A method and system for evaluating the suitability of arable land for mechanization

Abstract

This invention discloses a method and system for evaluating the suitability of arable land for mechanization transformation. It acquires the arable land to be transformed based on multi-source data, extracts image features of the arable land, establishes a suitability evaluation system for farmland mechanization transformation, and determines the objective weights of each indicator in the evaluation system based on machine learning methods to evaluate the suitability of arable land for mechanization transformation. It fully utilizes the rich and diverse ground feature information advantages of high-resolution remote sensing imagery, and achieves an objective evaluation of the suitability of arable land for mechanization transformation based on random forest classification and information processing by geographic software. This overcomes the shortcomings of traditional land suitability evaluation methods, such as insufficient comprehensiveness in the construction of the evaluation system and non-objective determination of indicator weights, saving workload in the early planning stage of farmland mechanization transformation projects and improving the efficiency of farmland mechanization transformation work.

Description

Technical Field

[0001] This invention relates to the field of agricultural technology, specifically to a method and system for evaluating the suitability of arable land for mechanization transformation. Background Technology

[0002] Two problems exist in my country's agricultural mechanization process: First, the existing land productivity is insufficient to meet population demand, necessitating an increase in arable land. However, there is currently no additional high-quality farmland available for grain production. Hilly and mountainous areas account for two-thirds of China's land area, and arable land and crop planting area each account for one-third of the total land area. Therefore, the focus should be on converting hilly and mountainous land into arable land through various means. Second, existing agricultural machinery is incompatible with the land. The development of agricultural machinery is trending towards larger scales, without considering the specific terrain of hilly and mountainous areas. While agricultural mechanization is rapidly developing in plains areas, the level of agricultural mechanization in hilly and mountainous areas lags far behind. Clearly, the lagging level of agricultural mechanization in hilly and mountainous areas will severely restrict the overall progress of agricultural and rural modernization nationwide, becoming a bottleneck in China's agricultural mechanization development. Improving farmland to suit mechanization conditions can promote the development of agricultural mechanization in hilly and mountainous areas. However, before carrying out mechanization transformation, it is necessary to first identify the areas to be transformed. Since the conditions of land in different spatial locations are not the same, their suitability for transformation is also different. Therefore, it is necessary to conduct a suitability assessment of farmland mechanization transformation in the target area before transformation, and determine the sequence of farmland mechanization transformation in the target area based on the suitability assessment results, so that the transformation project can achieve the greatest benefits in the shortest time.

[0003] Common methods for evaluating the suitability of arable land include the analytic hierarchy process (AHP), network analysis, and ordered weighted method. Li Lingli et al., based on topographic complexity, selected three topographic limiting factors—elevation, slope, and plot fragmentation—and combined GIS spatial analysis functions and Data Envelopment Analysis (DEA) to evaluate the suitability of arable land for mechanization. Specifically, the evaluation method involved combining topographic limiting factors such as slope, elevation, and fragmentation into a topographic complexity index. Using the Delphi scoring method in conjunction with DEA, experts from several related disciplines scored the three topographic complexity factors (elevation, slope, and plot fragmentation) to obtain weights, and then calculated the comprehensive topographic complexity result. Finally, using ArcGIS 10.6 software, a natural segmentation method was employed to divide the surface complexity index of the study area into different segments to evaluate the suitability of arable land for transformation.

[0004] Current methods for evaluating the suitability of arable land mainly focus on the construction evaluation of high-standard farmland, and most of them are evaluations after construction. In farmland mechanization transformation projects, it is essential to conduct a suitability evaluation of the land before transformation. Through mechanization transformation suitability evaluation, the transformation area can be accurately identified before construction, and the focus of transformation work in each area can be determined based on the evaluation results, which can greatly reduce the workload in the planning stage.

[0005] The core of suitability assessment for farmland transformation to suitability for mechanization is establishing a sound evaluation system and determining the weights of each indicator within that system. Since suitability assessments for farmland transformation to suitability for mechanization involve multiple complex and interrelated factors such as climate conditions, soil quality, topography, hydrological environment, and socio-economic development, most current methods for evaluating farmland suitability for mechanization only select indicators based on the natural conditions of the farmland itself and supporting farmland facilities, neglecting factors such as the convenience of agricultural machinery operation and the ecological environment, even though these factors play a crucial role in facilitating efficient agricultural machinery operation, improving farmland productivity, and enhancing the farmland ecosystem. Furthermore, the weights established using the Delphi expert scoring method for determining the various evaluation indicators are primarily based on expert experience, resulting in a structure that is not objective or accurate enough. Summary of the Invention

[0006] Therefore, this invention provides a method and system for evaluating the suitability of arable land for mechanization transformation, in order to solve the above-mentioned technical problems.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] According to a first aspect of the present invention, a method for evaluating the suitability of arable land for mechanization transformation is proposed, the method comprising:

[0009] The original vector data of the spatial range of the farmland to be transformed and the high-resolution remote sensing image data were obtained and preprocessed. Constrained segmentation was performed on the high-resolution remote sensing image to construct a set of farmland objects to be transformed.

[0010] Establish an evaluation system for the suitability of farmland for mechanization transformation, and create a sample dataset. Train a random forest classification model based on the sample dataset and conduct validation tests. Obtain the weights of each evaluation index through the random forest classification model.

[0011] Based on the evaluation system and the weights of each indicator, the comprehensive evaluation results of the suitability of the cultivated land to be transformed are obtained, and the evaluation results are classified into suitable transformation levels according to the pre-constructed suitability evaluation level standards.

[0012] Furthermore, the data preprocessing specifically includes:

[0013] Geometric correction is performed on high-resolution remote sensing images, and projection conversion is carried out based on the vector data projection space of the cultivated land to be transformed.

[0014] Furthermore, the construction of the set of cultivated land objects to be transformed specifically includes:

[0015] By using the vector data of the farmland to be transformed to eliminate land use types other than farmland, the high-resolution remote sensing image is further constrained for segmentation, providing a reasonable range for subsequent feature extraction of the farmland to be transformed, and ensuring that the final segmented and extracted image data is consistent with the existing vector range of the farmland to be transformed.

[0016] Based on the characteristics of the images, high-resolution remote sensing images are segmented at multiple scales, and the corresponding parameters are continuously adjusted to obtain the optimal segmentation scale, generating image objects of farmland to be transformed. Furthermore, the vector data of the farmland to be transformed is used to determine whether the objects in the constrained segmentation results are farmland objects to be transformed, and all farmland objects to be transformed are combined into a set of farmland objects to be transformed.

[0017] Furthermore, the aforementioned evaluation system for the suitability of farmland for mechanization includes:

[0018] Evaluation indicators were constructed by selecting factors affecting the mechanization of farmland from four aspects: natural geographical conditions of arable land, convenience of agricultural machinery operation, completeness of field infrastructure, and limitations of farmland ecological environment.

[0019] Regarding the natural geographical conditions of cultivated land, field slope and field elevation are selected as evaluation indicators. The elevation data of the field can be obtained directly from the digital elevation data of the cultivated land to be transformed, and the slope of the field is calculated using the slope analysis tool of ArcGIS 10.6.

[0020] Regarding the convenience of agricultural machinery operations, the following indicators were selected: field shape index, field density, field clustering, accessibility of farm roads, and distance to agricultural machinery service stations.

[0021] The formula for calculating the field shape index is as follows:

[0022]

[0023] SI is the shape index of the field, E is the perimeter of the field, and A is the area of ​​the field.

[0024] The formula for calculating field density is:

[0025]

[0026] FD represents patch density, NP represents the number of plots, and A represents the total cultivated land area of ​​the evaluation region.

[0027] The formula for calculating the clustering degree of farmland is:

[0028]

[0029] BA represents the field clustering degree, and Pi represents the perimeter of the cultivated plot.

[0030] Accessibility of farm roads can be calculated using the buffer analysis tool in ArcGIS 10.6, and distance to agricultural machinery service stations can be obtained using the Euclidean distance tool in ArcGIS 10.6.

[0031] Regarding the completeness of field infrastructure, irrigation and drainage guarantee rate and farmland protection engineering completeness were selected as evaluation indicators. Irrigation and drainage guarantee rate refers to the ratio of the area of ​​cultivated land that can be directly irrigated to the total cultivated land area of ​​the evaluation unit. The irrigation and drainage guarantee rate can be calculated by extracting ditches from the land use data and using the buffer analysis and overlay analysis tools in ArcGIS 10.6. Farmland protection engineering completeness refers to the ratio of the area of ​​cultivated land that can be protected by farmland shelterbelts to the total cultivated land area. The farmland shelterbelts can be extracted from the land use data and the farmland protection engineering completeness can be calculated by using the buffer analysis and overlay analysis tools in ArcGIS 10.6.

[0032] Regarding the farmland ecological environment, soil organic matter content, soil pH, and vegetation coverage were selected as evaluation indicators. Soil organic matter content and soil pH were obtained from on-site soil sampling and testing. Vegetation coverage was represented by the vegetation normalized index (NDVI), which was obtained from the interpretation of high-precision remote sensing images. The calculation formula was: NDVI = ((NIR-R) / (NIR+R)), where NIR is the pixel value in the infrared band and R is the pixel value in the red band.

[0033] Furthermore, a sample dataset is established, specifically including:

[0034] Areas with a high degree of mechanization were selected as training sample areas. Several farmland attribute data with a high degree of mechanization were extracted from the sample areas using the geographic software ArcGIS 10.6 and marked as 1. Then, several farmland attribute data with a low degree of mechanization were extracted from the sample areas and marked as 0. These two types of data were merged into one dataset, which contains multiple indicator attributes and one category attribute. This dataset was divided into two parts: training samples and validation samples.

[0035] Furthermore, a random forest classification model is trained based on the sample dataset and validated. The weights of each evaluation metric are obtained through the random forest classification model, specifically including:

[0036] The random forest classifier is selected for learning and classification. The learning process of the random forest classification model includes:

[0037] Step 1: If there are N samples, randomly select N samples with replacement, that is, randomly select one sample each time, then return and continue to select. The N selected samples are used to train a decision tree, which serves as the sample at the root node of the decision tree.

[0038] Step 2: When each sample has M attributes, when each node of the decision tree needs to be split, randomly select m attributes from these M attributes, satisfying the condition m << M, and then select 1 attribute from these m attributes using a certain strategy as the splitting attribute of this node.

[0039] Step 3: Each node in the process of forming the decision tree should be split according to Step 2 until it cannot be split anymore. No pruning is performed during the entire process of forming the decision tree.

[0040] Step 4: For the dataset to be classified, after the decision of each tree, the final classification result is determined according to the classification with the highest number of votes obtained in the decision.

[0041] Furthermore, based on the sample dataset, the random forest classification model is trained and verified. The weights of each evaluation index are obtained through the random forest classification model, specifically including:

[0042] In the random forest, the weight of a feature is calculated according to the Gini coefficient. Suppose the set T contains k classifications, and the Gini index is calculated as:

[0043]

[0044] where P j represents the frequency of the occurrence of class j;

[0045] If the set T is divided into n parts Ti, i = 1, 2,..., m, then in order to calculate the Gini index, calculate the Gini index of the variable xi used for splitting at each splitting node. The calculation formula for the Gini index of this split is:

[0046]

[0047] where, N i is the number of samples at the child node T i ; N is the number of samples at the parent node T;

[0048] The average decrease in the Gini index of each variable of all the trees in the forest is used to estimate the importance of the variable. Therefore, the weight of the feature is:

[0049]

[0050] Among them, D j It represents the importance of the j-th feature.

[0051] Furthermore, the suitability evaluation levels for the modification include unsuitable, marginally suitable, suitable, and very suitable.

[0052] According to a second aspect of the present invention, a suitability evaluation system for arable land adapted for mechanization is proposed, the system comprising:

[0053] The module for constructing the set of cultivated land to be transformed is used to acquire the original spatial range vector data and high-resolution remote sensing image data of the cultivated land to be transformed, and to perform data preprocessing, constrain the segmentation of the high-resolution remote sensing image, and construct the set of cultivated land to be transformed.

[0054] The evaluation system establishment module is used to establish an evaluation system for the suitability of farmland for mechanization transformation, and to establish a sample dataset. Based on the sample dataset, a random forest classification model is trained and validated. The weights of each evaluation index are obtained through the random forest classification model.

[0055] The suitability evaluation module is used to obtain the comprehensive evaluation results of the suitability of the cultivated land to be transformed based on the evaluation system and the weight of each indicator, and to classify the suitability level of the evaluation results according to the pre-constructed suitability evaluation level standard.

[0056] According to a third aspect of the present invention, a computer storage medium is provided, the computer storage medium containing one or more program instructions, the one or more program instructions being configured to be executed by a farmland suitability evaluation system for mechanization-oriented transformation as described in any of the preceding methods.

[0057] The present invention has the following advantages:

[0058] This invention proposes a method and system for evaluating the suitability of arable land for mechanization transformation. It fully utilizes the rich and diverse ground feature information of high-resolution remote sensing images and achieves an objective evaluation of the suitability of arable land for mechanization transformation based on random forest classification and information processing by geographic software. It can overcome the shortcomings of traditional land suitability evaluation methods, such as insufficient comprehensiveness in the construction of the evaluation system and non-objective determination of indicator weights. It saves workload in the early planning stage of farmland mechanization transformation projects and improves the efficiency of farmland mechanization transformation work. Attached Figure Description

[0059] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0060] Figure 1 This is a flowchart illustrating a method for evaluating the suitability of arable land for mechanization transformation, as provided in an embodiment of the present invention.

[0061] Figure 2 This is a schematic diagram illustrating the specific implementation process of a method for evaluating the suitability of arable land for mechanization transformation, as provided in an embodiment of the present invention.

[0062] Figure 3 The flowchart of random forest classification in a method for evaluating the suitability of arable land for mechanization transformation provided in an embodiment of the present invention. Detailed Implementation

[0063] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] like Figure 1 As shown in the figure, this invention proposes a method for evaluating the suitability of arable land for mechanization transformation, the method comprising:

[0065] S100. Obtain the original vector data of the spatial range of the cultivated land to be transformed and the high-resolution remote sensing image data, and perform data preprocessing. Perform constrained segmentation on the high-resolution remote sensing image to construct the set of cultivated land objects to be transformed.

[0066] S200. Establish an evaluation system for the suitability of farmland for mechanization transformation, and establish a sample dataset. Train a random forest classification model based on the sample dataset and conduct validation tests. Obtain the weights of each evaluation index through the random forest classification model.

[0067] S300. Based on the evaluation system and the weight of each indicator, obtain the comprehensive evaluation result of the suitability of the cultivated land to be transformed, and classify the suitability of the evaluation result into a transformation suitability level according to the pre-constructed transformation suitability evaluation level standard.

[0068] This invention proposes a method for evaluating the suitability of farmland for mechanization. Based on multi-source data, it acquires the farmland to be transformed, extracts image features of the farmland, establishes a suitability evaluation system for farmland mechanization, and determines the objective weights of each indicator in the evaluation system using machine learning methods to evaluate the suitability of farmland for mechanization. Figure 2 As shown, the specific content is as follows:

[0069] 1. Acquisition of farmland to be transformed

[0070] (1) Data preprocessing

[0071] The initial relevant data includes vector data of the spatial extent of the farmland to be transformed and high-resolution remote sensing imagery. In order to accurately evaluate the plots using the imagery features of the farmland to be transformed, the resolution of the high-resolution remote sensing imagery used should be within 2 meters, and the accuracy of the vector data of the spatial extent of the farmland to be transformed should be within this range.

[0072] Data preprocessing mainly involves geometric correction of high-resolution remote sensing images and projection transformation based on the projection space of the vector data of the farmland to be transformed. Satellite remote sensing images, aerial remote sensing images, or UAV images all exhibit various geometric distortions during acquisition. When applying the vector data of the farmland to be transformed, high-resolution images require geometric correction and projection transformation. To ensure spatial consistency with the vector data of the farmland to be transformed, the geometric correction accuracy must not be lower than the accuracy of the vector data of the farmland to be transformed, and must not exceed 2 pixels. Specific correction and projection transformation methods can be found in relevant literature on remote sensing image processing or software user manuals; these will not be elaborated upon in this specification.

[0073] (2) Construction of the set of cultivated land to be transformed

[0074] This invention uses vector data of the arable land to be transformed and high-resolution remote sensing image data to obtain the arable land objects to be transformed through remote sensing image segmentation. Land use types other than arable land are eliminated using the vector data of the arable land to be transformed, further constraining the segmentation of the high-resolution remote sensing image. This provides a reasonable range for subsequent feature extraction of the arable land objects to be transformed, and ensures that the final segmented image data is consistent with the existing vector range of the arable land to be transformed. Based on the image characteristics, multi-scale segmentation is performed on the high-resolution remote sensing image, and the corresponding parameters are continuously adjusted to obtain the optimal segmentation scale, generating the image objects of the arable land to be transformed. The vector data of the arable land to be transformed is then used to determine whether the objects in the constrained segmentation results are indeed arable land objects to be transformed, and all arable land objects to be transformed are combined into a set of arable land objects to be transformed.

[0075] 2. Evaluation system construction and corresponding farmland information acquisition

[0076] Evaluation indicators were constructed by selecting factors affecting the mechanization of farmland from four aspects: natural geographical conditions of arable land, convenience of agricultural machinery operation, completeness of field infrastructure, and limitations of farmland ecological environment.

[0077] Regarding the natural geographical conditions of cultivated land, field slope and field elevation are selected as evaluation indicators. The elevation data of the field can be obtained directly from the digital elevation data of the cultivated land to be transformed, and the slope of the field is calculated using the slope analysis tool of ArcGIS 10.6.

[0078] Regarding the convenience of agricultural machinery operations, the following indicators were selected: field shape index, field density, field clustering, accessibility of farm roads, and distance to agricultural machinery service stations.

[0079] The formula for calculating the field shape index is as follows:

[0080]

[0081] SI is the shape index of the field, E is the perimeter of the field, and A is the area of ​​the field.

[0082] The formula for calculating field density is:

[0083]

[0084] FD represents patch density, NP represents the number of plots, and A represents the total cultivated land area of ​​the evaluation region.

[0085] The formula for calculating the clustering degree of farmland is:

[0086]

[0087] BA represents the field clustering degree, and Pi represents the perimeter of the cultivated plot.

[0088] Accessibility of farm roads can be calculated using the buffer analysis tool in ArcGIS 10.6, and distance to agricultural machinery service stations can be obtained using the Euclidean distance tool in ArcGIS 10.6.

[0089] Regarding the completeness of field infrastructure, irrigation and drainage guarantee rate and farmland protection engineering completeness were selected as evaluation indicators. Irrigation and drainage guarantee rate refers to the ratio of the area of ​​cultivated land that can be directly irrigated to the total cultivated land area of ​​the evaluation unit. The irrigation and drainage guarantee rate can be calculated by extracting ditches from the land use data and using the buffer analysis and overlay analysis tools in ArcGIS 10.6. Farmland protection engineering completeness refers to the ratio of the area of ​​cultivated land that can be protected by farmland shelterbelts to the total cultivated land area. The farmland shelterbelts can be extracted from the land use data and the farmland protection engineering completeness can be calculated by using the buffer analysis and overlay analysis tools in ArcGIS 10.6.

[0090] Regarding the farmland ecological environment, soil organic matter content, soil pH, and vegetation coverage were selected as evaluation indicators. Soil organic matter content and soil pH were obtained from on-site soil sampling and testing. Soil sampling methods can be based on the requirements of relevant national standards or the soil sampling methods used in the literature, which will not be elaborated here. Vegetation coverage is represented by the normalized vegetation index (NDVI), which is obtained from the interpretation of high-precision remote sensing images. The calculation formula is: NDVI = ((NIR-R) / (NIR+R)), where NIR is the pixel value of the infrared band and R is the pixel value of the red band.

[0091] 3. Determining Indicator Weights

[0092] A sample area was selected, and attribute data of the sample plots was extracted using ArcGIS 10.6 to form a dataset. A random forest model was trained and validated to obtain the importance of each evaluation indicator, which was then converted into the weights of each indicator. This process includes the following three steps:

[0093] (1) Establishment of sample dataset

[0094] Areas with high mechanization suitability were selected as training sample areas. Using ArcGIS 10.6, several farmland attribute data points with high mechanization suitability were extracted from these areas and labeled as 1. Then, several farmland attribute data points with low mechanization suitability were extracted and labeled as 0. Note that the number of samples with high mechanization suitability should be the same as the number of samples with low mechanization suitability. These two types of data were merged into a single dataset containing 12 indicator attributes and 1 categorical attribute. This dataset was then divided into training and validation samples, with the training samples comprising 70% of the total sample set and the validation samples comprising 30%. The training samples were used for feature selection and random forest model building, while the validation samples were used for accuracy evaluation.

[0095] (2) Feature Importance Calculation

[0096] As an effective prediction tool, the random forest algorithm has the characteristics of high efficiency, flexibility, accuracy, and strong selection ability. Therefore, it is appropriate to select this algorithm to calculate the index weights for the suitability of farmland mechanization transformation. Random forest is an ensemble learning algorithm based on the tree decision-making process. This is a supervised learning method, and random forest can be used for classification and regression. The calculation of the weights of evaluation indicators is realized based on its classification function. In the multiple decision trees in the random forest, for each decision tree, when each sample is input, a classification result will be obtained. The random forest algorithm can handle high-dimensional data. And due to the randomness of feature subset selection, random forest classification does not require feature selection and fully ensures the independence between each tree. Compared with other algorithms, the random forest has a faster training speed, is easy to implement parallel computing, and can detect the mutual influence between features. It can still maintain accuracy when applied to datasets with missing features. As Figure 3 shown, the present invention selects a random forest classifier for learning and classification, constructs a random forest classification model based on the sample dataset, and calculates the importance of each evaluation indicator. The learning process of the random forest classification model is divided into the following steps:

[0097] Step 1: If there are N samples, randomly select N samples with replacement (each time randomly select a sample and then return to continue selection). The selected N samples are used to train a decision tree, which serves as the sample at the root node of the decision tree.

[0098] Step 2: When each sample has M attributes, when each node of the decision tree needs to be split, randomly select m attributes from these M attributes, satisfying the condition m << M. Then, select 1 attribute from these m attributes using a certain strategy as the splitting attribute of this node.

[0099] Step 3: Each node in the process of forming the decision tree should be split according to Step 2. Keep splitting until it cannot be split anymore. Note that no pruning is performed during the entire process of forming the decision tree.

[0100] Step 4: For the dataset to be classified, after the decision of each tree, determine the final classification result according to the classification with the highest number of votes obtained in the decision.

[0101] (3) Calculation of the weights of each indicator

[0102] There are many features in the dataset. In the random forest, the weight of a feature is calculated based on the Gini coefficient. Assume that the set T contains k classifications, and the Gini index is calculated as:

[0103] [[ID=二十五]]

[0104] If a set T is divided into n parts Ti (i = 1, 2, ..., m), then to calculate the Gini index, the Gini index of the variable xi used for splitting at each split node is calculated. The formula for calculating the Gini index of this split is:

[0105]

[0106] Generally, the average Gini index decrease for each variable across all trees in a forest is often used to estimate the importance of the variable. Therefore, the weights of the features are:

[0107]

[0108] 4. Modify the zoning

[0109] Based on the comprehensive evaluation results of the suitability of the target arable land for transformation, the suitability was divided into four levels according to the natural discontinuity method. To unify the indicator dimensions, a graded scoring method was used for assignment, where unsuitable = 25 points, marginally suitable = 50 points, suitable = 75 points, and very suitable = 100 points. The regional division results were visualized using the geographic software ArcGIS 10.6, which can intuitively show the suitability of the target area for arable land transformation. Based on the visualization results, the priority order for carrying out farmland mechanization transformation work was determined, and the transformation focus of different areas was identified.

[0110] This invention proposes a method for evaluating the suitability of arable land for mechanization transformation. It fully utilizes the rich and diverse ground feature information of high-resolution remote sensing imagery, and achieves an objective evaluation of arable land suitability based on random forest classification and geographic software information processing. The evaluation system fully considers various influencing factors, including economic and social factors and the ecological environment affecting farmland mechanization transformation, making the evaluation system more comprehensive. Machine learning methods are used to determine the weights of indicators, employing the random forest algorithm driven by a large amount of data to calculate the objective weights of each indicator. This overcomes the shortcomings of traditional land suitability evaluation methods, such as insufficient comprehensiveness in evaluation system construction and non-objective determination of indicator weights. It saves workload in the early planning stage of farmland mechanization transformation projects and improves the efficiency of farmland mechanization transformation work.

[0111] Corresponding to the above embodiments, this invention proposes a farmland suitability evaluation system for mechanization-oriented transformation, the system comprising:

[0112] The module for constructing the set of cultivated land to be transformed is used to acquire the original spatial range vector data and high-resolution remote sensing image data of the cultivated land to be transformed, and to perform data preprocessing, constrain the segmentation of the high-resolution remote sensing image, and construct the set of cultivated land to be transformed.

[0113] The evaluation system establishment module is used to establish an evaluation system for the suitability of farmland for mechanization transformation, and to establish a sample dataset. Based on the sample dataset, a random forest classification model is trained and validated. The weights of each evaluation index are obtained through the random forest classification model.

[0114] The suitability evaluation module is used to obtain the comprehensive evaluation results of the suitability of the cultivated land to be transformed based on the evaluation system and the weight of each indicator, and to classify the suitability level of the evaluation results according to the pre-constructed suitability evaluation level standard.

[0115] The functions performed by each component in the farmland suitability evaluation system for mechanization transformation provided in this embodiment of the invention have been described in detail in the above embodiments, so they will not be repeated here.

[0116] Corresponding to the above embodiments, this embodiment proposes a computer storage medium containing one or more program instructions, which are used by a farmland suitability evaluation system for mechanization-oriented transformation to execute the method of the above embodiments.

[0117] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for evaluating the suitability of arable land for mechanization transformation, characterized in that, The method includes: The process involves acquiring original vector data of the spatial extent of the farmland to be transformed and high-resolution remote sensing image data, performing data preprocessing, constrained segmentation of the high-resolution remote sensing image, and constructing a set of farmland objects to be transformed. Specifically, this construction includes: using the vector data of the farmland to be transformed to remove land use types other than farmland, further constraining the segmentation of the high-resolution remote sensing image to provide a reasonable range for subsequent feature extraction of the farmland objects to be transformed, and ensuring that the final segmented and extracted image data is consistent with the existing vector extent of the farmland to be transformed; performing multi-scale segmentation of the high-resolution remote sensing image based on the image characteristics, continuously adjusting the corresponding parameters to obtain the optimal segmentation scale, and generating image objects of the farmland to be transformed; further using the vector data of the farmland to be transformed to determine whether the objects in the constrained segmentation results are farmland objects to be transformed, and constructing a set of farmland objects to be transformed from all the farmland objects to be transformed. An evaluation system for the suitability of farmland for mechanization was established, and a sample dataset was created. A random forest classification model was trained based on the sample dataset and validated. The weights of each evaluation index for assessing the suitability of farmland for mechanization were obtained through the random forest classification model. Specifically, the evaluation system for the suitability of farmland for mechanization includes: selecting factors influencing farmland for mechanization from four aspects: natural geographical conditions of arable land, convenience of agricultural machinery operation, completeness of field infrastructure, and limitations of the farmland ecological environment to construct evaluation indicators. Regarding the convenience of agricultural machinery operation, the field shape index, field density, field clustering, accessibility of farm roads, and distance to agricultural machinery service stations were selected as evaluation indicators. The sample dataset specifically includes: selecting areas with high mechanization suitability as training sample areas, using the geographic software ArcGIS 10.6 to extract several farmland attribute data with high mechanization suitability in the sample areas, and marking them as 1; then extracting several farmland attribute data with low mechanization suitability in the sample areas, and marking them as 0; merging these two types of data into one dataset, which contains multiple indicator attributes and one category attribute, and dividing this dataset into two parts: training samples and validation samples; Based on the evaluation system and the weights of each indicator, the comprehensive evaluation results of the suitability of the cultivated land to be transformed are obtained. According to the pre-constructed evaluation level standard for suitability of transformation, the evaluation results are classified into levels of suitability for transformation. The results of regional division are visualized using the geographic software ArcGIS 10.

6. Based on the visualization results, the priority order of carrying out farmland mechanization transformation is determined, and the transformation focus of different regions is determined.

2. The method for evaluating the suitability of arable land for mechanization transformation according to claim 1, characterized in that, The data preprocessing specifically includes: Geometric correction is performed on high-resolution remote sensing images, and projection conversion is carried out based on the vector data projection space of the cultivated land to be transformed.

3. The method for evaluating the suitability of arable land for mechanization transformation according to claim 1, characterized in that, The aforementioned evaluation system for the suitability of farmland for mechanization includes the following indicators: Regarding the natural geographical conditions of cultivated land, field slope and field elevation were selected as evaluation indicators. The field elevation data was obtained directly from the digital elevation data of the cultivated land to be transformed, and the field slope was calculated using the slope analysis tool of ArcGIS 10.

6. The formula for calculating the field shape index is as follows: ； SI is the shape index of the field, E is the perimeter of the field, and A is the area of ​​the field. The formula for calculating field density is: ； FD represents patch density, NP represents the number of plots, and A represents the total cultivated land area of ​​the evaluation region; The formula for calculating the clustering degree of farmland is: ； BA represents the clustering degree of the fields, and Pi represents the perimeter of the cultivated plot. Accessibility of farm roads was calculated using the buffer analysis tool in ArcGIS 10.6, and distance to agricultural machinery service stations was obtained using the Euclidean distance tool in ArcGIS 10.

6. Regarding the completeness of field infrastructure, irrigation and drainage guarantee rate and farmland protection engineering completeness were selected as evaluation indicators. Irrigation and drainage guarantee rate refers to the ratio of cultivated land area that can be directly irrigated to the total cultivated land area of ​​the evaluation unit. The irrigation and drainage guarantee rate can be calculated by extracting ditches from the land use data and using the buffer analysis and overlay analysis tools in ArcGIS 10.

6. Farmland protection engineering completeness refers to the ratio of cultivated land area that can be protected by farmland shelterbelts to the total cultivated land area. The farmland shelterbelts can be extracted from the land use data and the farmland protection engineering completeness can be calculated by using the buffer analysis and overlay analysis tools in ArcGIS 10.

6. Regarding the farmland ecological environment, soil organic matter content, soil pH, and vegetation coverage were selected as evaluation indicators. Soil organic matter content and soil pH were obtained from on-site soil sampling and testing. Vegetation coverage was represented by the vegetation normalized index (NDVI), which was obtained from the interpretation of high-precision remote sensing images. The calculation formula is: NDVI = ((NIR - R) / ( NIR + R)), where NIR is the pixel value in the infrared band and R is the pixel value in the red band.

4. The method for evaluating the suitability of arable land for mechanization transformation according to claim 1, characterized in that, A random forest classification model was trained based on a sample dataset and validated. The weights of each evaluation metric were obtained through the random forest classification model, including: The random forest classifier is selected for learning classification. The learning process of the random forest classification model includes: Step 1: If there are N samples, then randomly select N samples with replacement. That is, randomly select one sample each time, and then return to continue selecting. The selected N samples are used to train a decision tree and serve as the samples at the root node of the decision tree. Step 2: When each sample has M attributes, when each node of the decision tree needs to split, randomly select m attributes from these M attributes, satisfying the condition m << M. Then, use a certain strategy to select one attribute from these m attributes as the splitting attribute for that node. Step 3: During the formation of the decision tree, each node must be split according to Step 2 until it can no longer be split. No pruning is performed during the entire decision tree formation process. Step 4: For the dataset to be classified, after each tree's decision, determine the final classification result based on the classification that receives the highest number of votes.

5. The method for evaluating the suitability of arable land for mechanization transformation according to claim 1, characterized in that, A random forest classification model was trained based on a sample dataset and validated. The weights of each evaluation metric were obtained through the random forest classification model, including: In random forests, feature weights are calculated based on the Gini coefficient. Assuming a set T contains k categories, the Gini coefficient is calculated as follows: ； Where P j This indicates the frequency of occurrence of category j; If we divide the set T into n parts Ti, i = 1, 2, ..., m, then to calculate the Gini index, we calculate the Gini index of the variable xi used for splitting at each split node. The formula for calculating the Gini index of this split is: ； Where, N i It is in child node T i The number of samples at point T; N is the number of samples at the parent node T. The average Gini index decrease for each variable across all trees in the forest is used to estimate the importance of the variable; therefore, the weights of the features are: ； Among them, D j It represents the importance of the j-th feature.

6. The method for evaluating the suitability of arable land for mechanization transformation according to claim 1, characterized in that, The suitability evaluation levels for the renovation include unsuitable, marginally suitable, suitable, and very suitable.

7. A suitability evaluation system for arable land adapted for mechanization, characterized in that, The system includes: The module for constructing a set of farmland to be transformed is used to acquire original vector data of the spatial extent of farmland to be transformed and high-resolution remote sensing image data, and to perform data preprocessing, constrained segmentation of the high-resolution remote sensing image to construct a set of farmland to be transformed objects. Specifically, constructing the set of farmland to be transformed objects includes: using the vector data of the farmland to be transformed to remove land use types other than farmland, further constraining the segmentation of the high-resolution remote sensing image to provide a reasonable range for subsequent feature extraction of farmland to be transformed objects, and ensuring that the final segmented and extracted image data is consistent with the existing vector range of farmland to be transformed; performing multi-scale segmentation of the high-resolution remote sensing image based on image characteristics, continuously adjusting the corresponding parameters to obtain the optimal segmentation scale, and generating image objects of farmland to be transformed; further using the vector data of the farmland to be transformed to determine whether the objects in the constrained segmentation results are farmland to be transformed objects, and constructing a set of farmland to be transformed objects from all the farmland to be transformed objects. The evaluation system establishment module is used to establish an evaluation system of indicators for the suitability of farmland for mechanization transformation, and to establish a sample dataset. A random forest classification model is trained based on the sample dataset and validated. The weights of each evaluation indicator are obtained through the random forest classification model. Specifically, the evaluation system for the suitability of farmland for mechanization transformation includes: selecting factors influencing farmland mechanization transformation from four aspects: natural geographical conditions of arable land, convenience of agricultural machinery operation, completeness of field infrastructure, and limitations of the farmland ecological environment to construct evaluation indicators. Regarding the convenience of agricultural machinery operation, the field shape index, field density, field clustering, accessibility of farm roads, and distance to agricultural machinery service stations are selected as evaluation indicators. The sample dataset specifically includes: selecting areas with high mechanization suitability as training sample areas, and using the geographic software ArcGIS 10.

6. Extract several farmland attribute data with high mechanization suitability from the sample area and label them as 1; then extract several farmland attribute data with low mechanization suitability from the sample area and label them as 0; merge these two types of data into one dataset, which contains multiple indicator attributes and one category attribute, and divide this dataset into two parts: training samples and validation samples. The suitability evaluation module is used to obtain the comprehensive evaluation results of the suitability of the cultivated land to be transformed based on the evaluation system and the weight of each indicator, and to classify the suitability of the evaluation results into levels according to the pre-constructed suitability evaluation level standards. The results of the regional division are visualized using the geographic software ArcGIS 10.

6. Based on the visualization results, the priority order of carrying out farmland mechanization transformation work is determined, and the transformation focus of different regions is determined.

8. A computer storage medium, characterized in that, The computer storage medium contains one or more program instructions, which are used by a farmland suitability evaluation system for mechanization-oriented transformation to execute the method described in any one of claims 1-6.

Images