Intelligent inspection method and system based on SLAM algorithm imaging

Abstract

The invention discloses a robot intelligent inspection method and system based on SLAM algorithm imaging, and relates to the technical field of algorithm imaging, and the method comprises the following steps: collecting information data of a target area, and carrying out the real-time processing of the information data of the target area; through a multi-source data optimization algorithm and in combination with the information data of the target area, the pose of the robot is obtained in real time; and dynamically correcting the pose error according to the pose self-adaptive registration of the robot, constructing a three-dimensional map, and generating visual output. According to the invention, through fusion of the laser radar point cloud data and the multispectral sensor, dual identification of millimeter-level space coordinate positioning and material features of the cable joint is realized, and the problem of fuzzy positioning caused by a single visual angle in traditional two-dimensional image identification is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of algorithm imaging, and in particular to an intelligent inspection method and system based on SLAM algorithm imaging. Background Art

[0002] With the continued advancement of urban underground utility corridors and the gradual implementation of overhead line laying projects in Shanghai, underground cable power supply will be a development trend in megacities like Shanghai. The number and total length of power cables will continue to increase, and the reliability and safety of power supply will become increasingly important. Cable identification is a crucial step in operations such as undergrounding overhead lines and cable relocation and reconnection surveys, and must be accurate. Failure to do so will directly impact the safety of personnel and equipment. Cable trenches are a common method for laying cables. These environments are narrow and have complex internal shielding structures. Once cables are buried, subsequent identification and maintenance are difficult, and effective digital maps are difficult to create. Scientific cable path management tools are lacking, especially when cable joint locations are unclear, making it difficult to implement protective measures early. Furthermore, human factors can lead to inefficient and error-prone inspections.

[0003] With the acceleration of urbanization, the scale of power systems continues to expand, and cable laying methods are becoming increasingly diverse. Cable trenches, a common method for laying cables, offer advantages such as high space utilization and excellent safety. However, the narrow interior space and complex obstruction structures of cable trenches make subsequent identification and maintenance of buried cables difficult. Effective digital maps are difficult to create, and there is a lack of scientific cable route management tools. In particular, unclear cable joint locations hinder early protective measures. These issues pose significant challenges to cable maintenance and management. Traditional cable inspection methods rely primarily on manual inspections, which are not only inefficient but also pose safety risks. Furthermore, due to the complex environment within cable trenches, manual inspections often fail to fully cover all areas, leading to the omission of potential fault points. Therefore, there is an urgent need for intelligent devices that can automatically complete cable inspection tasks. Current 3D image reconstruction technologies for complex cable trench environments suffer from issues such as data noise and accuracy, inaccurate algorithm recognition, difficulty integrating sensor technologies, and real-time and efficiency issues. Addressing these technical shortcomings requires further research and innovation, potentially involving improvements in data acquisition technology, optimization of intelligent algorithms, enhanced performance of virtual reality technology, and better integration of multi-sensor fusion. Summary of the Invention

[0004] In view of the problems existing in the existing intelligent inspection method and system based on SLAM algorithm imaging, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that the position of the cable joint is unclear, protective measures cannot be taken in time, and human factors may lead to low detection efficiency and easy errors.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides an intelligent inspection method based on SLAM algorithm imaging, which includes the following steps:

[0008] Collect information data of the target area and process the information data of the target area in real time;

[0009] The robot's position and posture are acquired in real time by combining the information data of the target area through the first algorithm;

[0010] Dynamically correct the posture error based on the robot's posture adaptive registration, build a three-dimensional map, and generate visual output.

[0011] As a preferred solution of the intelligent inspection method based on SLAM algorithm imaging described in the present invention, when collecting information data of the target area, multiple sensors are used to measure data.

[0012] Among them, the information data is used to analyze the robot's posture.

[0013] As a preferred solution of the intelligent inspection method based on SLAM algorithm imaging described in the present invention, the step of obtaining the robot posture includes:

[0014] Establish an initial pose benchmark through environmental perception data;

[0015] Perform motion error correction processing on sensor information data;

[0016] The robot's motion state is recursively optimized to achieve continuous estimation of posture parameters.

[0017] As a preferred solution of the intelligent inspection method based on SLAM algorithm imaging of the present invention, when constructing a three-dimensional map, the following steps are included:

[0018] LiDAR point cloud data provides the spatial structure of the scene;

[0019] Provide cable joint material characteristics through multispectral sensor data;

[0020] The wheel odometer and IMU data are then used to provide the robot's motion trajectory for map stitching and scale correction.

[0021] As a preferred solution of the intelligent inspection method based on SLAM algorithm imaging described in the present invention, wherein: when adaptive registration dynamically corrects the posture error, the steps include:

[0022] Define the robot carrier coordinate system and world coordinate system, and record the IMU posture of the initial laser frame;

[0023] By judging the thresholds of rotation error and translation error, the laser observation pose or IMU solution pose is selected for local pose optimization or global optimization.

[0024] As a preferred solution of the intelligent inspection method based on SLAM algorithm imaging of the present invention, wherein: the adaptive registration further includes:

[0025] When the rotation error exceeds the first threshold, the IMU is used to solve the posture and perform local posture optimization, and the number of optimizations is recorded;

[0026] When the translation error exceeds the second threshold, the pose is corrected by the point-to-point nearest neighbor matching method;

[0027] If the local optimization fails or the number of times exceeds the preset upper limit, the global pose optimization thread is started.

[0028] As a preferred solution of the intelligent inspection method based on SLAM algorithm imaging of the present invention, when the first threshold and the second threshold are set,

[0029] Adjust according to scene size, robot size, sensor accuracy parameters and map accuracy requirement parameters;

[0030] The first threshold and the second threshold are thresholds for performing backend global optimization.

[0031] In a second aspect, an embodiment of the present invention provides an intelligent inspection system based on SLAM algorithm imaging, which includes an environment perception sensor, a data processing module, a pose estimation module, a map construction module, and an imaging module;

[0032] The environmental perception sensor is used to collect environmental data of the target area;

[0033] The data processing module is used to process the environmental data in real time using a dynamic data window and to correct environmental data distortion during the acquisition process;

[0034] The posture estimation module is used to estimate the robot posture in real time by combining the robot's motion state and environmental observation data through a first algorithm;

[0035] The map construction module is used to dynamically correct the posture error through adaptive registration to construct a three-dimensional map;

[0036] The imaging module is used to mark the cable path and connector positions in combination with multimodal imaging technology to generate visual output.

[0037] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, any step of the above-mentioned intelligent inspection method based on SLAM algorithm imaging is implemented.

[0038] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned intelligent inspection method based on SLAM algorithm imaging is implemented.

[0039] The present invention achieves the following beneficial effects: By fusing LiDAR point cloud data with a multispectral sensor, it achieves millimeter-level spatial coordinate positioning and dual identification of material characteristics for cable joints, resolving the positioning ambiguity caused by a single viewing angle in traditional two-dimensional image recognition. Point cloud data can penetrate complex pipeline obstructions to accurately obtain the three-dimensional spatial coordinates of the joint (with an accuracy of ±3cm), while multispectral data can identify rubber and metal material characteristics (with an identification rate of >98%). This dual verification ensures target uniqueness.

[0040] An adaptive pose correction mechanism enables the system to maintain positioning accuracy in GPS-denied scenarios such as long corridors and tunnels. When the deviation between the laser observation and the IMU-solved pose exceeds a threshold (rotation error > 0.5°, translation error > 15cm), point-to-point ICP matching (registration error < 5cm) or a global optimization thread is initiated, reducing positioning error by 62% compared to traditional single-sensor SLAM.

[0041] A multi-source data sliding window mechanism achieves a data fusion processing speed of 30 frames per second, and coupled with an online pose optimization algorithm, enables simultaneous mapping and inspection. Compared to traditional manual inspections (average 2km / person / day), the robot's inspection speed is increased to 8km / h and can operate continuously 24 hours a day. The visual 3D map integrates material heat maps and spatial coordinate annotations. Operations and maintenance personnel can quickly locate potential hazards through color coding (red: aging joints, yellow: abnormal displacement), shortening decision-making response time from the traditional 48 hours to real-time alarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0043] Figure 1 This is a flow chart of the intelligent inspection method based on SLAM algorithm imaging.

[0044] Figure 2This is a flowchart of the pose calculation of the intelligent inspection method based on SLAM algorithm imaging. DETAILED DESCRIPTION

[0045] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0046] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0047] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0048] The present invention is described in detail with reference to schematic diagrams. For ease of illustration, cross-sectional views of device structures may be partially enlarged and not to scale when describing embodiments of the present invention. Furthermore, the schematic diagrams are merely illustrative and should not limit the scope of the present invention. Furthermore, in actual production, the three-dimensional dimensions of length, width, and depth should be included.

[0049] In the description of the present invention, it should be noted that the terms "upper, lower, inner, and outer" and other references to orientations or positional relationships are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first, second, or third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0050] In this disclosure, unless otherwise specified or limited, the terms "mounted," "connected," and "connected" should be interpreted broadly. For example, they may refer to fixed, removable, or integral connections. They may also refer to mechanical, electrical, or direct connections, indirect connections through an intermediary, or internal communication between two components. Those skilled in the art will understand the specific meanings of these terms in this disclosure.

[0051] Example 1

[0052] Reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides an intelligent inspection method based on SLAM algorithm imaging, comprising the following steps:

[0053] S1. Collect information data of the target area and process the information data of the target area in real time.

[0054] In an optional embodiment, when collecting information data of the target area, the data is measured by multiple sensors, including but not limited to potential sensor types such as lidar, infrared, and sonar;

[0055] For example, the distance information of the target is obtained by emitting a laser beam and measuring the time or phase difference of the reflected laser beam. The main output of the laser radar is three-dimensional point cloud data.

[0056] The infrared sensor detects the infrared radiation emitted or reflected by the object to perceive the target and generate a thermal image. Each pixel represents the temperature value of the corresponding area.

[0057] Sonar detects targets in the air by emitting sound waves and receiving echoes. It can measure the distance from the target to the sensor, usually outputting it in the form of a single point or multiple points.

[0058] In an optional embodiment, the information data of the target area includes spatial structure information and physical attribute information of the target object.

[0059] All sensor-measured information data is kept in a sliding window to ensure real-time data processing and retain historical correlation information.

[0060] In an optional embodiment, real-time processing of the target area's information data includes denoising, smoothing, and distortion correction. In particular, for visual data, image distortion caused by optical lens characteristics must be corrected to ensure accurate feature extraction and matching. Efficient real-time data processing can significantly improve the robot's ability to self-locate in complex environments. This is particularly important in applications such as cable trenches, where space is limited and obstacles abound, as it directly impacts the ability to accurately identify the location and status of cable connectors.

[0061] During wireless transmission or in harsh environments, data may be lost or corrupted. Real-time processing strategies should incorporate appropriate fault-tolerance mechanisms, such as redundant coding or predictive compensation techniques, to ensure that even if some data is unavailable, overall system performance is not significantly impacted. Efficient data processing not only improves detection efficiency but also effectively saves energy and extends equipment life. Intelligent management and analysis of data streams can avoid unnecessary computational and storage overhead.

[0062] S2. Using the first algorithm and combining it with the information data of the target area, the robot's position and posture are obtained in real time.

[0063] In an optional embodiment, the first algorithm is a core algorithm for obtaining the robot's position and posture in real time, and its main function is to fuse sensor data and environmental information to calculate the robot's position and posture in the target area;

[0064] The first algorithm can use multi-sensor fusion positioning based on the extended Kalman filter (EKF), which uses the extended Kalman filter (EKF) to fuse the data of the IMU (inertial measurement unit), lidar and vision sensor to estimate the position of the robot in the cable trench in real time;

[0065] Alternatively, a SLAM algorithm based on graph optimization can be used to construct a global map of the cable trench and simultaneously estimate the robot's pose. Graph optimization ensures the consistency of the map and pose by minimizing the error function.

[0066] Alternatively, multi-hypothesis positioning based on particle filtering can be used. The particle filtering algorithm is used to generate multiple possible robot pose hypotheses in the cable trench and gradually converge to the correct position through sensor data.

[0067] The steps to obtain the robot pose include:

[0068] Establish an initial pose benchmark through environmental perception data;

[0069] In an optional embodiment, the environmental perception data includes but is not limited to geometric structure data, visual feature data, and multi-sensor fusion scenarios;

[0070] In this embodiment, taking the multi-sensor fusion scenario as an example, the initial radar pose is set;

[0071] Assume that when a frame of point cloud starts scanning, the radar posture is:

[0072]

[0073] Where T0 is the initial radar pose transformation matrix, R0 is a 3x3 rotation matrix, representing the initial radar pose (rotation), and t0 is a 3x1 translation vector, representing the initial radar position (translation).

[0074] The coordinates of the i-th laser point are Pi = [pix piy piz]T. The radar pose during acquisition is:

[0075]

[0076] Perform motion error correction processing on sensor information data;

[0077] The motion error correction process includes motion distortion compensation, sensor calibration, etc. In an optional embodiment, the distortion is compensated and the coordinates of the laser point are obtained;

[0078] After compensating for the distortion, the coordinates of the i-th laser point are:

[0079]

[0080] Where, is the coordinate of the i-th laser point in the world coordinate system, is the inverse matrix of the initial pose transformation matrix T0, which is used to transform the point from the radar coordinate system back to the world coordinate system, T i is the pose transformation matrix at the current moment, which is used to transform the point from the world coordinate system to the current radar coordinate system, P i is the coordinate of the original laser point in the world coordinate system;

[0081] Since the number of point clouds of the laser radar is relatively small and the accuracy is low, the compensation can be performed using the mean value [V, w] of the velocity solution calculated by the IMU within the time interval of adjacent laser frames. That is, the rotation and translation compensation is:

[0082] R i =wd time , t i =Vd time ;

[0083] Where R i is the rotation matrix calculated by the angular velocity and time interval measured by the IMU, t i It is the translation vector calculated by the linear velocity and time interval measured by the IMU.

[0084] After preprocessing the laser data, the state estimation system based on tight coupling optimization begins to run, reading in the observation data after sensor sampling and processing. To reduce the amount of computation, the robot state estimation is updated through sliding window constrained data processing. Since the robot state estimation problem can be expressed as a maximum a posteriori estimation (MAP) problem, assuming that the noise is zero Gaussian noise, the maximum a posteriori probability problem is equivalent to the problem of finding the minimum sum of the cost of multi-sensor observation data. That is:

[0085]

[0086] Where {rp,Gp} is the prior information of the system state, z is the sum of the observation data of each sensor, the r(.) sub-term represents the error-constrained residual function between the actual observation data and the predicted observation data of each sensor module, and ||.||M is the Markov norm, which can be solved by the nonlinear least squares method;

[0087] The robot's motion state is recursively optimized to achieve continuous estimation of posture parameters.

[0088] In an optional embodiment, examples of the spatiotemporal correlation model include, but are not limited to, a sliding window method, factor graph optimization, Kalman filtering, etc.;

[0089] If we take the sliding window method as an example, the robot state estimation is updated by sliding window constrained data processing;

[0090] Taking factor graph optimization as an example, by modeling the problem as a graph structure, global optimization is performed using the nodes (state variables) and edges (constraint relationships) in the graph. This method is suitable for scenarios that require high-precision maps and pose estimation;

[0091] For example, the Kalman filter is a recursive estimation algorithm that estimates the system state by fusing predictions and observations. The robot uses a lidar to measure the distance between the robot and the cable trench wall as observation data. The robot's posture is updated by fusing the predictions and observations according to the Kalman gain formula.

[0092] To summarize: when acquiring each frame of point cloud, it is assumed that each laser point has its corresponding radar pose Ti at the time of acquisition, which reflects the change in radar pose due to the motion of the carrier. In order to compensate for the point cloud distortion caused by this change, each laser point needs to be compensated. The compensation process involves converting the laser point from the original radar coordinate system back to the world coordinate system, and then converting it to the new radar coordinate system based on the current radar pose. This process can be accomplished by using the angular velocity and linear velocity information provided by the IMU, which can be used to estimate the rotation and translation of the radar in the time interval between adjacent frames;

[0093] Specifically, during the compensation process, IMU data is used to calculate a rotation matrix Ri and a translation vector ti, which respectively represent the change in attitude and translation of the radar relative to the initial moment when the i-th laser point was acquired. These calculation results are then applied to each laser point to correct for distortion caused by the carrier's motion. This method relies on the high-frequency data output of the IMU and is often combined with IMU pre-integration techniques to effectively utilize IMU data for point cloud dedistortion.

[0094] The robot is able to achieve accurate self-localization in complex environments. This process not only relies on high-quality sensor data (such as lidar and IMU), but also requires effective algorithms to process this data. For example, the use of a sliding window strategy can reduce computational costs while ensuring accuracy, which is crucial for real-time operation of mobile robots in dynamic environments. In addition, the use of nonlinear least squares method to solve the MAP problem ensures that the optimal solution can be obtained even in the presence of noise. Such a system design enables the robot to maintain stable and accurate position perception in a constantly changing environment, which is crucial for autonomous navigation, obstacle avoidance, and the performance of specific tasks.

[0095] S3. Dynamically correct the posture error based on the robot's posture adaptive registration, build a three-dimensional map, and generate visual output.

[0096] When adaptive registration dynamically corrects the pose error, the steps include:

[0097] Define the robot carrier coordinate system and world coordinate system, and record the IMU posture of the initial laser frame;

[0098] In an optional embodiment, assuming that the motion between two consecutive odometry frames is in the same two-dimensional plane, the pose transformation between the two frames is the translation in the two-dimensional plane and the rotation around the normal perpendicular to the plane. That is:

[0099]

[0100] Where POB and ROB are the translation and rotation from the wheel odometer to the IMU, respectively. Because the wheel odometer rotation state only rotates around the z-axis, that is, The third element in the vector represents the angular error between the predicted rotation and the observed rotation.

[0101] Then the rotation part of the wheel odometry residual is:

[0102]

[0103] By performing angle approximation on the relevant rotation matrix, the angle error can be solved:

[0104]

[0105] By the same principle, the error of the translation part can be approximated as:

[0106]

[0107] Where r OR represents the residual error of the wheel odometry in the rotation part, represents the observation value from the current moment j to the next moment j+1, x represents the state variable, e3 is a unit direction, represents the amount of rotation error, represents the noise term.

[0108] By judging the thresholds of rotation error and translation error, the laser observation pose or IMU solution pose is selected for local pose optimization or global optimization.

[0109] Assume that the IMU observation model is:

[0110]

[0111] The IMU pre-integration residual term between the current key frame j and the previous observation frame i is expressed as follows:

[0112]

[0113]

[0114] in,

[0115]

[0116] Where, represents the measured angular velocity (rotation) at time i, represents the true angular velocity (rotation) at time i, represents the bias of the angular velocity sensor, η g represents the noise of the angular velocity sensor, represents the measured acceleration (translation) at time i, represents the rotation matrix from the world coordinate system to the robot coordinate system, represents the true acceleration (translation) at time i, X represents the pre-integrated residual vector, represents the attitude error, represents the angle error, represents the position error, δb a Denotes the acceleration bias error, δb w represents the angular velocity bias error, r p represents the position residual, rq represents the attitude residual, r v represents the velocity residual.

[0117] In the sliding window, assuming the first laser frame is at time i, the observation value at time i+1 of the next laser frame is zi+1i=(dx,dy,dθ). The coordinates of these two frames are obtained through IMU / Odom, namely xi and xi+1. The predicted pose at time i+1 is zi+1i′=Xi-1Xi+1, where Xi represents the transformation matrix corresponding to xi. The error function between the predicted value and the true value is:

[0118]

[0119] t2v means converting the transformation matrix into the corresponding pose. The error function is transformed into a matrix form:

[0120]

[0121] The corresponding Jacobian matrix is:

[0122]

[0123] That is, the error function can be transformed into:

[0124]

[0125] The error between adjacent laser frames can be solved using a nonlinear least squares method. This error constraint enables the LiDAR to quickly and effectively optimize the laser observation data when matching observations in structured scenes.

[0126] Adaptive registration and dynamic correction of pose errors are key components of the entire SLAM (Simultaneous Localization and Mapping) system. By defining the robot's carrier coordinate system and the world coordinate system and recording the IMU pose of the initial laser frame, a baseline is provided for subsequent pose calculations. In this process, wheel odometry and IMU data are used to estimate the robot's motion state, and the optimal pose update method is selected by applying thresholds to rotational and translational errors.

[0127] This adaptive approach helps improve pose estimation accuracy, especially in complex environments, such as underground cable trenches, where irregular terrain can degrade lidar data quality. When large rotational or translational errors are detected, the system can select a more reliable IMU solution for local or global pose optimization, thereby reducing error accumulation due to environmental factors.

[0128] In an optional embodiment, for adaptive inter-frame registration: Because underground cable trench mapping lacks loop constraints, the tightly coupled optimization of the laser odometry and IMU pre-integration only works well in cable trenches with flat surfaces. Therefore, the accuracy of the laser odometry is crucial in the SLAM system designed in this paper. The laser odometry is primarily derived from the registration of adjacent laser frames. The accuracy of the laser odometry is primarily affected by the following two factors:

[0129] 1) The underground cable trench is covered with rubble, sand pits and steep ups and downs. When the inspection robot conducts inspections in this environment,

[0130] LiDAR devices experience severe physical vibrations. When the robot passes through a sand pit or uphill, a large number of laser points may even hit the ground, causing a sharp increase in inter-frame error, which seriously affects odometry accuracy.

[0131] 2) The interior of the underground cable trench is mainly composed of cable supports and structures on both sides of the parallel walls where cables are placed on the cable supports. Since the laser point cloud only has the pose information of the scene structure and lacks feature description information, even if the inspection robot does not encounter serious situations such as laser points hitting the ground or ceiling during movement, the highly structured scene and repeated textures will cause mismatching between laser frames, making the map obtained by splicing based on relative pose constraints obtained by laser frame matching inevitably degraded and scaled.

[0132] Adaptive registration further includes,

[0133] When the rotation error exceeds the first threshold, the IMU is used to solve the posture and perform local posture optimization, and the number of optimizations is recorded;

[0134] When the translation error exceeds the second threshold, the pose is corrected by the point-to-point nearest neighbor matching method;

[0135] If the local optimization fails or the number of times exceeds the preset upper limit, the global pose optimization thread is started.

[0136] When the first threshold and the second threshold are set,

[0137] Adjust according to scene size, robot size, sensor accuracy parameters and map accuracy requirement parameters;

[0138] The first threshold and the second threshold are thresholds for performing backend global optimization.

[0139] The settings of the first and second thresholds can be flexibly adjusted based on the specific scene characteristics (such as size and complexity), the physical size of the robot, and the accuracy of the sensor. This means that the system can make the most optimized choice for different application scenarios, thereby improving the accuracy of positioning. When the rotation error exceeds the first threshold, the IMU is used to solve the posture for local posture optimization, which can quickly correct small-scale deviations caused by short-term motion, ensuring the continuity and accuracy of position tracking.

[0140] For larger translation errors, the pose is corrected using a point-to-point nearest neighbor matching method. This method can provide more reliable matching results in feature-rich environments and is particularly suitable for structured scenes, helping to reduce the probability of mismatches and enhance system stability. If local optimization fails or the number of attempts exceeds a preset upper limit, a global pose optimization thread is initiated. This strategy ensures that even in extreme cases (such as long-term accumulated errors), the system can find a more accurate solution, avoiding situations where it is trapped by a local optimal solution.

[0141] In an optional embodiment, in order to deal with the above two situations, the following two steps are mainly composed:

[0142] 1) Data Preprocessing: Define the robot carrier coordinate system as B, align the IMU coordinate system with the robot coordinate system, define the world coordinate system as W, and align the z-axis coordinate of the world coordinate system with the direction of gravity in the world coordinate system. Record the pose obtained by the IMU solution for the first laser frame.

[0143] 2) Adaptive registration. Through the comparison and screening of three state quantities, state quantity 1 is the predicted pose obtained by performing IMU integration on the pose of the previous frame of lidar data in step 2, state quantity 2 is the observed pose obtained by lidar inter-frame registration, and state quantity 3 is the initial pose of the robot.

[0144] In an optional embodiment, when constructing a three-dimensional map, the following steps are included:

[0145] S3.1,LiDAR point cloud data provides the scene spatial structure;

[0146] S3.2. Provide cable connector material characteristics using multispectral sensor data;

[0147] S3.3. The wheel odometer and IMU data are then used to provide the robot's motion trajectory for map stitching and scale correction.

[0148] The fusion of wheel odometry and IMU data not only helps improve the robot's self-localization capabilities but can also be used for map stitching and scale correction. This approach effectively addresses the limitations of a single sensor, such as potential slippage in wheel odometry or long-term drift in IMUs.

[0149] The point cloud data provided by LiDAR is crucial for building 3D maps. It not only provides information about the spatial structure of the scene but also enables point cloud registration technology to align data collected at different points in time to form a consistent map representation. During this process, point cloud data may be distorted due to robot motion, so compensation processing is required to ensure data consistency and accuracy.

[0150] This method significantly improves the robot's navigation capabilities and map quality in complex environments. The adaptive registration mechanism flexibly handles various error conditions, ensuring accurate and robust pose estimation. LiDAR point cloud data combined with IMU pre-integration processing improves map accuracy. Multispectral sensors enhance the map's information content. The fusion of multiple sensor data further enhances the system's overall performance. These technologies work together to construct a highly detailed and consistent 3D map, meeting the demands of practical applications.

[0151] Within the sliding window, when the sum between the two most recent laser keyframes is below the threshold, the laser alignment is in good condition and the keyframe pose is considered to be the laser observation pose. However, when the rotation error between laser keyframes exceeds the threshold, the pose information of the current laser keyframe is considered unreliable. The inertial navigation solution pose at the corresponding time is assigned to the local pose optimization, and the number of optimizations is recorded. When the rotation error is below the threshold, the translation error is thresholded. If it exceeds the threshold, the pose is re-corrected using the point-to-point nearest neighbor matching method. If the point-to-point nearest neighbor matching fails or the number of local optimizations exceeds the threshold, the global pose optimization thread is started for global pose optimization.

[0152] In summary, the fusion of LiDAR point cloud data and multispectral sensors enables millimeter-level spatial coordinate positioning and dual identification of material characteristics of cable joints, resolving the positioning ambiguity caused by a single viewing angle in traditional 2D image recognition. Point cloud data can penetrate complex pipeline obstructions to accurately obtain the 3D spatial coordinates of joints (with an accuracy of ±3cm), while multispectral data can identify rubber / metal material characteristics (with an identification rate of >98%). This dual verification ensures target uniqueness.

[0153] An adaptive pose correction mechanism enables the system to maintain positioning accuracy in GPS-denied scenarios such as long corridors and tunnels. When the deviation between the laser observation and the IMU-solved pose exceeds a threshold (rotation error > 0.5°, translation error > 15cm), point-to-point ICP matching (registration error < 5cm) or a global optimization thread is initiated, reducing positioning error by 62% compared to traditional single-sensor SLAM.

[0154] A multi-source data sliding window mechanism achieves a data fusion processing speed of 30 frames per second, and coupled with an online pose optimization algorithm, enables simultaneous mapping and inspection. Compared to traditional manual inspections (average 2km / person / day), the robot's inspection speed is increased to 8km / h and can operate continuously 24 hours a day. The visual 3D map integrates material heat maps and spatial coordinate annotations. Operations and maintenance personnel can quickly locate potential hazards through color coding (red: aging joints, yellow: abnormal displacement), shortening decision-making response time from the traditional 48 hours to real-time alarms.

[0155] Example 2

[0156] On the basis of the first embodiment, this embodiment further provides an intelligent inspection system based on SLAM algorithm imaging, including an environment perception sensor, a data processing module, a posture estimation module, a map construction module, and an imaging module;

[0157] The environmental perception sensor is used to collect environmental data of the target area;

[0158] The data processing module is used to process the environmental data in real time using a dynamic data window and to correct environmental data distortion during the acquisition process;

[0159] The posture estimation module is used to estimate the robot's posture in real time by combining the robot's motion state and environmental observation data through a multi-source data fusion optimization algorithm;

[0160] The map construction module is used to dynamically correct the posture error through adaptive registration to construct a three-dimensional map;

[0161] The imaging module is used to mark the cable path and connector positions in combination with multimodal imaging technology to generate visual output.

[0162] This embodiment also provides a computer device suitable for the intelligent inspection method based on SLAM algorithm imaging, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the intelligent inspection method based on SLAM algorithm imaging proposed in the above embodiment.

[0163] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0164] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the intelligent inspection method based on SLAM algorithm imaging proposed in the above embodiment is implemented.

[0165] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. Technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0166] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A robot intelligent inspection method based on SLAM algorithm imaging, characterized by: The following steps are included: Collect information data of the target area and process the information data of the target area in real time; The robot's position and posture are acquired in real time by combining the information data of the target area through the first algorithm; Dynamically correct the posture error based on the robot's posture adaptive registration, build a three-dimensional map, and generate visual output.

2. The robot intelligent inspection method based on SLAM algorithm imaging as claimed in claim 1, characterized in that: When collecting information data of the target area, the data is measured by a variety of sensors. Among them, the information data is used to analyze the robot's posture.

3. The robot intelligent inspection method based on SLAM algorithm imaging as claimed in claim 2, characterized in that: The steps to obtain the robot pose include: Establish an initial pose benchmark through environmental perception data; Perform motion error correction processing on sensor information data; The robot's motion state is recursively optimized to achieve continuous estimation of posture parameters.

4. The robot intelligent inspection method based on SLAM algorithm imaging as claimed in claim 3, characterized in that: When building a 3D map, the following steps are included: LiDAR point cloud data provides the spatial structure of the scene; Provide cable joint material characteristics through multispectral sensor data; The wheel odometer and IMU data are then used to provide the robot's motion trajectory for map stitching and scale correction.

5. The robot intelligent inspection method based on SLAM algorithm imaging as claimed in claim 4, characterized in that: When adaptive registration dynamically corrects the pose error, the steps include: Define the robot carrier coordinate system and world coordinate system, and record the IMU posture of the initial laser frame; By judging the thresholds of rotation error and translation error, the laser observation pose or IMU solution pose is selected for local pose optimization or global optimization.

6. The robot intelligent inspection method based on SLAM algorithm imaging as claimed in claim 5, characterized in that: The adaptive registration further comprises, When the rotation error exceeds the first threshold, the IMU is used to solve the posture and perform local posture optimization, and the number of optimizations is recorded; When the translation error exceeds the second threshold, the pose is corrected by the point-to-point nearest neighbor matching method; If the local optimization fails or the number of times exceeds the preset upper limit, the global pose optimization thread is started.

7. The robot intelligent inspection method based on SLAM algorithm imaging as claimed in claim 6, characterized in that: When the first threshold and the second threshold are set, Adjust according to scene size, robot size, sensor accuracy parameters and map accuracy requirement parameters; The first threshold and the second threshold are thresholds for performing backend global optimization.

8. A robot intelligent inspection system based on SLAM algorithm imaging, based on the robot intelligent inspection method based on SLAM algorithm imaging according to any one of claims 1 to 7, characterized in that: Includes environment perception sensor, data processing module, pose estimation module, map construction module, and imaging module; The environmental perception sensor is used to collect environmental data of the target area; The data processing module is used to process the environmental data in real time using a dynamic data window and to correct environmental data distortion during the acquisition process; The posture estimation module is used to estimate the robot posture in real time by combining the robot's motion state and environmental observation data through a first algorithm; The map construction module is used to dynamically correct the posture error through adaptive registration to construct a three-dimensional map; The imaging module is used to mark the cable path and connector positions in combination with multimodal imaging technology to generate visual output.

9. A computer device comprising a memory and a processor, wherein the memory and an environment sensing sensor store a computer program, wherein: When the processor executes the computer program, the steps of the robot intelligent inspection method based on SLAM algorithm imaging according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the robot intelligent inspection method based on SLAM algorithm imaging according to any one of claims 1 to 7 are implemented.

Images