CN-121982559-A - Narrow space exploration method based on multi-criterion decision and adaptive light projection

CN121982559ACN 121982559 ACN121982559 ACN 121982559ACN-121982559-A

Abstract

The invention discloses a narrow space exploration method based on multi-criterion decision and self-adaptive light projection, and relates to the technical field of ecological restoration; the method comprises the steps of calculating flood time distribution of a river beach under an over-frequency flood scene through a Weibull formula, extracting information such as vegetation type, coverage and the like of the beach based on multi-source remote sensing images, dividing proper plant subareas by combining the flood time, constructing a habitat quality evaluation system from dimensions such as vegetation conditions, hydrologic conditions, topography characteristics and the like to finish the habitat quality rating of each subarea, and finally providing ecological restoration subareas according to the flood time characteristics, the suitability of the plants and the habitat quality rating. The invention realizes the precise coupling of the flood hydrologic process and the beach ecosystem, solves the problem that the traditional ecological restoration measures lack the "hydrologic-ecological" cooperative basis, improves the scientificity and feasibility of the restoration scheme, and is suitable for the ecological restoration engineering of the water area surrounding such as the northern river beach, the lake wetland and the like in arid and semi-arid areas.

Inventors

YANG YUDONG
LIU ZHANZHU
LIANG XUESONG
ZHAO SHUANG
HAN YUWEI

Assignees

长春市万易科技有限公司

Dates

Publication Date: 20260505
Application Date: 20251225

Claims (5)

1. A narrow space exploration method based on multi-criterion decision and adaptive light projection is characterized by comprising the following steps: step one, calculating the flooding time of the super-frequency flood, namely collecting annual maximum flood peak flow serial data of a target river beach control hydrologic station, and calculating the beach flooding time space distribution under different super-frequency flood reproduction periods after pretreatment and oversized flood treatment; dividing plants in proper subareas, namely drawing a beach space flooding time distribution map by combining flooding time; Acquiring multisource remote sensing image data, and extracting beach vegetation type, coverage and terrain elevation information; performing beach field plant investigation, flooding stress experiments and literature investigation, and dividing beach land plant partitions according to plant growth characteristics; thirdly, determining the beach repair subareas, namely evaluating the quality of the beach current situation, constructing a habitat quality evaluation index system comprising vegetation status, hydrologic conditions, topography characteristics and soil quality, determining index weights by adopting an analytic hierarchy process, and grading the quality of each plant subarea; And (3) coupling plants and habitats to perform multidimensional parameter partition simulation to form a beach ecological restoration partition scheme.
2. The method for exploring a narrow space based on multi-criterion decision-making and adaptive light projection of claim 1, wherein the calculating the super-frequency flood flooding time in the first step comprises: Step A1, data preprocessing, namely collecting a annual maximum flood peak flow series of a control hydrologic station, and carrying out consistency correction, missing value supplementation and abnormal value inspection to ensure that the sample series meets the consistency, representativeness and independence requirements; Step A2, extra-large flood treatment, namely identifying extra-large flood of actual measurement and history investigation, determining the length N of a history investigation period, adjusting the sequencing numbers of the extra-large flood and common flood, and calculating the experience frequency and the reproduction period by adopting a corrected Weibull formula; a3, building a flooding time model, namely building a response relation between flood peak flow and a beach flooding range and a flooding depth, and calculating flooding time of each space unit of the beach under different super-frequency flood reproduction periods by combining a flood drainage curve; And A4, generating flooding time space distribution, namely performing on-site measurement based on a beach Digital Elevation Model (DEM), correcting a beach area of the constructed DEM with accuracy of +/-2 cm, and obtaining the flooding time space distribution by spatial interpolation.
3. The narrow space exploration method based on multi-criterion decision and self-adaptive light projection of claim 1, wherein in the second step, the multi-source remote sensing image comprises a 10m resolution Sentinel-2 image and a 30m resolution Landsat-9 image, the vegetation information extraction adopts an object-oriented classification method in combination with spectral indexes, and the plant suitable partition is divided into a long-term flooding-resistant plant area, a short-term flooding-resistant plant area, a flooding-avoiding plant area and a pioneer plant improvement area in an unsuitable area according to flooding time.
4. The narrow space exploration method based on multi-criterion decision and self-adaptive light projection of claim 1, wherein the habitat quality evaluation index system in the third step comprises vegetation coverage, vegetation diversity, flooding time suitability, terrain gradient, soil organic matter content, soil water content and saline-alkali degree, wherein each index weight is determined by combining an analytic hierarchy process with an entropy weight method, and evaluation grade is divided into four grades of excellent, good, medium and poor.
5. The method for exploring a narrow space based on multi-criterion decision-making and adaptive light projection of claim 1, wherein the method for comprehensively analyzing plant parameters and habitat parameters in the fourth step comprises the following steps: s1, evaluating the current situation of a plant partition ecological system by adopting a standardized method; s2, multi-factor weighted coupling, namely calculating an analog partition by adopting methods such as random forests and the like; And S3, generating an ecological restoration partition grid map according to the partition result of the random forest and the DEM.

Description

Narrow space exploration method based on multi-criterion decision and adaptive light projection Technical Field The invention relates to the technical field of ecological restoration, in particular to a narrow space exploration method based on multi-criterion decision and self-adaptive light projection. Background The exploration of autonomous robots in unknown or partially unknown environments is a fundamental and key research problem, and the core task of the autonomous robots is to guide the robots to efficiently and comprehensively sense the environment structure and construct an accurate map. The technology is a core enabling technology for search and rescue (SAR), facility inspection, military reconnaissance and other applications. Among the numerous exploration strategies, "front-of-the-line exploration (front-BasedExploration)" is recognized as one of the most classical and widely used methods, proposed by Yamauchi in 1997. The method defines a "leading edge" as the boundary of a known region with an unknown region. The goal of the robot is to continually find and move to these leading edge points, gradually expanding the range of the known map. The frontiers method becomes a reference algorithm in the field and is the closest prior art because of clear concept and relatively simple implementation. The basic implementation flow of front-edge exploration generally includes: Leading edge detection, namely identifying a grid set (namely leading edge point cluster) bordering an unknown area through algorithms such as edge detection or area growth based on occupancygridmap (occupying a grid map) which is constructed currently. Candidate target generation, namely clustering the front edge point clusters, and calculating the center or representative point of each cluster as a candidate exploration target. Target selection-typically, an optimal target is selected from the candidate targets according to some criteria (e.g., shortest distance, maximum leading edge size). Navigation and execution the robot plans a path and moves to the target point, repeating this process until no new leading edge appears. With the development of technology, researchers have performed a number of improvements on classical leading edge exploration, wherein a modified leading edge exploration scheme combining information gain evaluation and multi-objective optimization is more similar to the concept of the invention. The protocol generally comprises the steps of: Candidate point sampling-instead of simply taking the center of the leading edge cluster as the candidate point, random sampling (similar to RRT) in the leading edge region or robot-surrounding free space, or generating a series of candidate views (NBV) within the sensor cone. Multiple criteria evaluation multiple objective functions are evaluated for each candidate point, the three most common criteria being: Information gain (InformationGain) estimates the volume or area of the unknown region that can be observed from the candidate point. Path cost PathCost the euclidean distance or estimated travel time is typically used to represent the cost of reaching the candidate point. Accessibility (Reachability) simply determines if the point is physically reachable (e.g., if there is a collision with an obstacle). Decision model the multi-objective decision is made using a weighted sum model (WeightedSumModel, WSM). That is, each criterion is assigned a fixed weight (e.g., w_info, w_cost, w_reach), and then the criterion scores for each candidate point are weighted and summed to obtain a comprehensive Utility value (Utility). And finally selecting the candidate point with the highest utility value as a target. Utility=W_info*Score_info+W_cost*Score_cost+W_reach*Score_reach The method improves the performance of classical front-edge exploration to a certain extent by considering a plurality of optimization targets, is a more advanced and common implementation scheme in the current research, and is also the most similar prior art implementation scheme with the invention. Despite the advances made in the above-described prior art, there are still the following significant drawbacks and limitations when applied to narrow, closed, complex multi-layer environments: the candidate point generation mechanism is inefficient and does not adapt to a narrow environment: classical leading edge exploration creates a large number of redundant leading edge points in complex structures, and the computational overhead of clustering and processing is large. The random sampling or NBV method is effective in a wide space, but in a narrow channel, a large number of sampling points can be positioned in an obstacle or blocked, the invalid sampling rate is high, and the waste of calculation resources is serious. The generation of candidate points is static and cannot be adaptively adjusted according to the spaciousness or narrowness of the environment, resulting in insufficient fine exploration in a