CN-121979231-A - Inspection method for unmanned plane tunnel and drainage culvert

CN121979231ACN 121979231 ACN121979231 ACN 121979231ACN-121979231-A

Abstract

The invention discloses an inspection method of an unmanned aerial vehicle roadway and a drainage culvert, which comprises the steps of constructing an environment priori model based on structural information of the roadway or the drainage culvert, controlling the unmanned aerial vehicle to autonomously explore in an environment without a global navigation satellite system signal, constructing an environment map in real time and carrying out registration correction with real geographic data to generate a dynamic reference navigation map, inspecting according to the dynamic reference navigation map, identifying a key structure, carrying out semantic annotation on the dynamic reference navigation map, constructing a repositioning anchor point library according to the identified key structure, repositioning when the unmanned aerial vehicle fails to be positioned, predicting the residual available electric quantity of the unmanned aerial vehicle in real time, and planning a return route according to a prediction result by combining the dynamic reference navigation map to ensure the safe recycling of the unmanned aerial vehicle. The invention realizes the full-autonomous intelligent inspection with high efficiency, high precision, high robustness and high safety of the unmanned aerial vehicle in GNSS refused environments such as roadways, drainage culverts and the like.

Inventors

NIE WEN
ZHAO WENBIN
DAI BIBO
HOU JUN
Xin Zekun
WANG YUNMIN
WANG XING

Assignees

中钢集团马鞍山矿山研究总院股份有限公司
长春黄金研究院有限公司

Dates

Publication Date: 20260505
Application Date: 20251229

Claims (10)

1. The inspection method for the unmanned aerial vehicle roadway and the drainage culvert is characterized by comprising the following steps of: Constructing an environment priori model based on structural information of a roadway or a drainage culvert, and controlling the unmanned aerial vehicle to explore autonomously in an environment without a global navigation satellite system signal according to the environment priori model; In the autonomous exploration process, an environment map is built in real time and registered and corrected with real geographic data, so that a dynamic reference navigation map is generated; Performing inspection according to the dynamic reference navigation map, identifying a key structure, performing semantic annotation on the dynamic reference navigation map, constructing a repositioning anchor point library according to the identified key structure, and repositioning when the unmanned aerial vehicle fails to be positioned; And predicting the residual available electric quantity of the unmanned aerial vehicle in real time, and planning a return route according to a prediction result and combining with a dynamic reference navigation map to ensure the safe recovery of the unmanned aerial vehicle.
2. The method of claim 1, wherein constructing an environment prior model based on structural information of a roadway or a drainage culvert, controlling unmanned aerial vehicle autonomous exploration according to the environment prior model, specifically comprises: Based on a building information model or a computer aided design drawing, constructing a light three-dimensional nerve radiation field as an environment priori model; identifying the front edge points of unexplored areas according to the three-dimensional occupied grid map output by the real-time synchronous positioning and mapping system; and comprehensively evaluating information gain, topological consistency with the environment priori model and estimated energy consumption cost reaching the leading edge point aiming at each leading edge point, selecting an optimal exploration direction and controlling the unmanned aerial vehicle to conduct autonomous exploration.
3. The method according to claim 2, wherein the lightweight three-dimensional neural radiation field is constructed by a hash grid coding and small multi-layer perceptron network, and is subjected to pruning and quantization processing to adapt the computational power and storage conditions of an unmanned aerial vehicle on-board edge computing platform.
4. The method according to claim 1, wherein the environment map is constructed in real time and registered and corrected with the real geographic data in an autonomous exploration process, specifically comprising: Generating a six-degree-of-freedom pose and dense point cloud map in real time by fusing a tight coupling synchronous positioning and mapping algorithm of laser radar, a depth camera and inertial measurement unit data; When the preset triggering condition is met, carrying out spatial registration on the real-time point cloud map and real geographic data acquired through a ground laser scanning or geographic information system through an iterative nearest point algorithm, and solving an optimal rigid body transformation matrix; And performing global optimization on the historical track of the unmanned aerial vehicle and the constructed map by using the optimal rigid body transformation matrix, eliminating accumulated errors and generating the dynamic reference navigation map.
5. The method of claim 4, wherein the predetermined triggering condition comprises the cumulative flight distance of the unmanned aerial vehicle reaching a first predetermined threshold, or the trace value of the pose covariance matrix in the back-end optimization process of the synchronous positioning and mapping system exceeding a second predetermined threshold.
6. The method according to claim 1, characterized in that the identification of key structures and the semantic annotation on a dynamic reference navigation map comprises in particular: Processing data acquired in real time through a lightweight three-dimensional target detection network deployed in an unmanned aerial vehicle-mounted edge computing unit, and identifying a key structure in real time, wherein the key structure at least comprises a branch well, a water collecting well and a secondary culvert structure; and overlapping the identification result to the dynamic reference navigation map in a graphical labeling mode to form a semantic enhancement layer.
7. The method according to claim 1, wherein constructing a relocation anchor point library based on the identified critical structure, and performing relocation when the unmanned aerial vehicle positioning fails, specifically comprises: Extracting geometric feature descriptors of the identified key structures and global poses thereof, and constructing a repositioning anchor point library; When the positioning confidence of the unmanned aerial vehicle is lower than a preset threshold, extracting a characteristic descriptor of current observation data and matching the characteristic descriptor with a candidate descriptor in the anchor point library; and if the matching similarity exceeds a set threshold and the geometric relation verification is passed, resetting the pose of the unmanned aerial vehicle to be the global pose corresponding to the anchor point successfully matched.
8. The method according to claim 1, wherein the remaining available power of the unmanned aerial vehicle is predicted in real time, and the return route is planned according to the prediction result, specifically comprising: Estimating the residual available flight time or electric quantity through a multivariable prediction model based on the state of charge and the state of health of the battery, the current flight state, the environmental airflow resistance, the task load power consumption and the estimated energy consumption of the residual path; setting a dual-electric-quantity threshold return mechanism, triggering early warning and pre-planning a return path when the electric quantity is lower than a first threshold, and triggering emergency return when the electric quantity is lower than a second threshold; and planning an optimal safe return route avoiding a dangerous area according to the dynamic reference navigation map.
9. The method of claim 8, wherein the estimated energy consumption of the remaining path is obtained by performing an integral calculation based on a path length, an estimated flight speed, and a terrain slope angle of each point on the path by an energy consumption coefficient reflecting a dynamics characteristic of the unmanned aerial vehicle.
10. The method of claim 1, wherein the dynamic reference navigation map adopts a hierarchical storage structure including a base geometry layer, a semantic annotation layer, an obstacle layer, a secure channel layer, and a historical track layer.

Description

Inspection method for unmanned plane tunnel and drainage culvert Technical Field The invention belongs to the technical field of mine safety inspection, and particularly relates to an inspection method for an unmanned aerial vehicle roadway and a drainage culvert. Background The underground tunnel and the drainage culvert are used as key infrastructure for energy exploitation, and the structural safety of the underground tunnel and the drainage culvert directly relates to public safety and production efficiency. Currently, manual inspection is mainly relied on, operators need to go deep into severe environments with toxic gas, collapse risks or sudden ponding, and personal safety threats are prominent. The traditional detection mode has low efficiency, the hiking investigation is long in time consumption, the periodic investigation requirement of a large-scale pipe network system is difficult to deal with, and the response is delayed especially under the emergency working conditions after heavy rain and disaster. In addition, hidden damages such as crack development, pipe wall corrosion and the like can not be accurately captured by naked eye observation, the omission ratio of high-altitude and underwater areas is high, and great potential safety hazards are easily left. To improve inspection safety, orbital robots and track detection devices have been introduced into use. However, such devices rely on pre-paved tracks or flat road surfaces with severely limited traffic capacity in slump blocking or slush road segments. The carried visible light camera is easy to lose efficacy due to lens fog in the culvert high-humidity environment, and the laser scanning device is difficult to adapt to a complex curve structure. More prominently, the coverage area of a single device is limited, the long-distance pipe network needs to be densely distributed, the deployment and maintenance cost is far higher than that of the traditional manual mode, and the popularization value is restricted due to the lack of economy. The successful application of the consumer unmanned aerial vehicle in the open area excites the exploration of the consumer unmanned aerial vehicle in the inspection of the closed space. However, the special environment of the tunnel culvert causes multiple obstacles to the conventional unmanned aerial vehicle, namely the GPS signal is lack to cause serious drift of positioning, errors of an inertial navigation system are continuously accumulated, visual recognition capability is obviously reduced due to dark and water mist interference, flight instability is aggravated due to air flow disturbance in a narrow space, and the conventional unmanned aerial vehicle is difficult to support for long-distance continuous operation in a medium-distance manner, so that the practicability is fundamentally limited. The defects of the prior art are overcome, a set of special unmanned aerial vehicle roadway and drainage culvert inspection system integrating autonomous positioning, high-precision perception and intelligent management of cruising is developed, and the system becomes an urgent need for breaking through the safety monitoring bottleneck of the underground limited space. Disclosure of Invention Aiming at the technical problems that the existing unmanned aerial vehicle has strong autonomous exploration blindness, large map construction accumulated error, weak key structure identification capability, easy failure in positioning, insufficient return safety and the like in underground environments without Global Navigation Satellite System (GNSS) signals, repeated structures, poor illumination conditions and limited space, such as a roadway, a drainage culvert and the like, the invention provides a fully autonomous intelligent inspection method for the unmanned aerial vehicle roadway and the drainage culvert, so as to realize high efficiency, high precision, high robustness and high safety. The invention provides a patrol method for an unmanned aerial vehicle roadway and a drainage culvert, which comprises the following steps: Constructing an environment priori model based on structural information of a roadway or a drainage culvert, and controlling the unmanned aerial vehicle to explore autonomously in an environment without a global navigation satellite system signal according to the environment priori model; In the autonomous exploration process, an environment map is built in real time and registered and corrected with real geographic data, so that a dynamic reference navigation map is generated; Performing inspection according to the dynamic reference navigation map, identifying a key structure, performing semantic annotation on the dynamic reference navigation map, constructing a repositioning anchor point library according to the identified key structure, and repositioning when the unmanned aerial vehicle fails to be positioned; And predicting the residual available electric quantity of the unmanned aerial veh