CN-121977531-A - Underground space detection and positioning method combining laser information and gas concentration information

CN121977531ACN 121977531 ACN121977531 ACN 121977531ACN-121977531-A

Abstract

An underground space detection and positioning method combining laser information and gas concentration information belongs to the technical field of synchronous positioning and mapping. The invention aims at the problems that the accumulated error of the existing SLAM system in the underground closed space is obvious and the consistency of the map is difficult to guarantee. The method comprises the steps of providing initial value and direction constraint of physical field guidance for laser interframe matching by estimating gas concentration spatial gradient, improving tracking robustness under characteristic degradation scene, constructing and jointly optimizing a geometric map and a parameterized gas concentration field model in a local map construction stage, restraining local drift by utilizing concentration consistency constraint anchor pose, generating a bimodal descriptor fusing geometric and gas distribution characteristics in a loop detection link, realizing scene identification with high discriminant, and effectively distinguishing geometric similarity areas. The invention can be used for reliable navigation and mapping without satellite signals.

Inventors

CHEN LIANG
CHEN YUANBO
BAI CHENGCHAO
ZHENG HONGXING
YAN PENG
GUO JIFENG

Assignees

哈尔滨工业大学
西安现代控制技术研究所

Dates

Publication Date: 20260505
Application Date: 20260212

Claims (10)

1. A method for detecting and positioning underground space by combining laser information and gas concentration information is characterized by comprising the following steps, Space-time synchronous calibration is carried out on a laser radar and a gas sensor which are arranged on the robot, and 3D point cloud data of the underground space are acquired by adopting the laser radar at the moment t And processing to obtain a t-moment characteristic point set Acquiring an observed value of the gas concentration at the moment t by adopting a gas sensor Simultaneously using inertial sensor to obtain initial pose at t moment , In the following Is a three-dimensional coordinate of the initial position, Is a gesture comprising pitch angle, roll angle and yaw angle; employing sliding window selection includes And I=1, 2,3, the term "t-1, The three-dimensional gradient vector observation value of the gas concentration at the moment t is obtained by adopting least square fitting ; Pose transfer matrix based on t moment Three-dimensional gradient vector observation value of gas concentration at time t The method comprises the steps of calculating a predicted value of gas concentration change at the moment t relative to the moment t-1, establishing a gas gradient error item from an observed value of gas concentration change at the moment t relative to the moment t-1 and the predicted value of gas concentration change, carrying out weighted fusion on a traditional geometric error item and the gas gradient error item, constructing an interframe matching pose transfer matrix optimization function by taking the minimum weighted fusion result as a target, and solving the pose transfer matrix at the moment t by adopting a Gauss Newton method ; Pose transfer matrix based on t moment Initial pose is taken Performing geometric feature matching with the map obtained at the previous t-1 moment to calculate geometric feature matching errors, establishing an objective function by taking the weighted sum of the geometric feature matching errors and gas concentration errors as the minimum target, and solving to obtain the initial pose at the t moment T-moment pose correction value of (2) ; Pose correction value based on t moment Feature point set at t moment Carrying out feature point matching and point cloud fusion on the map obtained at the previous t-1 moment to obtain a local map at the previous t moment; Selecting 3D point cloud data with geometrical remarkable characteristics from historical detection data as key frames, establishing a key frame set, taking the 3D point cloud data at the moment t as a current frame, performing loop detection based on the key frames and the current frame, and performing global consistency optimization on a local map at the moment t after loop confirmation to obtain an optimized map.
2. The method for detecting and locating an underground space by combining laser information and gas concentration information according to claim 1, wherein a three-dimensional gradient vector observation value of the gas concentration at time t is obtained The method of (1) is as follows: constructing a linear equation set: , In the middle of Is that Is the initial position difference of the ith behavior adjacent moment , Is that Concentration difference matrix of (i) the i-th behavior gas concentration observation difference ; Solving to obtain a three-dimensional gradient vector observation value of the gas concentration at the moment t : 。
3. The method for detecting and locating a subterranean space utilizing laser light and gas concentration information according to claim 2, wherein, The predicted value of the gas concentration change at the time t relative to the time t-1 is as follows: , pose transfer matrix In the form of a4 x 4 dimensional homogeneous transformation matrix, To be used in Initial pose of dimension Conversion to A conversion function in the form of a dimensional matrix, To be used in Dimension(s) Conversion to Is a conversion function of (2); The translation part is constituted as a translation vector, Conversion to a rotation matrix by the rondrigues transform Is combined into the standard form of homogeneous transformation matrix Dimension matrix ; The observed value of the gas concentration change at the corresponding time t relative to the time t-1 is Establishing a gas gradient error term The method comprises the following steps: 。
4. The method for detecting and locating a subterranean space utilizing laser light and gas concentration information according to claim 3, The pose transfer matrix optimization function is constructed as follows: , In the middle of As a matter of conventional geometric error terms, Is the gas gradient error weight; , Where N is the number of 3D point cloud matches, The nth 3D point cloud matching position is the t moment; solving a pose transfer matrix optimization function to obtain a pose transfer matrix at the moment t 。
5. The method for detecting and locating a subterranean space combining laser information and gas concentration information according to claim 4, wherein the gas concentration error is an observed value of gas concentration at time t The difference value between the predicted value and the predicted value of the gas concentration at the time t is expressed as the predicted value of the gas concentration at the time t : , In the middle of Is a positional three-dimensional coordinate correction value.
6. The method for detecting and locating a subterranean space combining laser information and gas concentration information according to claim 5, wherein the objective function is established with the minimum of the weighted sum of geometric feature matching errors and gas concentration errors as the objective: , Wherein K is the geometric feature matching number, The k geometrical feature matching error is the k 3D point cloud matching position at the t moment Distance from the matching location in the local map; The weights are matched for the geometric features, Is the gas concentration error weight; solving the objective function by adopting an L-M nonlinear least square algorithm, and obtaining an optimal initial pose at the t moment by iterative solution As a position correction value at time t 。
7. The method of claim 6, wherein performing loop detection includes calculating geometric similarity between feature point sets of key frames and current frames Gas similarity of gas concentration observations corresponding to key and current frames 。
8. The method for detecting and locating a subsurface space by combining laser information and gas concentration information according to claim 7, wherein a sequence of neighboring gas concentration observations corresponding to a key frame is expressed as a feature vector Representing a neighbor gas concentration observation value sequence corresponding to the current frame as a feature vector Calculating gas similarity by cosine similarity : 。
9. The method for detecting and locating a subterranean space utilizing laser light and gas concentration information according to claim 8, wherein, Performing loop-back detection also includes calculating a loop-back candidate score : , In the middle of Is the weight of geometric similarity; If loop candidate score And if the key frame is larger than the set loop threshold, continuing geometric verification, ICP matching error verification and overlapping degree verification of the key frame and the current frame, and confirming the loop after passing all the verification.
10. The method for detecting and locating a subsurface space by combining laser information and gas concentration information according to claim 9, wherein the method for global consistency optimization of the local map at time t comprises a nonlinear least squares method or a graph optimization method.

Description

Underground space detection and positioning method combining laser information and gas concentration information Technical Field The invention relates to an underground space detection and positioning method combining laser information and gas concentration information, belonging to the technical field of synchronous positioning and mapping. Background Along with the development of underground engineering and emergency rescue, the requirements of high-precision positioning and mapping of an underground enclosed space are urgent. In such an environment, satellite signals are completely invalid, the structure is repeated, illumination is insufficient, accumulated errors of a SLAM system depending on laser or vision are obvious, and map consistency is difficult to guarantee. The existing method is improved by fusing inertial or geomagnetic data, but geomagnetism is easy to interfere, the inertial has drift, and the long-term stability is insufficient. Disclosure of Invention Aiming at the problems that the accumulated error of the existing SLAM system in the underground closed space is obvious and the consistency of the map is difficult to guarantee, the invention provides an underground space detection and positioning method combining laser information and gas concentration information. The invention relates to a method for detecting and positioning underground space by combining laser information and gas concentration information, which comprises the following steps, Space-time synchronous calibration is carried out on a laser radar and a gas sensor which are arranged on the robot, and 3D point cloud data of the underground space are acquired by adopting the laser radar at the moment tAnd processing to obtain a t-moment characteristic point setAcquiring an observed value of the gas concentration at the moment t by adopting a gas sensorSimultaneously using inertial sensor to obtain initial pose at t moment,In the followingIs a three-dimensional coordinate of the initial position,Is a gesture comprising pitch angle, roll angle and yaw angle; employing sliding window selection includes AndI=1, 2,3, the term "t-1,The three-dimensional gradient vector observation value of the gas concentration at the moment t is obtained by adopting least square fitting; Pose transfer matrix based on t momentThree-dimensional gradient vector observation value of gas concentration at time tThe method comprises the steps of calculating a predicted value of gas concentration change at the moment t relative to the moment t-1, establishing a gas gradient error item from an observed value of gas concentration change at the moment t relative to the moment t-1 and the predicted value of gas concentration change, carrying out weighted fusion on a traditional geometric error item and the gas gradient error item, constructing an interframe matching pose transfer matrix optimization function by taking the minimum weighted fusion result as a target, and solving the pose transfer matrix at the moment t by adopting a Gauss Newton method; Pose transfer matrix based on t momentInitial pose is takenPerforming geometric feature matching with the map obtained at the previous t-1 moment to calculate geometric feature matching errors, establishing an objective function by taking the weighted sum of the geometric feature matching errors and gas concentration errors as the minimum target, and solving to obtain the initial pose at the t momentT-moment pose correction value of (2); Pose correction value based on t momentFeature point set at t momentCarrying out feature point matching and point cloud fusion on the map obtained at the previous t-1 moment to obtain a local map at the previous t moment; Selecting 3D point cloud data with geometrical remarkable characteristics from historical detection data as key frames, establishing a key frame set, taking the 3D point cloud data at the moment t as a current frame, performing loop detection based on the key frames and the current frame, and performing global consistency optimization on a local map at the moment t after loop confirmation to obtain an optimized map. According to the underground space detection and positioning method combining the laser information and the gas concentration information, the gas concentration three-dimensional gradient vector observation value at the time t is obtainedThe method of (1) is as follows: constructing a linear equation set: , In the middle of Is thatIs the initial position difference of the ith behavior adjacent moment,Is thatConcentration difference matrix of (i) the i-th behavior gas concentration observation difference; Solving to obtain a three-dimensional gradient vector observation value of the gas concentration at the moment t: 。 According to the underground space detection and positioning method combining laser information and gas concentration information, the predicted value of the gas concentration change at the moment t relative to the moment t-1 is as fo