CN-121982538-A - Quantitative acceptance method for power grid trace ecological restoration based on spectral fingerprint library

CN121982538ACN 121982538 ACN121982538 ACN 121982538ACN-121982538-A

Abstract

The application discloses a grid locus ecological restoration quantitative acceptance method based on a spectrum fingerprint library, which comprises the steps of S1, constructing a full life cycle spectrum fingerprint library of a power grid typical habitat, S2, establishing a full life cycle digital twin file of a construction locus, S3, carrying out restoration quality quantitative calculation based on multi-feature fusion and intelligent identification, carrying out pseudo-greening intelligent screening on an acceptance period image based on a pseudo-greening sample feature training classifier in the spectrum fingerprint library, S4, introducing time sequence spectrum dynamics to carry out sustainability early warning and acceptance decision, and S5, outputting results. The intelligent classifier is trained based on the characteristics of the pseudo greening sample in the spectral fingerprint library, real vegetation and artificial camouflage materials can be effectively distinguished from the dimensions of spectral morphology, derivative characteristics and the like, a risk distribution map is automatically output, and a reliable false-making technical means is provided for acceptance.

Inventors

SHI JIANBO
ZHANG YING
ZHANG CHI
LIU PING
WANG CHENG

Assignees

国网湖北省电力有限公司电力科学研究院

Dates

Publication Date: 20260505
Application Date: 20260127

Claims (10)

1. The quantitative acceptance method for the ecological restoration of the power grid locus based on the spectral fingerprint library is characterized by comprising the following steps of: S1, constructing a full life cycle spectrum fingerprint library of a power grid typical habitat, wherein aiming at the type of the power grid typical habitat along the line, hyperspectral data observed at fixed points of a space-based satellite, an air-based unmanned aerial vehicle and a foundation are cooperatively collected, characteristic parameters of different ecological restoration stages and pseudo greening samples are extracted, and a characteristic database containing association relations between spectral characteristic vectors and corresponding ecological parameters is formed; S2, establishing a full life cycle digital twin file of a construction site, namely establishing a unique identifier for each construction disturbance point location, marking the boundary of each construction disturbance point location in a GIS, archiving multi-temporal hyperspectral data of a pre-construction background state, a pre-construction disturbance dynamic state, a post-repair acceptance state and a post-management monitoring state and derivative characteristics of the multi-temporal hyperspectral data according to a time axis, and forming a structured file; S3, performing repair quality quantitative calculation based on multi-feature fusion and intelligent recognition, namely training a classifier based on the characteristics of the pseudo greening sample in the spectrum fingerprint library, performing pseudo greening intelligent screening on the image in the acceptance period, and simultaneously obtaining the comprehensive ecological restoration index of the construction site by weighting and fusing the spectrum similarity restoration component, the vegetation health restoration component and the spatial structure restoration component; S4, introducing time sequence spectral dynamics to perform sustainable early warning and acceptance decision, namely analyzing a spectral change track from a disturbance state to an acceptance state, performing change detection and trend prediction on later-period management time sequence data, and triggering degradation risk early warning based on the result; s5, outputting results, namely automatically generating a comprehensive evaluation report comprising a pseudo greening risk distribution map, an ecological restoration degree index thematic map, a time sequence change track map, a degradation risk early warning report and an intelligent test conclusion receiving suggestion.
2. The quantitative acceptance method for power grid-trace ecological restoration based on a spectral fingerprint library according to claim 1, wherein in the step S1, the extracted spectral characteristic parameters comprise morphological characteristic parameters including red edge position, red edge slope, green peak reflectivity and water absorption valley depth; The vegetation index characteristic parameters comprise normalized vegetation index, enhanced vegetation index, photochemical vegetation index and normalized moisture index; And derivative spectrum characteristic parameters, namely a first-order differential spectrum and a second-order differential spectrum which are obtained after the reflection spectrum is smoothed.
3. The method for quantitative acceptance of grid-stitch ecological restoration based on spectral fingerprint library according to claim 1, wherein in S2, the digital twin file comprises the following multi-temporal data archiving phases: Stage T0, background state spectrum data before construction; the phase T1 is that spectrum data of the maximum disturbance state in construction; A T2 stage, namely repairing post-acceptance time point state spectrum data; and (3) a T & lt3+ & gt stage, namely later management and protection and long-term monitoring of state spectrum time sequence data.
4. The method for quantitative acceptance of grid-site ecological restoration based on spectral fingerprint library according to claim 3, wherein in S3, the calculating an ecological restoration degree index comprises: calculating a spectrum similarity recovery component, namely calculating the similarity of the spectrum curves of the acceptance period and the background period through a spectrum angle drawing method; Calculating a vegetation health recovery component based on the ratio or the difference value of a plurality of vegetation indexes in the acceptance period and the background period; Calculating a space structure recovery component, namely comparing and calculating vegetation coverage, plaque density, shape index and connectivity index based on the image in the acceptance period with a natural reference range; and (5) weighting and fusing the three components according to dynamically configured weights to obtain the comprehensive ecological restoration index.
5. The method for quantitative acceptance of grid-trace ecological restoration based on spectral fingerprint library according to claim 1, wherein in S3, the intelligent screening of pseudo greening comprises: Based on the pseudo greening sample and the real vegetation sample in the spectrum fingerprint library, extracting a distinguishing feature vector; Training a support vector machine or a convolutional neural network classifier, and optimizing model parameters to improve recall ratio; classifying pixels or objects of the image in the acceptance period to generate a pseudo greening risk probability distribution map; And if the area ratio of the suspected pseudo-greening area exceeds a preset threshold, triggering acceptance failing to pass the early warning.
6. The quantitative acceptance method for power grid and ground ecological restoration based on a spectral fingerprint library according to claim 1, wherein in the step S4, the time sequence spectral dynamics analysis comprises the steps of performing change vector analysis or minimum noise separation transformation on time sequence spectral data of a restoration process, and detecting vegetation state change; constructing a time sequence prediction model based on LSTM or GRU, and predicting vegetation state trend of a future monitoring period; and triggering yellow or red degradation risk early warning according to the change detection result and the prediction trend.
7. The quantitative acceptance method for power grid locus ecological restoration based on a spectrum fingerprint library according to claim 1, wherein in the step S4, the intelligent acceptance decision rule comprises judging that acceptance is not acceptable if false greening screening is not passed or an ecological restoration degree index is lower than an acceptable threshold; if the ecological restoration degree index is close to the threshold value but the restoration track is good and no early warning exists, judging that the ecological restoration degree index is qualified under the condition; And if all the indexes reach the standard and no early warning exists, judging that the acceptance is qualified.
8. The quantitative acceptance method for power grid-trace ecological restoration based on a spectral fingerprint library according to claim 1, wherein in S1, the spectral fingerprint library further comprises an ideal restoration path constructed based on a logical cliff growth model for comparison analysis with an actual restoration trace in S4.
9. The quantitative acceptance method for power grid-trace ecological restoration based on a spectral fingerprint library according to claim 1, wherein in the step S2, the digital twin file is further associated with and stores field investigation photographs, soil detection reports, restoration measure records and manual check records to form a complete evidence chain.
10. The quantitative acceptance method for grid-trace ecological restoration based on a spectral fingerprint library according to claim 1, wherein in S5, the comprehensive evaluation report is output in a structured document format and is associated to the digital twin file, so as to support WebGIS online interactive viewing and downloading.

Description

Quantitative acceptance method for power grid trace ecological restoration based on spectral fingerprint library Technical Field The application belongs to the technical field of ecological restoration monitoring and evaluation, and particularly relates to a quantitative acceptance method for power grid trace ecological restoration based on a spectrum fingerprint library. Background The construction of power grid engineering (such as a power transmission line, a transformer substation and the like) often causes significant disturbance to vegetation and soil along the surface of the line to form a construction site. In order to restore the ecological functions of the region, ecological restoration is needed to be implemented and scientific acceptance is needed. At present, the ecological restoration acceptance of the power grid site is dependent on manual field investigation and visual evaluation, and the following outstanding problems exist: the false greening behavior is difficult to find, part of construction units are subjected to acceptance, false instant greening artifacts are manufactured by adopting camouflage means such as laying inferior turfs, spraying green paint and the like, and the traditional visual or conventional remote sensing means are difficult to effectively identify. The existing method is mostly 'time point type' inspection and acceptance, and is difficult to continuously track and early warn the dynamic succession, long-term stability and potential degradation risk of the repair process. The data support is weak, and the lack of a systematic background spectrum database and a full life cycle data file leads to the lack of accurate and comparable data reference for the restoration effect evaluation. Although the remote sensing technology is gradually applied to ecological monitoring, the existing method is mostly dependent on mid-low resolution multispectral data, has limited inversion capability on fine ecological parameters such as vegetation physiological states, species composition and the like, and does not form a spectrum quantitative evaluation system which is deeply coupled with a power grid engineering disturbance-repair process, covers a plurality of platforms such as 'sky-air-ground', and runs through a full life cycle. Therefore, a quantitative acceptance method for achieving objective, efficient, accurate and dynamic ecological restoration of the power grid is needed. Disclosure of Invention The application provides a quantitative acceptance method for power grid trace ecological restoration based on a spectrum fingerprint library, and aims to solve the problems that the prior art is difficult to find pseudo greening behaviors and lacks of process dynamic monitoring. A quantitative acceptance method for power grid site ecological restoration based on a spectrum fingerprint library comprises the following steps: S1, constructing a full life cycle spectrum fingerprint library of a power grid typical habitat, wherein aiming at the type of the power grid typical habitat along the line, hyperspectral data observed at fixed points of a space-based satellite, an air-based unmanned aerial vehicle and a foundation are cooperatively collected, characteristic parameters of different ecological restoration stages and pseudo greening samples are extracted, and a characteristic database containing association relations between spectral characteristic vectors and corresponding ecological parameters is formed; S2, establishing a full life cycle digital twin file of a construction site, namely establishing a unique identifier for each construction disturbance point location, marking the boundary of each construction disturbance point location in a GIS, archiving multi-temporal hyperspectral data of a pre-construction background state, a pre-construction disturbance dynamic state, a post-repair acceptance state and a post-management monitoring state and derivative characteristics of the multi-temporal hyperspectral data according to a time axis, and forming a structured file; S3, performing repair quality quantitative calculation based on multi-feature fusion and intelligent recognition, namely training a classifier based on the characteristics of the pseudo greening sample in the spectrum fingerprint library, performing pseudo greening intelligent screening on the image in the acceptance period, and simultaneously obtaining the comprehensive ecological restoration index of the construction site by weighting and fusing the spectrum similarity restoration component, the vegetation health restoration component and the spatial structure restoration component; S4, introducing time sequence spectral dynamics to perform sustainable early warning and acceptance decision, namely analyzing a spectral change track from a disturbance state to an acceptance state, performing change detection and trend prediction on later-period management time sequence data, and triggering degradation risk early warni