CN-121994240-A - Unmanned aerial vehicle multifunctional automatic inspection method and system with real-time and low energy consumption

CN121994240ACN 121994240 ACN121994240 ACN 121994240ACN-121994240-A

Abstract

The invention relates to a real-time low-energy-consumption unmanned aerial vehicle multifunctional automatic inspection method and system, and belongs to the field of automatic inspection. The method comprises the steps of carrying out semantic field conversion based on a fuzzy geographic instruction, outputting a geographic semantic field, carrying out visual-field coupling navigation based on the geographic semantic field, outputting real-time flight track and environment preliminary perception data, carrying out topological risk recognition based on the real-time flight track and the environment preliminary perception data to obtain a 3D environment model with risk labeling, carrying out fractal energy consumption optimization based on the 3D environment model with risk labeling, outputting an optimized waypoint sequence, and carrying out multi-mode data acquisition and real-time risk verification based on the optimized waypoint sequence. The intelligent patrol method and the intelligent patrol system realize intelligent patrol for understanding natural language instructions, fusing multi-sensor information, dynamically evaluating risks and optimizing energy consumption.

Inventors

PAN HONGTAO
Liu Puyang
LIU HUI
KANG JIAWEI
ZHANG HAO
Yao Guiyue
WANG NING
SHI PAN
TANG XIAOQIANG
ZHOU GUOLIANG
LI JIAN
WANG HONGXU
ZHENG YI
XU ZHENG
LIU MIN
WU JIA
HUANG XIAOLONG
Qi Yunuo

Assignees

国网冀北电力有限公司技能培训中心
保定电力职业技术学院
国家电网有限公司

Dates

Publication Date: 20260508
Application Date: 20260204

Claims (7)

1. The unmanned aerial vehicle multifunctional automatic inspection method with real-time and low energy consumption is characterized by comprising the following steps of: Step S1, carrying out semantic field conversion based on a fuzzy geographic instruction, and outputting a geographic semantic field; Step S2, performing visual-field coupling navigation based on the geographic semantic field, and outputting real-time flight track and environment preliminary perception data; Step S3, performing topology risk identification based on the real-time flight trajectory and the environment preliminary perception data to obtain a 3D environment model with risk labeling; S4, carrying out fractal energy consumption optimization based on the 3D environment model with the risk annotation, and outputting an optimized waypoint sequence; and S5, carrying out multi-mode data acquisition and real-time risk verification based on the optimized waypoint sequence.
2. The method for multi-functional automated inspection of unmanned aerial vehicle with real-time and low energy consumption according to claim 1, wherein the semantic field conversion in step S1 is specifically: Inputting the fuzzy geographic instruction, carrying out natural language analysis on the fuzzy geographic instruction, and extracting key geographic entities and spatial relations; Pre-loading a digital map database, and identifying the position of the key geographic entity in the digital map and marking the position as a geographic reference point; constructing a geographical semantic field potential function to obtain geographical semantic field potential values of each position, wherein the geographical semantic field potential values are mathematically described as , wherein, Is the position The geographical semantic field potential value at which n is the number of geographical reference points, For semantic weights of the ith geographic reference point, Is the spatial coordinates of the ith geographic reference point, The semantic impact radius for the ith geographic reference point, Is taken as a point The projected distance to the geographic reference point in the direction specified by the ambiguous geographic command, Constraining the maximum distance for the direction; the geosemantic field is formed based on the geosemantic field potential values.
3. The method for multi-functional automated inspection of unmanned aerial vehicle with real-time and low energy consumption according to claim 1, wherein the visual-field coupling navigation in step S2 is specifically: According to the geographical semantic field flying to the position with the highest geographical semantic field potential value, continuously obtaining environment preliminary perception data in the process, and obtaining the geographical semantic field potential value of the current position in the geographical semantic field in real time; constructing a vision-field coupling navigation equation to obtain a flying speed vector, wherein the mathematical description is as follows , wherein, In the form of a vector of the speed of flight, The coefficients are guided for the field gradient, As a function of the speed scale factor, For a physical semantic field gradient, For the current position The geographical semantic field potential value at which, For the visual obstacle avoidance weight, the visual obstacle avoidance weight is calculated, In order to avoid the barrier scale factors, m is the number of barriers in the field of view, Is the obstacle avoidance weight of the jth obstacle, For a predicted speed vector to avoid the jth obstacle, In order to prevent a very small constant of zero removal, As the distance attenuation coefficient, For an estimated distance to the jth obstacle; And obtaining the real-time flight trajectory and the environment preliminary perception data based on the flight speed vector.
4. The method for multi-functional automated inspection of a unmanned aerial vehicle with low energy consumption in real time according to claim 1, wherein the topology risk identification in step S3 is specifically: reaching the vicinity of the maximum geographic semantic field potential value area, continuously collecting high-definition images, and reconstructing sparse three-dimensional point clouds of the environment; Meanwhile, analyzing the image, and identifying key elements to obtain a semantic identification result; calculating a risk value based on the 3D environmental model and the semantic recognition result, mathematically described as , wherein, Time t, position At risk value, P is the number of potential risk points detected, For the initial intensity of the risk point p, In order to be a risk attenuation rate, Is that And risk points Is used for the distance of (a), As a function of risk severity, dependent on risk category , For the risk time decay factor, t is the current time, The time at which the risk point p was first identified; And constructing a topological risk field based on the risk value, and combining the reconstructed 3D environment model to obtain the 3D environment model with the risk annotation.
5. The method for multi-functional automated inspection of unmanned aerial vehicle with real-time and low energy consumption according to claim 4, wherein the fractal energy consumption optimization in step S4 is specifically: Presetting candidate waypoint sets in a task area, obtaining fractal complexity C of paths between two adjacent candidate waypoints for each candidate waypoint set, and obtaining total energy consumption based on the fractal complexity, wherein the total energy consumption is mathematically described as , wherein, As a result of the total energy consumption, N is the number of candidate waypoints as an energy density factor, As a coefficient of the complexity of the energy consumption, For two candidate waypoints And The fractal complexity of the inter-path, As the distance between the candidate waypoints, Is taken as a point The air density at the location(s) is (are) high, Is the standard air density; calculating the amount of non-collected information based on the risk value, mathematically described as , wherein, For the amount of information not to be collected, Time t, position A risk value at the location of the risk value, In order to effectively cover the radius of the radius, Is a task area; obtaining an optimized navigation point sequence based on the total energy consumption and the non-collected information quantity, and mathematically describing as , wherein, In order to optimize the sequence of waypoints, As a set of candidate waypoints, Is the information-energy consumption balance coefficient.
6. The method for multi-functional automated inspection of a unmanned aerial vehicle with low energy consumption according to claim 5, wherein the real-time risk verification in step S5 is specifically: Flying according to the optimized waypoint sequence, and simultaneously activating a sensor to acquire multi-mode data; obtaining updated risk values based on the multi-modal data, mathematically described as , wherein, In order to update the risk value after the update, In order to reduce the risk of a risk reduction index, For the degree of anomaly measurement after normalization, For the normalized abnormality determination threshold value, As the maximum possible outlier value to be used, For the new risk enhancement coefficient, Confidence for newly detected risk; And repartitioning the risk areas according to the updated risk values and planning additional patrol tasks.
7. The unmanned aerial vehicle multifunctional automatic inspection system with real-time low energy consumption is characterized by being applied to the unmanned aerial vehicle multifunctional automatic inspection method with real-time low energy consumption according to any one of claims 1-6, and comprises a semantic conversion module, a navigation module, a risk marking module, an energy consumption optimization module and a risk verification module; The semantic conversion module is used for carrying out semantic field conversion based on the fuzzy geographic instruction and outputting a geographic semantic field; the navigation module is used for performing vision-field coupling navigation based on the geographic semantic field and outputting real-time flight track and environment preliminary perception data; The risk labeling module is used for carrying out topological risk recognition based on the real-time flight track and the environment preliminary perception data to obtain a 3D environment model with risk labeling; The energy consumption optimization module is used for carrying out fractal energy consumption optimization based on the 3D environment model with the risk annotation and outputting an optimized waypoint sequence; The risk verification module is used for carrying out multi-mode data acquisition and real-time risk verification based on the optimized waypoint sequence.

Description

Unmanned aerial vehicle multifunctional automatic inspection method and system with real-time and low energy consumption Technical Field The invention belongs to the technical field of automatic inspection, and particularly relates to a real-time low-energy-consumption unmanned aerial vehicle multifunctional automatic inspection method and system. Background At present, unmanned aerial vehicles gradually replace traditional manual inspection in inspection of infrastructure such as power lines, bridges and pipelines, and become an important means for improving operation safety and efficiency. The existing inspection technology mainly comprises three types, namely an automatic inspection based on preset waypoints, which depends on accurate geographic coordinates and pre-coding Cheng Hangxian, has poor flexibility and cannot cope with sudden task or area change, an autonomous flight based on GPS and inertial navigation, which has certain adaptability, but easily generates positioning drift in complex terrain or signal shielding environment and cannot understand semantic instructions, and an inspection system combined with computer vision, which mostly only has obstacle avoidance or target recognition functions and lacks understanding and dynamic risk modeling capabilities of fuzzy task instructions. In addition, the existing method often neglects energy consumption optimization and risk-driven self-adaptive data acquisition in the inspection process, so that the unmanned aerial vehicle is difficult to efficiently cover a high-risk area or repeatedly acquire low-value information under the condition of limited electric quantity. Disclosure of Invention In order to solve the problems in the prior art, the invention provides a real-time low-energy-consumption unmanned aerial vehicle multifunctional automatic inspection method and system. The aim of the invention can be achieved by the following technical scheme: The implementation of the unmanned aerial vehicle multifunctional automatic inspection method with real-time low energy consumption comprises the following steps: Step S1, carrying out semantic field conversion based on a fuzzy geographic instruction, and outputting a geographic semantic field; Step S2, performing visual-field coupling navigation based on the geographic semantic field, and outputting real-time flight track and environment preliminary perception data; Step S3, performing topology risk identification based on the real-time flight trajectory and the environment preliminary perception data to obtain a 3D environment model with risk labeling; S4, carrying out fractal energy consumption optimization based on the 3D environment model with the risk annotation, and outputting an optimized waypoint sequence; and S5, carrying out multi-mode data acquisition and real-time risk verification based on the optimized waypoint sequence. Preferably, the semantic field conversion in the step S1 is specifically: Inputting the fuzzy geographic instruction, carrying out natural language analysis on the fuzzy geographic instruction, and extracting key geographic entities and spatial relations; Pre-loading a digital map database, and identifying the position of the key geographic entity in the digital map and marking the position as a geographic reference point; constructing a geographical semantic field potential function to obtain geographical semantic field potential values of each position, wherein the geographical semantic field potential values are mathematically described as , wherein,Is the positionThe geographical semantic field potential value at which n is the number of geographical reference points,For semantic weights of the ith geographic reference point,Is the spatial coordinates of the ith geographic reference point,The semantic impact radius for the ith geographic reference point,Is taken as a pointThe projected distance to the geographic reference point in the direction specified by the ambiguous geographic command,Constraining the maximum distance for the direction; the geosemantic field is formed based on the geosemantic field potential values. Preferably, the visual-field coupling navigation in the step S2 specifically includes: According to the geographical semantic field flying to the position with the highest geographical semantic field potential value, continuously obtaining environment preliminary perception data in the process, and obtaining the geographical semantic field potential value of the current position in the geographical semantic field in real time; constructing a vision-field coupling navigation equation to obtain a flying speed vector, wherein the mathematical description is as follows , wherein,In the form of a vector of the speed of flight,The coefficients are guided for the field gradient,As a function of the speed scale factor,For a physical semantic field gradient,For the current positionThe geographical semantic field potential value at which,For the visual obstacle a