CN-121582794-B - Wetland plant identification method and system based on multidimensional feature similarity comparison

CN121582794BCN 121582794 BCN121582794 BCN 121582794BCN-121582794-B

Abstract

The invention relates to the technical field of machine vision and discloses a wetland plant identification method and system based on multidimensional feature similarity comparison, wherein the method comprises the steps of taking color features as a macroscopic information basis, taking texture features as microcosmic detail supplement, and constructing basic feature description; the method comprises the steps of counting color distribution and texture trend in image data to obtain multi-scale feature representation, matching geometric positions and descriptors in the multi-scale feature representation to establish a feature manifold of the multi-scale feature representation, guiding the multi-scale feature representation based on the feature manifold to generate fusion feature representation, selecting a primary comparison feature path according to association relation strength in fusion feature identification to obtain comparison strategy, comparing the fusion feature representation with a known wetland plant feature database to obtain similarity scores, and carrying out consistency check on growth environment context information and climatic period context information to obtain category identification results.

Inventors

Fang Guojing
ZHANG FENG
CHEN CHENG
XU XIAOWEI
YE ZHOUXING
Zhai Xuheng

Assignees

浙江鸿森生态科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260126

Claims (10)

1. A wetland plant identification method based on multidimensional feature similarity comparison, which is characterized by comprising the following steps: S1, constructing basic feature description of the wetland plant by taking color features of the wetland plant image data as a macroscopic information basis and taking texture features as microcosmic detail supplement; S2, based on the basic feature description, counting the color distribution and texture trend of the visual saliency area in the image data to obtain multi-scale feature representation of the wetland plant; S3, matching geometric positions and descriptors of features among different scales in the multi-scale feature representation to establish a feature manifold of the multi-scale feature representation, and guiding the multi-scale feature representation based on the feature manifold to generate a fusion feature representation of the wetland plant; S4, selecting a primary comparison characteristic path according to the association relation strength among different characteristic dimensions in the fusion characteristic representation to obtain a comparison strategy of the wetland plant; s5, applying the comparison strategy, and comparing the fusion characteristic representation with a known wetland plant characteristic database in similarity to obtain a similarity score of the wetland plant; and S6, based on the similarity score, carrying out consistency check on the growth environment context information and the climatic period context information of the wetland plants to obtain a type identification result of the wetland plants.
2. The method for identifying wetland plants based on multidimensional feature similarity comparison according to claim 1, wherein the constructing basic feature description content of the wetland plants based on the color features of the wetland plant image data as a macroscopic information basis and the texture features as microcosmic detail supplement comprises the following steps: Suppressing high-frequency noise in the wetland plant image data to obtain denoising image data of the wetland plant; Expanding the gray level distribution of the denoising image data to obtain contrast enhancement image data of the wetland plants; Dividing a color space from the contrast enhanced image data, and counting the distribution rule of the dominant hue in the color space; according to the distribution rule, the color of the contrast enhancement image data is used as a macroscopic information base of the wetland plants; supplementing microcosmic details of the wetland plants with texture features of the contrast enhanced image data; Combining the macro information base with the micro detail supplement to construct a base characterization of the wetland plant.
3. The method for identifying wetland plants based on multidimensional feature similarity comparison according to claim 2, wherein the step of counting the color distribution and texture trend of the visual saliency areas in the image data based on the basic feature description to obtain the multi-scale feature representation of the wetland plants comprises the steps of: Detecting a difference image area in the contrast enhancement image data based on the basic feature description to obtain a visual saliency area of the wetland plant; Establishing a concentric analysis window of the wetland plant by taking the geometric center of the visual saliency area as an origin; in the concentric analysis window, tensor synthesis is carried out on the main hue and the distribution proportion of the contrast enhancement image data, so as to obtain the color distribution vector of the wetland plant; Tracking the gray level change gradient direction of the contrast enhancement image data to obtain the texture trend description of the wetland plants; And superposing the texture trend description on the color distribution vector to obtain the multi-scale characteristic representation of the wetland plant.
4. The method of claim 1, wherein the matching geometric locations and descriptors of features between different scales in the multi-scale feature representation to create a feature manifold of the multi-scale feature representation, and guiding the multi-scale feature representation based on the feature manifold to generate a fused feature representation of the wetland plant comprises: Under a unified coordinate system, aligning the geometric positions of the coarse scale features and the fine scale features in the multi-scale feature representation to obtain the spatial correspondence between scales of the wetland plants; screening out a feature corresponding group of the wetland plants according to the descriptor similarity degree of different scale features in the multi-scale feature representation; constructing a topological skeleton represented by the multi-scale features according to the spatial correspondence between scales, and merging the topological skeleton into the feature correspondence group to obtain a feature manifold of the wetland plant; And transmitting and fusing the detail information of the fine-scale features to the coarse-scale features along the topological structure of the feature manifold to obtain the fused feature representation of the wetland plants.
5. The method for identifying wetland plants based on multidimensional feature similarity comparison according to claim 4, wherein the topological structure along the feature manifold transfers and fuses the detailed information of the fine-scale features to the coarse-scale features to generate a fused feature representation of the wetland plants, and the method comprises the following steps: identifying the information transmission paths of the fine-scale features and the coarse-scale features according to the topological connection relation of the feature manifold; Transmitting local texture structure information contained in the fine-scale features to the region of the coarse-scale features along the information transmission path to realize the trans-scale migration of the fine-scale features; In the trans-scale migration process, eliminating feature expression conflict between the coarse-scale features and the fine-scale features according to the geometric constraint of the feature manifold to obtain coordinated feature expression of the wetland plants; Optimizing the distribution continuity of the coordination feature expression in the manifold space to obtain the fusion feature expression of the wetland plant.
6. The method for identifying wetland plants based on multidimensional feature similarity comparison according to claim 1, wherein the selecting a primary comparison feature path according to the association relationship strength between different feature dimensions in the fusion feature representation to obtain the comparison strategy of the wetland plants comprises: Determining the connection strength between the feature dimensions according to the co-occurrence relation between the feature dimensions in the fusion feature representation; Constructing a characteristic topological graph of the wetland plant by taking the characteristic dimension as a node and the connection strength as an edge; obtaining a centrality index of the wetland plant according to the relative proportion of the number of the associated edges of the nodes in the characteristic topological graph in the total number of the associated edges of the characteristic topological graph; And screening the highest node in the centrality index as a path starting point, and determining a characteristic comparison sequence according to the edge with the highest connection strength in the node to obtain the comparison strategy of the wetland plants.
7. The method for identifying wetland plants based on multidimensional feature similarity comparison according to claim 1, wherein the applying the comparison strategy to compare the similarity between the fusion feature representation and a database of known wetland plant features to obtain similarity scoring content of the wetland plants comprises the following steps: Extracting features to be compared of the wetland plants from the fusion feature representation according to the feature path sequence determined in the comparison strategy; comparing the features to be compared with features in a known wetland plant feature database item by item to obtain the feature consistency degree of the wetland plants; And quantifying the similarity relation between the fusion characteristic representation and the known wetland plant characteristic database according to the characteristic consistency degree between the characteristics to obtain the similarity score of the wetland plants.
8. The method for identifying wetland plants based on multidimensional feature similarity comparison according to claim 6, wherein the similarity score is calculated according to the following formula: ; in the formula, The degree of similarity is scored as such, For the total number of feature dimensions of the fused feature representation, The first to represent the fusion feature The values of the dimensions of the features, A first of the corresponding features in the database of known wetland plant characteristics The values of the dimensions of the features, For a preset characteristic dimension Is used for the wide tolerance coefficient of the (c), For the feature dimension in the feature topological graph And (3) with The connection strength between the two components is high, Is the characteristic dimension A node centrality index in the feature topology graph, For a preset neighbor influencing factor, And the sum of the centrality indexes of all the feature dimension nodes in the feature topological graph is used for normalizing the weight.
9. The method for identifying wetland plants based on multidimensional feature similarity comparison according to claim 1, wherein the step of performing consistency check on the growth environment context information and the climatic period context information of the wetland plants based on the similarity score to obtain the type identification result of the wetland plants comprises the following steps: Screening out the candidate wetland plant types of the wetland plants according to the candidate type sequence in the similarity score; taking the water area type and soil humidity information of the area where the wetland plants are located as the growth environment context information of the wetland plants; taking the corresponding relation between the current growth stage and seasons of the wetland plants as the climatic period context information of the wetland plants; Removing the environmental difference types in the wetland plants according to the compatibility of the types of the candidate wetland plants and the growing environment context information to obtain environmental compatibility candidate types of the wetland plants; Checking the coincidence degree of the candidate wetland plant types and the climatic period context information, and screening out candidate types with consistent climatic conditions of the wetland plants; and combining the environment compatible candidate category and the weather consistent candidate category to determine a category identification result of the wetland plant.
10. A wetland plant identification system based on multidimensional feature similarity comparison, for implementing the wetland plant identification method based on multidimensional feature similarity comparison of claim 1, the system comprising: the basic feature construction module is used for constructing basic feature description of the wetland plant by taking the color feature of the wetland plant image data as a macroscopic information basis and taking the texture feature as microcosmic detail supplement; The multi-scale feature representation module is used for counting the color distribution and the texture trend of the visual saliency area in the image data based on the basic feature description to obtain multi-scale feature representation of the wetland plant; the fusion feature generation module is used for matching geometric positions and descriptors of features among different scales in the multi-scale feature representation to establish a feature manifold of the multi-scale feature representation, guiding the multi-scale feature representation based on the feature manifold and generating a fusion feature representation of the wetland plant; The comparison strategy making module is used for selecting a primary comparison characteristic path according to the association relation strength among different characteristic dimensions in the fusion characteristic representation to obtain comparison strategy of the wetland plant; The similarity comparison scoring module is used for applying the comparison strategy, and comparing the fusion characteristic representation with a known wetland plant characteristic database in similarity to obtain a similarity score of the wetland plant; And the type identification result generation module is used for carrying out consistency check on the growth environment context information and the climatic period context information of the wetland plants based on the similarity score to obtain the type identification result of the wetland plants.

Description

Wetland plant identification method and system based on multidimensional feature similarity comparison Technical Field The invention relates to the technical field of machine vision, in particular to a wetland plant identification method and system based on multidimensional feature similarity comparison. Background The existing wetland plant identification technology has significant limitation in the feature extraction link, and often only focuses on the utilization of single type features, and the macro global information carried by color features and the micro detail information corresponding to texture features cannot be effectively combined, so that the constructed basic feature description lacks comprehensiveness and integrity. Meanwhile, the technology is not perfect enough for preprocessing the original image data, cannot fully inhibit high-frequency noise interference, and does not reasonably expand gray level distribution to improve image contrast, so that the quality of the extracted features is low, the recognition degree is insufficient, hidden hazards are buried for subsequent feature matching and recognition work, and the reliability of recognition results is directly affected. In the process of feature fusion and identification judgment, the prior art fails to properly solve the problem of coordination among multi-scale features, lacks accurate alignment of geometric positions of features with different scales and effective matching of descriptors, lacks scientific guidance in construction of feature manifold, causes hindered transmission of cross-scale feature information and poor fusion effect, and is difficult to form fusion feature representation with global relevance and local detail. In addition, the existing comparison method lacks pertinence, does not optimize the comparison path according to the association relation strength among characteristic dimensions, neglects the verification effect of the context information such as the growth environment, the weather period and the like on the identification result, not only causes low comparison efficiency, but also is easy to generate identification deviation due to characteristic expression conflict or environment and weather suitability mismatch, and cannot meet the requirements of efficient and accurate identification on wetland plants in practical application, so that how to improve the efficiency of wetland plant identification becomes a problem to be solved urgently. Disclosure of Invention The invention provides a wetland plant identification method and system based on multidimensional feature similarity comparison, which are used for solving the problems in the background technology. In order to achieve the above purpose, the invention provides a wetland plant identification method based on multidimensional feature similarity comparison, which comprises the following steps: S1, constructing basic feature description of the wetland plant by taking color features of the wetland plant image data as a macroscopic information basis and taking texture features as microcosmic detail supplement; S2, based on the basic feature description, counting the color distribution and texture trend of the visual saliency area in the image data to obtain multi-scale feature representation of the wetland plant; S3, matching geometric positions and descriptors of features among different scales in the multi-scale feature representation to establish a feature manifold of the multi-scale feature representation, and guiding the multi-scale feature representation based on the feature manifold to generate a fusion feature representation of the wetland plant; S4, selecting a primary comparison characteristic path according to the association relation strength among different characteristic dimensions in the fusion characteristic identification to obtain a comparison strategy of the wetland plant; s5, applying the comparison strategy, and comparing the fusion characteristic representation with a known wetland plant characteristic database in similarity to obtain a similarity score of the wetland plant; and S6, based on the similarity score, carrying out consistency check on the growth environment context information and the climatic period context information of the wetland plants to obtain a type identification result of the wetland plants. In a preferred embodiment, the constructing the description of the basic features of the wetland plant based on the color features of the wetland plant image data and the microcosmic details of the texture features comprises: Suppressing high-frequency noise in the wetland plant image data to obtain denoising image data of the wetland plant; Expanding the gray level distribution of the denoising image data to obtain contrast enhancement image data of the wetland plants; Dividing a color space from the contrast enhanced image data, and counting the distribution rule of the dominant hue in the color space; according to the d