Intelligent river system classification method and device based on local feature standardization

Abstract

The present invention belongs to the field of computer map making technology, and specifically relates to a method and apparatus for intelligent river system classification based on local feature standardization. The method comprises: first, preprocessing and extracting river segment features based on existing river system classification results; second, local standardization of river segment features; then, inputting the model into a random forest classification model for supervised learning, and applying the trained model to the classification of main tributaries in unclassified river systems; and finally, establishing a hierarchical tree of the river system based on the main tributary classification results, assigning hierarchical codes to the river segments in the river system. The automatic river system classification of the present invention has a high degree of consistency with manual classification results, can achieve high-quality automatic classification of river systems, and supports further applications such as river system selection and cartographic expression.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer map making, and in particular relates to a method and device for intelligently grading river systems based on local standardization of features. Background Art

[0002] River system classification is the key point of river system selection (Tan Xiao, Wu Fang, Huang Qi, et al. Multi-criteria decision model for mainstream identification and its application in river system structuring [J]. Acta Geodaetica et Cartographica Sinica, 2005, 34(2):154-160.DOI:10.3321 / j.issn:1001-1595.2005.02.012.). Common classification methods include Horton classification (HORTON RE. Erosional development of streams and their drainage basins; hydrophysical approach to quantitative morphology [J]. Geological Society of America bulletin, 1945, 56(3):275-370.DOI:10.1130 / 0016-7606(1945)56[275:EDOSAT]2.0.CO; 2.), Strahler classification (STRAHLER A N. Quantitative analysis of Watershed geomorphology[J]. Eos, Transactions American Geophysical Union, 1957, 38(6):913-920. DOI:10.1029 / TR038i006p00913.), Shreve classification (SHREVE R L. Statistical law of stream numbers[J]. The Journal of Geology, 1966, 74(1):17-37.), etc. Compared with the simple increase in the hierarchy starting from the terminal tributary, the Horton classification defines the smallest unbranched tributary as the first level and the mainstream as the highest level to establish the river system hierarchy. This has become the main principle of river system classification. The difficulty lies in how to accurately identify the mainstream.Therefore, relevant scholars have carried out a lot of research. Tan Xiao et al. (2005) determined the main streams of river systems at all levels by the maximum membership weighted average programming method in the multi-criteria decision-making model; Zhang Yuanyu et al. (Zhang Yuanyu, Li Lin, Jin Yuping, et al. Research on the structural drawing model of tree-like river system based on graph theory [J]. Journal of Wuhan University (Information Science Edition), 2004, 29 (6): 537-539, 543.), Zhai Renjian et al. (Zhai Renjian, Xue Benxin. Research on the structural model of river system for automatic synthesis [J]. Journal of Surveying and Mapping Science and Technology, 2007, 24 (4): 294-298, 302.) and Guo Qingsheng et al. (Guo Qingsheng, Huang Yuanlin. Automatic reasoning of the main stream of tree-like river system [J]. Journal of Wuhan University (Information Science Edition), 2008, 33 (9): 978-981.) Based on the 180° hypothesis (PAIVA J, EGENHOFER MJ, FRANK AU. Spatial reasoning about flow directions:towards an ontologyfor river networks[C] / / Proceedings of the XV International Congress for Photogrammetryand Remote Sensing.[S.1.]:ISPRS,1992:224-318.), respectively, based on different judgment rules, identifying the mainstream of the river system by searching the river sections. Li Chengming et al. (Li Chengming, Yin Yong, Wu Wei, et al. Construction and simplification of tree-like river system hierarchical relationships based on stroke feature constraints [J]. Acta Geodaetica et Cartographica Sinica, 2018, 47(4): 537-546. DOI: 10.11947 / i.AGCS.2018.20170141.) used stroke connections to determine the hierarchical relationship of river systems based on the stroke "good continuity" principle (THOMSON RC, RICHARDSON DE. The 'good continuation' principle of perceptual organization applied to the generalization of road networks [C] / / Proceedings of the ICA 19th international cartographic conference. Ottawa, Canada, 1999: 1215-1223.). The above method mainly starts from the local structure of the river system, and the defined rules are difficult to apply to all situations.To this end, Duan Peixiang et al. (Duan Peixiang, Qian Haizhong, He Haiwei, et al. Naive Bayesian tree-based river system automatic classification method based on case studies [J]. Acta Geodaetica et Cartographica Sinica, 2019, 48(8): 975-984. DOI: 10.11947 / j.AGCS.2019.20180370.) used the naive Bayesian algorithm to avoid the problem of subjective setting of indicator weights through case studies of existing river system classification results. However, the influence of river basin characteristics was not considered in feature selection. Summary of the Invention

[0003] The present invention proposes a method and device for intelligent river system classification based on local feature standardization. The method of the present invention focuses on the automation and intelligence of river system classification to better meet the requirements of applications such as river system selection and cartographic expression. It first designs six river section feature indicators based on the geometric structure of the river section and the local and overall structure of the river system. After the river section features are locally standardized, they are input into the random forest algorithm for supervised training. The unclassified river system is input into the trained model, and a hierarchical tree of the river system is established according to the main and tributary classification results, and hierarchical codes are assigned to the river sections in the river system.

[0004] In order to achieve the above object, the present invention provides the following technical solutions:

[0005] The intelligent river system classification method based on local feature standardization includes the following steps:

[0006] S1: Based on the existing river system classification results, perform topological preprocessing and river segment feature extraction;

[0007] The river section characteristics are the characteristics of the main tributaries of the river section, which are selected based on the geometry of the river section and the local and overall structure of the river system. They include six river section characteristic indicators:

[0008] (1) River section length (RLen). This indicates the length of the current river section. The larger the RLen value, the greater the probability that it is a mainstream river section.

[0009] (2) Upstream reach length (URL). This represents the total length of the current river reach and its upstream reaches. The larger the URL value, the greater the probability that it is a mainstream river reach.

[0010] (3) River reach drainage area (WArea). This refers to the area of ​​the region that the current river reach flows through. WArea reflects the influence range of the river reach. The larger the area, the greater the probability that it is a mainstream river reach.

[0011] (4) Upstream drainage area (URA). This represents the total area of ​​the region drained by the current river reach and its upstream reaches. The URA reflects the influence range of the river reach and its upstream reaches. The larger the area, the greater the probability that it is a mainstream river reach.

[0012] (5) Confluence angle (BAngle). This represents the confluence angle between the upstream and downstream reaches. The closer the BAngle is to 180°, the greater the probability that it is a mainstream reach.

[0013] (6) Number of tributaries (TNum). This indicates the number of upstream river sections of the current river section. The larger the TNum value, the greater the probability that it is a mainstream river section.

[0014] S2: Locally standardize the river section characteristics. Local standardization of river section characteristics is to standardize the characteristics using the maximum value of the characteristics of the adjacent upstream river section. Specifically, the characteristic value of the current river section is divided by the maximum value of the characteristics of the adjacent upstream river section.

[0015] Among them, the three types of features, URL, URA and TNum, are locally normalized, while the three types of features, RLen, WArea and Bangle, are normalized using Min-Max.

[0016] S3: Input the river section characteristics obtained after local normalization into the random forest classification model for supervised learning, and apply the trained model to the classification of main tributaries of unclassified river systems; the application means that after the training is completed, the trained model is used to identify the main tributaries of each river section in the river system to be classified, and derive the mainstream probability of each river section.

[0017] S4: Identification and hierarchical structuring of river system mainstreams: handle abnormal situations according to the mainstream probability, and then use the main tributary classification results to establish a river system hierarchy tree and assign hierarchical codes to the river sections in the river system.

[0018] The specific operation of handling anomalies is: compare the mainstream probability of each river section, select the river section with the optimal value as the mainstream, and set the remaining river sections as tributaries.

[0019] The specific operations of assigning hierarchical codes are:

[0020] S4.1: Create a new hierarchical tree root node to store the river section ID of the river system estuary, and start from the river section and its tree node to execute S4.2; after all the main tributaries of the river section are determined and the hierarchical tree is constructed, execute S4.4.

[0021] S4.2: Determine the type of the upstream river section of the river section. If the upstream river section is a mainstream river section, store its ID in the tree node and continue to execute S4.2; otherwise, execute S4.3.

[0022] S4.3: Create a new hierarchical tree child node, which stores the river segment ID and continues to determine its upstream river segment. If the upstream river segment is the mainstream river segment, continue to execute S4.2; otherwise, execute S4.3.

[0023] S4.4: According to the hierarchical classification of the hierarchical tree, the river sections contained in all leaf nodes are assigned a value of 1, the river sections contained in the upper-level tree nodes are assigned a value of 2, and so on; the river sections contained in the root node are assigned the highest level, and the hierarchical code of the river system is established.

[0024] The present invention also provides an intelligent river system classification device based on local feature standardization, comprising:

[0025] River section feature extraction module, which is used to extract 6 river section features and perform local standardization of river section features based on the existing river system classification results.

[0026] The training prediction module is used to input the standardized river segment characteristics and main tributary labels into the random forest classification model for supervised training. After training, the trained model is used to identify the main tributaries of each river segment in the river system to be classified and derive the probability of mainstream for each river segment.

[0027] The result output module is used to process abnormal situations according to the mainstream probability, and then use the main tributary classification results to establish a river system hierarchy tree and assign hierarchical codes to the river sections in the river system.

[0028] The present invention also provides an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for intelligent river system classification based on local standardization of features is implemented.

[0029] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent river system classification method based on local standardization of features.

[0030] The present invention has the following beneficial effects:

[0031] The intelligent river system classification results of the present invention are highly consistent with those of manual classification, enabling high-quality automatic river system classification. Compared with other machine learning algorithms, the automatic river system classification supported by the random forest algorithm has higher accuracy and better classification results. Local feature normalization can significantly improve the classification accuracy of machine learning algorithms. Compared with other river system classification methods, the river system classification results of the present invention can improve the accuracy of river system selection. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 Flowchart for the implementation of an intelligent river system classification method based on local normalization of features.

[0033] Figure 2 Schematic diagram of river section characteristics.

[0034] Figure 3Schematic diagram for confluence angle calculation.

[0035] Figure 4 Construct a river system hierarchy tree and assign hierarchical codes.

[0036] Figure 5 This is the test data for automatic classification of river systems.

[0037] Figure 6 These are the mainstream classification results of different standardization methods.

[0038] Figure 7 The results of mainstream identification of River System 9 using different standardization methods.

[0039] Figure 8 The feature importance results of different normalization methods.

[0040] Figure 9 These are the mainstream classification results of different machine learning algorithms. DETAILED DESCRIPTION

[0041] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0042] Example 1

[0043] The present invention firstly designs six river section characteristic indicators as the features for main tributary classification based on the geometric structure of river sections and the local and overall structures of river systems. Secondly, a local standardization processing strategy for river section characteristics is introduced in view of the differences in the numerical range and dimension of different river section characteristics. Then, six features of samples from existing river system classification results are extracted, and after local standardization, they are input into the random forest algorithm for supervised learning, and the trained model is applied to the main tributary classification of unclassified river systems. Finally, abnormal situations are handled according to the main tributary classification probability, and the river system hierarchy tree is established using the main tributary classification results to assign hierarchical codes to the river sections in the river system.

[0044] The implementation process of intelligent river system classification based on local standardization of characteristics is as follows: Figure 1 As shown in the figure, the specific implementation process includes:

[0045] S1: Based on the existing river system classification results, perform topological preprocessing and river segment feature extraction

[0046] The topology preprocessing is to use GIS software to perform topology checking, and then manually edit and process to ensure that the river systems are connected (that is, there are no suspensions or disconnections between the river systems).

[0047] Feature selection is a crucial step in applying machine learning methods. Selecting effective features can improve model accuracy, help uncover implicit knowledge about main-tributary classification, and ultimately, obtain a reliable main-tributary classification model. According to Horton's three laws of river system composition: the length law, the area law, and the number law, the length, number, and drainage area of ​​a river system vary geometrically with the system's hierarchy. This means that the mainstream should prioritize length, drainage area, and tributary number over its tributaries. Furthermore, according to the 180° hypothesis (PAIVA J, EGENHOFER MJ, FRANK AU. Spatial reasoning about flow directions: toward an ontology for river networks [C] / / Proceedings of the XV International Congress for Photogrammetry and Remote Sensing. [S.1.]: ISPRS, 1992: 224-318.), the flow directions of upstream and downstream reaches of a mainstream confluence are nearly aligned, so the upstream and downstream reaches of the mainstream should exhibit directional consistency. In summary, based on the geometric structure of river sections and the local and overall structure of the river system, six indicators were selected as characteristics for the classification of main and tributary rivers, including:

[0048] (1) River section length (RLen). This indicates the length of the current river section. The larger the RLen value, the greater the probability that it is a mainstream river section.

[0049] (2) Upstream reach length (URL). This represents the total length of the current river reach and its upstream reaches. The larger the URL value, the greater the probability that it is a mainstream river reach.

[0050] (3) River section drainage area (WArea). It indicates the area of ​​the area through which the current river section flows. Figure 2 As shown, river section R i The drainage area is WArea i , River Section R j The drainage area is WArea j , River Section R k The drainage area is WArea kWArea reflects the influence range of a river section. The larger the area, the greater the probability that it is a mainstream river section. Considering that the spatial density differences between river sections lead to changes in the spatial morphology of the river network, the river section watershed unit is constructed using the constrained Delaunay triangulation and skeleton line connection method (Ai Tinghua, Liu Yaolin, Huang Yafeng. Hierarchical triangulation and map synthesis of watershed areas [J]. Acta Geodaetica et Cartographica Sinica, 2007, 36(2): 231-236, 243.).

[0051] (4) Upstream drainage area (URA). It refers to the total area of ​​the area that the current river section and its upstream river sections flow through. Figure 2 As shown, river section R j The total area of ​​the region where the upstream river flows (i.e. the upstream drainage area) is URA. j , River Section R k The total area of ​​the region where the upstream river flows (i.e. the upstream drainage area) is URA. k URA reflects the influence range of a river reach and its upstream reaches; the larger the area, the greater the probability that it is a mainstream river reach.

[0052] (5) Confluence angle (BAngle). It indicates the confluence angle between the upstream and downstream river sections. Figure 3 As shown, river section R i The confluence angle is calculated as the angle between the downstream section R k BAngle i The closer the BAngle is to 180°, the greater the probability that it is the mainstream river section. Figure 3 The middle arrow indicates the flow direction.

[0053] (6) Number of tributaries (TNum). This indicates the number of upstream river sections of the current river section. The larger the TNum value, the greater the probability that it is a mainstream river section.

[0054] S2: Local normalization of river reach characteristics

[0055] Different characteristics of river reaches vary in their numerical ranges and dimensions. To facilitate data comparison, ensure better model convergence, and prevent overfitting, standardization methods are often used to eliminate these differences. Common data standardization methods include Z-Score and Min-Max standardization, which primarily standardize the data from a holistic perspective. To avoid weakening the correlation between upstream reach characteristics due to global standardization, local standardization of reach characteristics is introduced for optimization.

[0056] The local standardization of river section characteristics is to use the maximum value of the characteristics of the adjacent upstream river section to standardize the characteristics. Figure 2 As shown, taking the characteristic URA as an example, the river section R iThe adjacent upstream river sections are river sections R j and R k , River Section R j and R k The upstream basin areas are URA j and URAk, then URA j The local normalization is performed by dividing URAj and URAk by max(URAj, URAk). Local normalization is performed on the URL, URA, and TNum river system local and global structural features, while the RLen, WArea, and Bangle features are processed using Min-Max normalization. The final river system feature normalization results are obtained for training and learning.

[0057] S3: Training and application of river system main tributary classification model

[0058] The relationship between the main tributaries of a river system is included in the relationship between the main tributaries of a river section. The determination of the relationship between the main tributaries of a river section is the basis for realizing the automatic classification of a river system. Therefore, the problem of automatic classification of a river system can be transformed into a binary classification problem of identifying the main tributaries of a river section. In order to avoid the shortcomings of the traditional rule-based main tributary identification method in the subjective setting of indicator weights, the present invention introduces a machine learning algorithm to acquire the ability to classify main tributaries by learning the existing river system hierarchical classification results, and applies it to the automatic classification of the river system to be classified. Random forest is an integrated learning model based on the Bagging (Bootstrap Aggregating) strategy. It is based on the decision tree and integrates the prediction values ​​of multiple decision trees to obtain the final prediction result. Random forest has the characteristics of easy implementation, low computational overhead, excellent classification performance, and is good at processing a large number of samples and features. Therefore, the random forest algorithm is used as the basis of the classification and identification model of the main tributaries of a river system.

[0059] By extracting the characteristics of six river sections from existing river system classification results and applying local normalization to these sections, the standardized section characteristics and main tributary labels were input into a random forest classification model for supervised training. After training, the trained classification model was used to identify the main tributaries of each section in the river system to be classified and to derive the probability of mainstream flow for each section.

[0060] S4: Identification and hierarchical structuring of river system mainstreams

[0061] After the classification of the main and tributary streams of a river section is completed, there may be a situation where there is no main stream upstream of the river section or there are multiple main stream sections at the same time, which is inconsistent with the objective reality. In this case, the main stream probability of each river section is compared, and the river section with the optimal value is selected as the main stream, and the remaining river sections are set as tributaries (such as Figure 4 (as shown in (a)).

[0062] After the main stream of a river system is identified, a hierarchical tree (HTree) of the river system needs to be established based on the relationship between the main and tributary streams. The HTree then generates a hierarchical code (RFCoding, random forest coding) for each river segment in the river system. The specific steps are as follows:

[0063] S4.1: Create a new HTree root node, store the river section ID of the river estuary, and start from the river section and its tree node to execute S4.2; after all the main tributaries of the river section are judged, the HTree tree is constructed (such as Figure 4 (b), execute S4.4.

[0064] S4.2: Determine the type of the upstream river section of the river section. If the upstream river section is a mainstream river section, store its ID in the tree node and continue to execute S4.2; otherwise, execute S4.3.

[0065] S4.3: Create a new HTree tree child node, which stores the river segment ID and continues to determine its upstream river segment. If the upstream river segment is the mainstream river segment, continue to execute S4.2; otherwise, execute S4.3.

[0066] S4.4: According to the hierarchical classification of the HTree tree, the river sections contained in all leaf nodes are assigned a value of 1, the river sections contained in the upper level tree nodes are assigned a value of 2, and so on; the river sections contained in the root node are assigned the highest level, and the RFCoding code of the river system is established (such as Figure 4 (c)).

[0067] Example 2

[0068] In order to verify the effectiveness and reliability of the present invention, the main tributary classification effects of different standardization methods and different machine learning methods were tested. The experimental data were selected from the river system of 10 watershed units (such as Figure 5 As shown in the figure, main tributary samples were manually marked, and 5 of them were selected as the existing river system classification results to input the classification model for training. Another 5 were used as the river systems to be classified to input the trained classification model for testing. The training data contained 2882 samples and the test data contained 4037 samples.

[0069] Experiment 1 selection Figure 5 The river section characteristics are calculated using the data from the random forest classification model. The Min-Max standardization, Z-Score standardization and local standardization methods of the present invention are then used to generate main and tributary samples. The samples are then input into the random forest classification model for training and testing (the classification results are shown in Figure 2). Figure 6As shown in the figure, the analysis shows that Min-Max normalization and Z-Score normalization perform overall normalization on the data. From the classification results, the difference between the two is small. Compared with Min-Max normalization and Z-Score normalization, two commonly used feature normalization methods, the local normalization of the present invention has the best test set classification accuracy and mainstream recognition accuracy.

[0070] like Figure 7 As shown, river system 9 in the test data was selected, and its mainstream identification and hierarchical structured results were visually analyzed: the manual classification of river system 9 only contained 5 levels, while the classification results through Min-Max standardization and Z-Score standardization contained 9 levels. The river system level increased significantly, which was obviously inconsistent with the manual classification results; secondly, from the perspective of the highest-level mainstream, the mainstream they identified did not give sufficient consideration to length priority, basin area priority, and tributary number priority. The highest-level mainstream was shorter, and the controlled basin area and tributaries were smaller. In comparison, the number of levels of local standardized division and the highest-level mainstream identification results of the present invention were consistent with the manual classification results, and the mainstream identification results of the other levels were approximately consistent, achieving better mainstream identification results.

[0071] like Figure 8 As shown in the figure, the influence of different standardization methods on the importance of each feature in the random forest model is compared. Min-Max standardization and Z-Score standardization rely more on the feature Bangle, and give insufficient consideration to the features URL, URA, and TNum, which is consistent with the conclusion of visual analysis. The local standardization of the present invention pays more attention to the features URL and URA, while taking into account the features Bangle and TNum, and gives more comprehensive consideration to the length priority, watershed area priority, tributary number priority, and direction priority.

[0072] Experiment 2 compared the effects of different machine learning algorithms on the identification results of the main tributaries of the river system. Gaussian distribution naive Bayes and decision tree were selected as comparison algorithms. The training data were all processed by the feature local normalization of the present invention. The experimental results are as follows: Figure 9 As shown in the figure, both the Gaussian naive Bayes and decision tree algorithms can achieve high test set classification accuracy, but their accuracy decreases after main tributary identification, indicating that there is a deviation in the probability of mainstream classification between the two. In comparison, the random forest method has the highest test set classification accuracy, and after main tributary identification, the main tributary identification accuracy is further improved.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An intelligent river system classification method based on local standardization of features, characterized by: The following steps are involved: S1: Based on the existing river system classification results, perform topological preprocessing and river segment feature extraction; S2: Local standardization of river section characteristics. Local standardization of river section characteristics is to divide the characteristic value of the current river section by the maximum characteristic value of the adjacent upstream river section. S3: Input the river segment characteristics obtained after local normalization into the random forest classification model for supervised learning, and apply the trained model to classify the main tributaries of the unclassified river system; S4: Identification and hierarchical structuring of river system mainstreams: Anomalies are handled based on the mainstream probability. The river system hierarchy tree is then established using the main and tributary classification results. Hierarchical codes are assigned to river sections in the river system. The operation for handling anomalies is to compare the mainstream probability of each river section, select the section with the optimal value as the mainstream, and set the remaining sections as tributaries. The operation for assigning hierarchical codes is as follows: S4.1: Create a new hierarchical tree root node, store the river section ID of the river system estuary, and start from this river section and its tree node to execute S4.2; after all the main tributaries of the river section are determined and the hierarchical tree is constructed, execute S4.4; S4.2: Determine the type of the upstream river section of the river section. If the upstream river section is a mainstream river section, store its ID in the tree node and continue to execute S4.

2. Otherwise, execute S4.3; S4.3: Create a new hierarchical tree child node, store the river segment ID in the child node, and continue to determine its upstream river segment. If the upstream river segment is the mainstream river segment, continue to execute S4.2; Otherwise, execute S4.3; S4.4: According to the hierarchical classification of the hierarchical tree, the river sections contained in all leaf nodes are assigned a value of 1, the river sections contained in the upper-level tree nodes are assigned a value of 2, and so on; the river sections contained in the root node are assigned the highest level, and the hierarchical code of the river system is established.

2. The intelligent river system classification method based on local feature standardization according to claim 1 is characterized in that: In step S1, the river section characteristics include: (1) RLen , indicating the length of the current river section; (2) URL , represents the total length of the current river section and its upstream river section; (3) WArea , indicating the area of ​​the region through which the current river section flows; (4) URA , represents the total area of ​​the region flowed by the current river section and its upstream river sections; (5) BAngle , represents the confluence angle between upstream and downstream reaches; (6) TNum , indicating the number of upstream river sections of the current river section.

3. The intelligent river system classification method based on local feature standardization according to claim 2 is characterized in that: In step S2, URL 、 URA and TNum The three types of features are locally normalized. RLen 、 WArea and BAngle The three types of features are processed using Min-Max normalization.

4. The method for intelligent river system classification based on local feature standardization according to claim 1 is characterized in that: In step S3, the application refers to using the trained model to identify the main tributaries of each river section in the river system to be classified after the training is completed, and deriving the mainstream probability of each river section.

5. An intelligent river system classification device based on local standardization of features, characterized in that: include: River section feature extraction module, which is used to extract river section features and perform local standardization of river section features based on existing river system classification results; The training prediction module is used to input the standardized river segment characteristics and main tributary labels into the random forest classification model for supervised training; the trained model is used to identify the main tributaries of each river segment in the river system to be classified and derive the mainstream probability of each river segment; The result output module is used to handle abnormal situations according to the mainstream probability, and then use the main and tributary classification results to establish a river system hierarchy tree and assign hierarchical codes to the river sections in the river system. The operation of handling abnormal situations is to compare the mainstream probability of each river section, select the river section with the optimal value as the mainstream, and set the remaining river sections as tributaries. The operation of assigning hierarchical codes is as follows: S4.1: Create a new hierarchical tree root node, store the river section ID of the river system estuary, and start from this river section and its tree node to execute S4.2; after all the main tributaries of the river section are determined and the hierarchical tree is constructed, execute S4.4; S4.2: Determine the type of the upstream river section of the river section. If the upstream river section is a mainstream river section, store its ID in the tree node and continue to execute S4.

2. Otherwise, execute S4.3; S4.3: Create a new hierarchical tree child node, store the river segment ID in the child node, and continue to determine its upstream river segment. If the upstream river segment is the mainstream river segment, continue to execute S4.2; Otherwise, execute S4.3; S4.4: According to the hierarchical classification of the hierarchical tree, the river sections contained in all leaf nodes are assigned a value of 1, the river sections contained in the upper-level tree nodes are assigned a value of 2, and so on; the river sections contained in the root node are assigned the highest level, and the hierarchical code of the river system is established.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 4 is implemented.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images