Method and Device for Selecting Threshold for Extracting Farmland Roads Based on Data Mining

Abstract

The present invention discloses a method for selecting a threshold for extracting farmland roads based on data mining. The selection method includes: collecting farmland information through an image acquisition device to obtain a farmland information image; segmenting and dividing the pixel value range of the farmland information image, and obtaining the distribution data of farmland ground objects for each segment by mining segment by segment; summarizing the distribution data of farmland ground objects for each segment to obtain the distribution probability and pixel value range of farmland roads; and performing optimization processing according to the distribution probability and pixel value range of farmland roads to select a threshold for extracting farmland roads. After processing the image, the method performs segmented division, performs data mining segment by segment, and then conducts centralized analysis and processing through global summarization, obtains the distribution range and distribution probability of farmland road pixel values, and optimizes and selects the extraction threshold according to the distribution range and distribution probability of farmland roads, which can improve the accuracy of farmland road extraction and meet the requirements for rapid and automatic detection of large-scale farmland roads.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of land monitoring, and particularly relates to a method and device for selecting a threshold for extracting farmland roads based on data mining. Background Art

[0002] The construction of modern farmland requires the use of informatization means for monitoring and efficient construction management based on farmland information. Farmland road information is one of the important indicators in farmland construction projects, and accurately extracting farmland road information from farmland feature information is an important indicator for realizing large-scale rapid automatic detection of farmland roads.

[0003] Currently, the extraction of farmland road information mainly uses traditional image processing methods for information extraction, and it is necessary to set an extraction threshold for extracting farmland roads. However, the threshold setting is usually based on the habits and experience of operators, resulting in large errors, low accuracy of the farmland road extraction results, and general applicability, which cannot meet the requirements of large-scale rapid automatic detection of farmland roads. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art. The present invention provides a method and device for selecting a threshold for extracting farmland roads based on data mining. The method processes and segmentally divides the image, performs data mining on each segment, and then conducts centralized analysis and processing through global summarization to obtain the distribution range and distribution probability of farmland road pixel values, and optimally selects the extraction threshold according to the farmland road distribution range and distribution probability, which can improve the accuracy of farmland road extraction and meet the requirements of large-scale rapid automatic detection of farmland roads.

[0005] The present invention provides a method for selecting a threshold for farmland roads based on data mining, and the selection method includes:

[0006] Collecting farmland information through an image acquisition device to obtain a farmland information image;

[0007] Segmenting and dividing the pixel value range of the farmland information image, and obtaining the distribution data of farmland features for each segment through segment-by-segment mining;

[0008] Summarizing the distribution data of farmland features for each segment to obtain the distribution probability and pixel value range of farmland roads;

[0009] Performing optimization processing according to the distribution probability and pixel value range of farmland roads, and selecting a threshold for extracting farmland roads.

[0010] Further, the collecting farmland information through an image acquisition device to obtain a farmland information image includes:

[0011] Collecting farmland information by shooting with a drone to obtain the farmland information image.

[0012] Further, segment the pixel value range of the preprocessed farmland information image, and obtain the segmented farmland feature distribution data by segment-by-segment mining, including:

[0013] Preprocess the farmland information image to obtain the preprocessed farmland information image;

[0014] Segment the preprocessed farmland information image to obtain several segments of pixel value ranges;

[0015] Perform segment-by-segment mining on the several segments of pixel value ranges to obtain the segmented farmland feature distribution data.

[0016] Further, the preprocessing of the farmland information image to obtain the preprocessed farmland information image includes:

[0017] Use image processing software to perform grayscale processing on the farmland information image to obtain the preprocessed farmland information image.

[0018] Further, segmenting the preprocessed farmland information image to obtain several segments of pixel value ranges includes:

[0019] Segment the pixel set of the preprocessed farmland information image at a set step size to obtain several pixel value ranges with equal step sizes.

[0020] Further, the segment-by-segment mining of the several segments of pixel value ranges to obtain the segmented farmland feature distribution data further includes:

[0021] Input the grayscale value of each pixel point into each segment of the pixel value range in turn to obtain the feature distribution of each segment of the pixel value range, so as to obtain the segmented farmland feature distribution data.

[0022] Further, summarizing the segmented farmland feature distribution data to obtain the farmland road distribution probability and pixel value range includes:

[0023] Summarize and analyze the segmented farmland feature distribution data to obtain the pixel value range and distribution probability of the farmland features, so as to extract the pixel value range and distribution probability of the farmland roads.

[0024] Further, the optimization process includes optimizing the extraction threshold of the road with clear boundaries;

[0025] When the boundary between the farmland road distribution and the distribution ranges of other features is clear, select the boundary pixel value as the extraction threshold of the farmland road.

[0026] Further, the optimization process includes optimizing the extraction threshold of the road with overlapping boundaries;

[0027] When the distribution of the farmland roads intersects with the boundary of the distribution ranges of other ground objects, select the extraction threshold of the farmland roads within the pixel value range with a high probability of farmland road distribution according to the probability of farmland road distribution.

[0028] The present invention also provides a device for selecting an extraction threshold of farmland roads based on data mining. The selection device includes:

[0029] Information acquisition module: Collect farmland information through an image acquisition device to obtain a farmland information image;

[0030] Segment-by-segment mining module: Segment and divide the pixel value range of the farmland information image, and obtain segment-by-segment farmland ground object distribution data through segment-by-segment mining;

[0031] Summary and analysis module: Summarize the segment-by-segment farmland ground object distribution data to obtain the distribution probability of farmland roads;

[0032] Optimization processing module: Perform optimization processing according to the distribution probability of the farmland roads, and select the extraction threshold of the farmland roads.

[0033] The present invention provides a method and a device for selecting an extraction threshold of farmland roads based on data mining. The method performs gray-scale processing on the image, segments and divides the pixel set of the image after extraction, performs data mining segment by segment to obtain segment-by-segment farmland ground object distribution data, and through summarizing and centrally analyzing the segment-by-segment farmland ground object distribution data, obtains the distribution range and distribution probability of the pixel values of the farmland roads, and optimizes and selects the extraction threshold according to the distribution range and distribution probability of the farmland roads, which can improve the accuracy of farmland road extraction and meet the requirements of large-scale rapid automatic detection of farmland roads. Description of the Drawings

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0035] Figure 1 is a flowchart of the method for selecting an extraction threshold of farmland roads based on data mining in an embodiment of the present invention;

[0036] Figure 2 is a flowchart of the method for segment-by-segment mining of farmland road information in an embodiment of the present invention;

[0037] Figure 3 is a flowchart of the device for selecting an extraction threshold of farmland roads based on data mining in an embodiment of the present invention;

[0038] Figure 4 It is a schematic diagram of the distribution data of farmland features with pixel values ranging from 110 to 120 in the embodiments of the present invention;

[0039] Figure 5 It is a schematic diagram of the extraction result with a threshold of 140 in the embodiments of the present invention;

[0040] Figure 6 It is a schematic diagram of the extraction result with a threshold of 110 in the embodiments of the present invention. Detailed implementation manners

[0041] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0042] Embodiment 1:

[0043] Figure 1 It shows a flowchart of a method for selecting a threshold for extracting farmland roads based on data mining in the embodiments of the present invention. The method for selecting a threshold for extracting farmland roads based on data mining includes:

[0044] S11: Collect farmland information through an image acquisition device to obtain a farmland information image;

[0045] Specifically, the step of collecting farmland information through an image acquisition device to obtain a farmland information image includes:

[0046] Collect farmland information by using a drone to take pictures to obtain the farmland information image. Drone technology uses radio remote control equipment and self - contained program control devices to operate unmanned aircraft, or is completely or intermittently autonomously operated by an on - vehicle computer, and can adapt to various scenarios to perform flight tasks. By mounting a shooting device on the drone and using the drone to take pictures over the farmland, the monitoring efficiency of farmland projects can be improved.

[0047] S12: Segment the pixel value range of the farmland information image, and obtain the distribution data of farmland features in segments by segment - by - segment mining;

[0048] Specifically, the distribution data of farmland features is the distribution data of farmland feature pixel values, that is, it reflects the distribution of farmland feature pixel values in different pixel value ranges of the farmland information image.

[0049] Specifically, Figure 2 It shows a flowchart of a method for segment - by - segment mining of farmland road information in the embodiments of the present invention. The method includes:

[0050] S21: Preprocess the farmland information image to obtain the preprocessed farmland information image;

[0051] Specifically, use image processing software to perform grayscale processing on the farmland information image to obtain the preprocessed farmland information image.

[0052] Furthermore, select appropriate image processing software in the cross-platform computer vision and machine learning software library (Opencv) to perform grayscale processing on the farmland information image for subsequent analysis and processing.

[0053] S22: Segment the preprocessed farmland information image to obtain several pixel value ranges.

[0054] Specifically, segment the preprocessed farmland information image to obtain several pixel value ranges, including:

[0055] Segment the pixel set of the preprocessed farmland information image at a set step size to obtain several pixel value ranges with equal step sizes;

[0056] S23: Dig each of the several pixel value ranges one by one to obtain the farmland road distribution data for each segment;

[0057] Specifically, input the grayscale value of each pixel point in each pixel value range in turn to obtain the ground object distribution of each pixel value range, thereby obtaining the farmland road distribution data for each segment.

[0058] Furthermore, input the grayscale value of each pixel point in each pixel value range in turn to determine whether the pixel point is in the segmented pixel value range. If the pixel point is distributed in the segmented pixel value range, display the pixel point with a black dot; otherwise, keep the original color of the pixel point. Display the result image after segment-by-segment mining, thereby obtaining the ground objects represented by the black dot distribution, extracting the farmland road information in the analysis result image, and obtaining the farmland ground object distribution data for each segment.

[0059] S13: Aggregate the farmland ground object distribution data for each segment to obtain the farmland road distribution probability.

[0060] Specifically, aggregate the farmland ground object distribution data for each segment, compare and analyze the probabilities of farmland ground objects in the farmland road data for each segment, thereby obtaining the pixel value range and distribution probability of each farmland ground object, and extracting the pixel value range and distribution probability of the farmland road.

[0061] S14: Perform optimization processing according to the farmland road distribution probability and select the farmland road extraction threshold.

[0062] Specifically, the optimization processing includes optimizing the road extraction threshold with clear boundaries;

[0063] When the boundary between the distribution of the farmland roads and the distribution areas of other ground features is clear, select the boundary pixel value as the extraction threshold of the farmland roads.

[0064] Furthermore, the optimization processing includes the optimization of the extraction threshold for roads with interlaced boundaries.

[0065] When the boundary between the distribution of the farmland roads and the distribution of other ground features is interlaced, select the extraction threshold of the farmland roads within the pixel value range with a high probability of farmland road distribution according to the probability of farmland road distribution.

[0066] The embodiment of the present invention provides a method for selecting the extraction threshold of farmland roads based on data mining. The method performs gray-scale processing on the collected farmland information image, segments the processed image, performs data mining on each segment, obtains the farmland road information data for each segment, and then centrally summarizes to obtain the probability of farmland road distribution and the pixel range. The threshold is extracted through the probability of farmland road distribution and the pixel range, improving the accuracy of threshold extraction and realizing the rapid and automatic detection of farmland roads.

[0067] Embodiment 2:

[0068] Figure 3 The structural schematic diagram of the device for selecting the extraction threshold of farmland roads based on data mining in the embodiment of the present invention is shown. The device includes:

[0069] Information acquisition module 1: Collect farmland information through an image acquisition device to obtain a farmland information image.

[0070] Specifically, using a drone to photograph and extract the farmland information of the target plot to obtain the farmland information image can improve the information extraction efficiency.

[0071] Segment-by-segment mining module 2: Segment and divide the pixel value range of the farmland information image, and obtain the farmland road data for each segment through segment-by-segment mining.

[0072] Specifically, using image processing software, perform gray-scale processing on the farmland information image, extract one kind of pixel of the farmland information image as a pixel set to obtain a gray-scale image, divide the pixel set of the gray-scale image according to a preset step size into several pixel value ranges with equal step sizes, and perform data mining on the several pixel value ranges with equal step sizes.

[0073] Specifically, take the pixel value range, input the gray value of the pixel points in the grayscale image into the pixel value range, and determine whether the pixel points are within the pixel value range. If so, display the pixel points in black; if not, retain the original color of the pixel points and display them in the form of an image within the pixel value range. The pixel points distributed within the pixel value range will be displayed as black dots. According to the distribution of the black dots, the distribution of the corresponding ground objects within the pixel value range can be analyzed, thereby obtaining the farmland road data within the pixel value range.

[0074] Further, perform data mining on the several pixel value ranges in sequence to obtain segmented farmland road data.

[0075] Summary and analysis module 3: Summarize the segmented farmland road data to obtain the distribution probability of farmland roads.

[0076] Specifically, summarize and analyze the segmented farmland road data, sort out the distribution range and distribution probability of the gray values of farmland ground objects, and extract the distribution range and distribution probability of the gray values of farmland road data therefrom.

[0077] Optimization processing module 4: Perform optimization processing according to the distribution probability of farmland roads and select the extraction threshold for farmland roads.

[0078] Specifically, optimize and select the extraction threshold for farmland roads according to the distribution range and distribution probability of the gray values of the farmland road data.

[0079] Further, if there is a clear boundary between the distribution range of the gray values of farmland roads and the distribution range of the gray values of other ground objects, then select the boundary gray value as the extraction threshold for farmland roads.

[0080] Further, when the distribution range of the gray values of farmland roads and the distribution range of the gray values of other ground objects overlap, analyze according to the distribution probability of farmland roads and the distribution probability of other ground objects, and select the gray value with a high distribution probability of farmland roads and a low distribution probability of other ground objects as the extraction threshold for farmland roads.

[0081] Embodiment 3:

[0082] In the embodiment of the present invention, for the Ningxi Base of South China Agricultural University in Zengcheng, Guangzhou, Guangdong Province, the extraction threshold for farmland roads is selected. Under clear weather conditions, a drone is used to collect farmland information and obtain a farmland information image.

[0083] Specifically, use general python software to perform gray processing on the farmland information image, extract the gray values of the image pixels, and obtain a grayscale image.

[0084] Further, the grayscale processing equalizes the three-color pixels RGB in the farmland information image, namely the red pixel, the green pixel, and the blue pixel, to form a grayscale image. Extracting one of the three-color pixels can achieve grayscale conversion. For example, by extracting the blue pixel and converting the red pixel and the green pixel into the blue pixel, a grayscale image can be obtained.

[0085] Further, the grayscale value of the B image in the RGB image is extracted for analysis, or the R image or the G image in the RGB image can also be extracted for analysis and processing.

[0086] Specifically, the pixel value range of the grayscale image is equally divided. Assuming the step size of each segment is 10, the pixel value range of the grayscale image can be divided into 26 segments. Since the pixel value range of green plants is obvious and its pixel value is less than 90, excluding the pixel value range of green plants can reduce the amount of analysis and calculation and improve the analysis efficiency. That is, the pixel value range from 90 to 220 is equally divided with a step size of 10 into 16 segments of equal-step pixel value ranges.

[0087] Specifically, Figure 4 shows a schematic diagram of farmland road data with a pixel value range of 110 to 120 in the embodiment of the present invention. The ground object information of the farmland information image is converted into black by means of segment-by-segment excavation and displayed through the image.

[0088] Further, the segment-by-segment excavation method means that the grayscale value of each pixel of the grayscale image is input into the pixel value range, and it is judged whether the grayscale value of the pixel point is within the pixel value range. If so, the pixel point is displayed as a black point in the pixel value range. If not, the original color of the pixel point is retained.

[0089] Specifically, as Figure 4 shown, within the range of pixel values from 110 to 120, black dots appear in the upper left path, the right plot, and the left plot, indicating that the pixel grayscale value ranges of the upper left path, the left thin green plant layer, and the bare soil in the right plot are within the range of 110 to 120, and the pixel value ranges of the farmland ground objects are recorded.

[0090] Specifically, data mining is sequentially performed on the 16 segments of pixel value ranges to obtain segment-by-segment farmland ground object distribution data.

[0091] Further, segment-by-segment data mining and analysis are performed on the 16 segments of equal-step pixel value ranges to obtain segment-by-segment farmland road information, and the information is summarized and analyzed to obtain Table 1. Table 1 summarizes the distribution of several farmland ground objects, including farmland paths, upper left paths, upper right paths, bare soil, and thin green plant layers. The symbol "x" represents the situation where the ground object is distributed in this pixel value range and the distribution probability is relatively low, and "xx" represents the situation where the ground object is distributed in this pixel value range and the distribution probability is relatively high.

[0092] Table 1 Classification and summary of farmland road data by section

[0093]

[0094]

[0095] Specifically, the farmland road data are summarized and analyzed to obtain the pixel value range of the farmland features: 150<farmland road<220, 90<upper left corner road<130, 140<upper right corner road<180, 100<bare soil field<150, green thin layer<120.

[0096] Combined with the information in Table 1, farmland roads include farmland paths, upper left corner paths and upper right corner paths. Farmland roads and bare soil fields are staggered. If the threshold is set to 110, a large area of ​​bare soil will be misjudged as a road. If the threshold is set to 140, the interference of a large area of ​​bare soil can be eliminated, and only the upper left corner path will be missed. Based on the high probability of farmland roads being distributed in the pixel value range of 160 to 170, setting the threshold to 140 can improve the accuracy of detection.

[0097] Specifically, Figure 5 A schematic diagram of the extraction result when the threshold value is 140 in an embodiment of the present invention is shown. Figure 6 The schematic diagram of the extraction result with a threshold of 110 in the embodiment of the present invention is shown. Figure 5 and Figure 6 It can be concluded that using a pixel grayscale value of 140 as the extraction threshold will only miss the path in the upper left corner, and can highly accurately identify the farmland path and the path in the upper right corner. However, using a pixel grayscale value of 110 as the extraction threshold will identify a large area of ​​bare soil as a farmland road, affecting the farmland road extraction results and resulting in low extraction accuracy.

[0098] Furthermore, through segment-by-segment data mining and global summary analysis, the pixel value range and distribution probability of the farmland roads can be obtained. The selection of the farmland road extraction threshold is determined based on the pixel value range and distribution probability, which can improve the accuracy of farmland road extraction.

[0099] Specifically, the farmland road extraction threshold optimization selection method in the embodiment of the present invention is compared with the traditional threshold selection method, and Table 2 is obtained. It can be concluded from Table 2 that the farmland road extraction threshold optimization selection method can improve the accuracy of extraction, and at the same time has good applicability and can meet the needs of rapid and automated detection.

[0100] Table 2 Comparison between the farmland road extraction threshold optimization selection method and the traditional threshold selection method

[0101]

[0102]

[0103] Specifically, by selecting the threshold for farmland road extraction through the above-mentioned method for selecting the threshold for farmland road extraction based on data mining, the accuracy of the threshold selection can be guaranteed, and at the same time, it can be applied to the extraction of farmland roads in the same batch, effectively improving the efficiency and accuracy of farmland road extraction and meeting the requirements for rapid and automatic detection of farmland road extraction.

[0104] The embodiment of the present invention provides a method for selecting a threshold for farmland road extraction based on data mining. By segment-by-segment data mining combined with global summary analysis, the distribution range and distribution probability of the pixel values of farmland features are obtained. Optimally selecting the threshold for farmland road extraction according to the distribution range and distribution probability can improve the accuracy of the threshold, improve the efficiency and accuracy of farmland road extraction, and meet the requirements for automatic and rapid detection of farmland road extraction.

[0105] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program. This program can be stored in a computer-readable storage medium, and the storage medium can include: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, etc.

[0106] In addition, the above has introduced in detail a method and device for selecting a threshold for farmland road extraction based on data mining provided by the embodiments of the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for selecting a threshold for extracting farmland roads based on data mining, characterized in that, the selection method includes: collecting farmland information through an image acquisition device to obtain a farmland information image; segmenting and dividing the pixel value range of the farmland information image, and obtaining the segmented farmland feature distribution data through segment-by-segment mining, including: performing grayscale processing on the farmland information image to obtain a preprocessed farmland information image, segmenting the pixel set of the preprocessed farmland information image according to a set step size to obtain a number of pixel value ranges with equal step sizes; sequentially inputting the grayscale value of each pixel point in each segment of the pixel value range, determining whether the pixel point is in the segmented pixel value range, if the pixel point is distributed in the segmented pixel value range, then displaying the pixel point with a black dot, if not, keeping the original color of the pixel point; summarizing the segmented farmland feature distribution data to obtain the distribution probability and pixel value range of the farmland roads, including: summarizing the segmented farmland feature distribution data, comparing and analyzing the probabilities of the farmland features in the segmented farmland road data, so as to obtain the pixel value range and distribution probability of each farmland feature, and extracting the pixel value range and distribution probability of the farmland roads; performing optimization processing according to the distribution probability and pixel value range of the farmland roads, and selecting a threshold for extracting the farmland roads, including: when the boundary between the farmland road distribution and other feature distributions is clear, selecting the boundary pixel value as the threshold for extracting the farmland roads, when the boundary between the farmland road distribution and other feature distributions is staggered, according to the distribution probability of the farmland roads, selecting the threshold for extracting the farmland roads within the pixel value range with a high distribution probability of the farmland roads.

2. The method for selecting a threshold for extracting farmland roads according to claim 1, characterized in that, the collecting farmland information through an image acquisition device to obtain a farmland information image includes: collecting farmland information by shooting with a drone to obtain the farmland information image.

3. A device for selecting a threshold for extracting farmland roads based on data mining, characterized in that, the selection device includes: an information collection module: collecting farmland information through an image acquisition device to obtain a farmland information image; a segment-by-segment mining module: segmenting and dividing the pixel value range of the farmland information image, and obtaining the segmented farmland feature distribution data through segment-by-segment mining, including: performing grayscale processing on the farmland information image to obtain a preprocessed farmland information image, segmenting the pixel set of the preprocessed farmland information image according to a set step size to obtain a number of pixel value ranges with equal step sizes; sequentially inputting the grayscale value of each pixel point in each segment of the pixel value range, determining whether the pixel point is in the segmented pixel value range, if the pixel point is distributed in the segmented pixel value range, then displaying the pixel point with a black dot, if not, keeping the original color of the pixel point; Summary analysis module: Summarize the above-mentioned piece-by-piece farmland feature distribution data to obtain the distribution probability of farmland roads, including: Summarize the above-mentioned piece-by-piece farmland feature distribution data, compare and analyze the probabilities of farmland features in the piece-by-piece farmland road data, so as to obtain the pixel value range and distribution probability of each farmland feature, and extract the pixel value range and distribution probability of farmland roads; Optimization processing module: Perform optimization processing according to the above-mentioned farmland road distribution probability and select the farmland road extraction threshold, including: When the boundary between the farmland road distribution and other feature distribution areas is clear, select the boundary pixel value as the farmland road extraction threshold; when the boundary between the farmland road distribution and other feature distributions is staggered, according to the above-mentioned farmland road distribution probability, select the farmland road extraction threshold within the pixel value range with a high farmland road distribution probability.

Images