point cloud data, establish a prediction model for the corresponding rodent infestation risk value based on the number of rodent damage emergencies through a linear regression algorithm, predict the level of the rodent infestation risk, set a threshold for the level of the risk, and when the prediction result is higher than the threshold, the server will generate a rodent damage alarm message. The rodent damage alarm message specifically includes: the number of rodent damage incidents within a certain period of time and the predicted rodent infestation risk value.

[0116] The device for implementing the above method may include a lidar module, a power management system, an integrated device housing, a Linux embedded board for data processing, embedded software, and a display screen. The lidar module adopts a multi-layer printed PCB process, with a small volume and low power consumption. The display screen is connected to the embedded board through an HDMI interface, has a touch function, can display the real-time scanning effect through the display screen, and at the same time stores the scanned data in the hard disk in the industry-standard point cloud format, communicates with the embedded board through USB, and can be imported into a computer for post-processing. In terms of the mechanical structure, it is fixed in difficult-to-enter environments such as water channels and sewer pipes through a telescopic bracket and a thin connecting rod for scanning. When the device reaches a far distance, the touch screen can be separated from the device, and the user can see the scanning effect up close. By defining an area of interest where the target is located in the scanned scene, establishing a target model for it, and applying it to the above step method, target tracking and risk prediction are carried out.

[0117] The above are only embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structural or equivalent process transformation made using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

[0118] The program for implementing information control of the present invention can be written in one or more programming languages or combinations thereof to write computer program code for performing the operations of the present invention. The programming languages include object-oriented programming languages such as Java, Python, C++, and also include conventional procedural programming languages such as the C language or similar programming languages.

[0119] The program code can be executed entirely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, by connecting through the Internet using an Internet service provider).

[0120] In several embodiments provided by the present invention, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation.

[0121] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0122] In addition, each functional unit in various embodiments of the present invention can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of a combination of hardware and software functional units.

[0123] The above integrated units implemented in the form of software functional units can be stored in a computer-readable storage medium. The above software functional units stored in a storage medium include several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute some steps of the methods described in various embodiments of the present invention.

[0124] The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, external hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

Claims

1. A depth tracking and recognition method based on laser scanning, characterized in that, Including: Defining a key scanning area based on historical data of canal emergencies: obtaining and processing historical data of canal emergencies, and dividing the scanning area of the canal cross-section according to the historical data; Obtaining point cloud data of the defined area of the canal cross-section using a 3D laser scanner; Scanning for abnormal data in the point cloud data: correcting the above-water and underwater scanning data, and obtaining an abnormal point cloud data set; Identifying the type of emergency according to the abnormal data in the scanning: detecting and identifying foreign objects in the canal and their types, monitoring the morphological changes of the canal itself, and monitoring the abnormal water flow of the canal; Performing dynamic tracking based on the abnormal data and the type of emergency: dynamically monitoring foreign objects and dynamically monitoring abnormal water flow; Analyzing the abnormal point cloud data to generate emergency alarm information; Among them, the defining of the key scanning area based on historical data of canal emergencies includes: Obtaining and processing historical data of canal emergencies; Dividing the scanning area of the canal cross-section according to the historical data; Among them, the obtaining and processing of historical data of canal emergencies includes: Obtaining historical data of canal cross-section emergencies; among them, the historical data includes the type of emergency and its point cloud data, the number of occurrences of various emergencies, the location partitions where the emergencies occur, and the specific location coordinates within the partitions; the location partitions include the top, left and right sides, bottom, and water surface of the canal cross-section; Extracting the characteristics of different emergencies in the historical data, establishing a standard emergency matrix, with row vectors being each location partition and column vectors being emergency characteristics, and counting the types and numbers of emergencies occurring in each location partition in the historical data table; Among them, the dividing of the scanning area of the canal cross-section according to the historical data includes: Dividing the canal cross-section into a preset number of regions: Dividing the location partitions into a preset number of sub-regions of equal size; Performing clustering processing on the historical data in the above sub-regions; The historical data is divided into the corresponding sub-regions according to the specific location coordinates within its partition; Obtaining a risk model, and dividing the historical data, as the training data of the risk model, into high-risk, medium-risk, and risk-free regions; For the point cloud data to be processed, preprocessing it into data with the same characteristics as the training data of the risk model; According to the risk model, determining whether the preprocessed data conforms to the characteristics of the high-risk region. If so, determining that the sub-region belongs to the high-risk region; Performing the above processing on all sub-regions of all location partitions and merging adjacent same-risk regions within the same location partition; among them, high-risk and medium-risk regions are used as the areas to be scanned, and risk-free regions are non-scanning areas; Among them, the obtaining of point cloud data of the defined area of the canal cross-section using a 3D laser scanner includes: Setting the scanning range of the 3D laser scanner according to the scanning division and using the 3D laser scanner to scan the canal cross-section; among them, the 3D laser scanner includes: A support assembly, a rotary drive assembly arranged above the support assembly, a laser sensor rotary assembly, a GPS positioning assembly arranged above the support assembly, a laser sensor arranged on the laser sensor rotary assembly, and a control box assembly arranged below the support assembly; The rotation drive assembly is used to drive the rotation of the laser sensor rotation assembly; The control box assembly is connected to the rotation drive assembly and the laser sensor through a cable; Obtain the required point cloud data: scan high-risk areas and medium-risk areas, do not scan risk-free areas, and set different scan cycles for high-risk areas and medium-risk areas; Dynamically adjust the scan cycle according to the update of historical data: set the scan cycle according to the actual situation, set a short scan cycle for high-risk areas, and set a long scan cycle for medium-risk areas; The point cloud data obtained from each scan is included in the historical data to update the historical data; If the determination result of the risk area changes after the historical data is updated, adjust the scan cycle of the area according to the changed determination result.

2. The method according to claim 1, wherein The abnormal data in the scanned point cloud data includes: Scan a standard and hazard-free water channel to obtain a standard point cloud data set; the standard point cloud data set includes data scanned above water and data scanned underwater; Correct the scanned data above and below water, including: Preprocess the obtained point cloud data to obtain preprocessed point cloud data; Perform edge recognition on the preprocessed point cloud data to obtain a number of edge points; Connect each of the edge points in sequence to obtain the water surface-water bottom boundary line and the water channel boundary line; Divide the point cloud data into two parts, above water and underwater, according to the water surface-water bottom boundary line; Perform the above processing on the standard point cloud data set to obtain the water channel boundary lines obtained from scanning above water and underwater; Compare the water channel boundary lines obtained from the standard point cloud data set with the point cloud data to be corrected. If there is an offset, obtain the offset of each corresponding position coordinate; Perform a summation and averaging process on the offset to obtain the final correction amount and correct the position coordinates; Obtain an abnormal point cloud data set, including: Take the standard point cloud data set as a sample set, and extract the reflection intensity information and coordinate data of the data as feature values; Train a data point prediction model through a decision tree algorithm, input the point cloud data to be processed into the above prediction model, and output a normal point cloud data set; Remove the normal point cloud data set from the original point cloud data to obtain the abnormal data in the point cloud data.

3. The method according to claim 1, characterized in that Identifying the type of hazard according to the abnormal data in the scan, including: Foreign object detection and type identification in the water channel: The foreign object detection includes monitoring of floating objects and rodent damage. Map the abnormal point cloud data to generate a distance image and a reflection intensity image; Perform point cloud segmentation and clustering on the abnormal point cloud data according to the distance image and the reflection intensity image to obtain a number of point cloud regions; Extract features from each point cloud region, and classify them by the extracted feature vectors to identify the target, including: training a target detection model for rats; input the extracted feature vectors into the target detection model, and the output recognition result is rodent damage; extract the feature values of the point cloud data within a preset range near the water surface-water bottom boundary line and compare them with the standard point cloud data set to determine whether there is abnormal data; if so, the output recognition result is a floating object; Monitoring of the morphological changes of the water channel itself: After the above foreign object detection, floating objects and rodent damage data in the abnormal data are removed to obtain the remaining point cloud data; Edge data is extracted from the remaining point cloud data through an edge extraction algorithm; Similarly, edge data is extracted from the standard point cloud data through the above edge extraction algorithm, and the Robert operator is used for the edge extraction algorithm; The edge data in the remaining point cloud data and the edge data in the standard point cloud data are compared to obtain the similarity; A threshold is set. When the similarity is higher than the threshold, it is considered that the water channel has not deformed; Monitor the abnormal water flow of the water channel: Based on the pre-obtained water surface and underwater demarcation line, the horizontal height of the demarcation line is obtained with the lowest point of the water channel as the reference, and different height thresholds are set corresponding to different water flow conditions.

4. The method according to claim 1, characterized in that, The dynamic tracking based on abnormal data and danger types includes: Dynamic monitoring of foreign objects: A three-dimensional laser scanner is installed at a preset site of the water channel. After abnormal data is obtained at a certain site and its type is identified, the abnormal data is uploaded to the server; The server dynamically tracks the floating object and rodent damage danger types in the abnormal data, including: extracting the characteristic values of the floating object or rodent damage danger; the server determines the sites that need to be called for dynamic tracking according to the water flow direction and the position information of the abnormal data, and sends a scanning instruction to the three-dimensional laser scanner at the site. The point cloud data obtained by scanning at the site is transmitted back to the server; the server identifies the danger of the uploaded data according to the characteristic values, whether it is the danger to be tracked; if so, the instruction is sent to the next site, if not, the site returns to the normal working state; obtain the original scanned point cloud data of the water channel, and perform three-dimensional surface reconstruction after preprocessing to construct a three-dimensional model of the water channel; obtain the abnormal data and its type, and mark the danger on the three-dimensional model according to its position information, and connect the same danger monitored at different sites to obtain the danger movement trajectory; Dynamic monitoring of abnormal water flow: Obtain the water flow conditions of each site, and compare the water flow differences between adjacent sites one by one; Set a threshold for the difference. When the threshold is exceeded, it means that there is a blockage between the two sites being compared; The scanning periods of the two sites are adjusted to a preset value, and the maximum number of scans is set; If the water flow difference between the two sites drops below the threshold before the maximum number of scans is reached, no processing is performed, otherwise the position information and water flow difference of the two sites are uploaded to the server.

5. The method according to claim 1, characterized in that The analysis of abnormal point cloud data to generate danger alarm information includes: Obtain the abnormal point cloud data monitored dynamically, and count the number of floating objects and the sites where floating objects are monitored within a certain period of time; Establish a prediction model of the blockage probability corresponding to the number of floating objects through a linear regression algorithm to predict the size of the blockage probability; Set a threshold for the size of the probability. When the prediction result is higher than the threshold, the server will generate a blockage alarm information, where the blockage alarm information includes the number of floating objects, the monitored sites, and the predicted blockage probability; Obtain the monitoring data of the morphological changes of the water channel itself; Set the minimum value of the edge similarity according to the construction requirements of the water channel. When the monitored data is lower than the set minimum value, the server will generate a water channel deformation alarm message, where the water channel deformation alarm message includes the edge data and similarity of the water channel; Count the number of occurrences of rodent damage danger situations within a certain period of time in the abnormal point cloud data; Establish a prediction model of the rodent damage risk value corresponding to the number of rodent damage danger situations through a linear regression algorithm to predict the level of rodent damage risk; Set the threshold of the risk level. When the prediction result is higher than this threshold, the server will generate a rodent damage alarm message, and the rodent damage alarm message includes the number of rodent damage within a certain period of time and the predicted rodent damage risk value.

Images