A method for making intelligent interpretation samples of land disturbance in power grid line engineering based on ArcGIS

Abstract

The present invention provides a method for making samples of land disturbance intelligent interpretation of power grid line engineering based on ArcGIS, and the method includes: 1) sorting out the verification data of land disturbance of environmental protection and water conservation of historical power grid line engineering, obtaining drone orthophotos and land disturbance vectors; 2) performing buffer zone analysis based on land disturbance vectors, and then performing minimum boundary geometry processing to obtain the minimum boundary vector as the sample size of land disturbance intelligent interpretation; 3) using the minimum boundary vector to batch cut out sample images from drone orthophotos; 4) obtaining label vectors through erasing and merging operations based on land disturbance vectors and minimum boundary vectors; 5) converting label vectors into sample labels in a raster format; 6) one-to-one correspondence between sample images and sample labels, that is, forming land disturbance intelligent interpretation samples. The method realizes the rapid production of land disturbance samples, makes the sample production process simpler and more convenient, and greatly reduces the cost of manual labeling of traditional interpretation samples.

Description

Technical Field

[0001] The present invention relates to the field of on-site verification of environmental protection and water conservation during the construction of power grid line projects, and in particular to a method for preparing land disturbance intelligent interpretation samples for power grid line projects based on ArcGIS, which can realize rapid preparation of land disturbance samples based on historical engineering drone verification data, save time and manpower spent on manual annotation of traditional sample preparation, and improve the efficiency of intelligent interpretation sample preparation. Background Art

[0002] The control of environmental protection and water conservation during the construction of power grid line projects mainly focuses on the monitoring of land disturbance, which is an important means to control the impact of project construction on the ecological environment. Because power grid line projects span multiple provinces and regions, with large construction scales and long distances, involving the construction of thousands of towers, the monitoring scope of land disturbance is characterized by many points and a wide range. The current monitoring method has changed from traditional reliance on manpower verification to the use of drone remote sensing technology to conduct on-site verification, thus accumulating a large amount of engineering land disturbance verification data.

[0003] Drone verification has been widely used, but a large amount of remote sensing data still relies on manual visual interpretation, which is also quite time-consuming. Therefore, the use of intelligent interpretation technology to replace manual visual interpretation is widely studied. The application of intelligent interpretation first requires sufficient interpretation samples to build an interpretation sample library, and then train an accurate interpretation model. However, the production of traditional samples first involves segmenting the image, and then manually vectorizing and annotating it into intelligent interpretation samples, which consumes a lot of manpower and a lot of time. Therefore, considering that drone verification has accumulated a large amount of drone remote sensing data and visually interpreted vector data, a method for producing intelligent interpretation samples is needed that can make full use of the accumulated historical data to complete the production of interpretation samples, save the workload of traditional manual interpretation sample annotation, and improve the utilization rate of historical data value. Summary of the invention

[0004] Aiming at the problem that land disturbance interpretation samples need to be produced in the land disturbance areas of the intelligent interpretation of UAV orthophotos in production and construction projects, but the traditional manual labeling of samples takes a long time and requires a lot of manpower, the present invention proposes a method for producing land disturbance intelligent interpretation samples for power grid line projects based on ArcGIS, which can realize batch production of land disturbance samples and save the cost of manual labeling.

[0005] The main technical solution steps of the present invention are as follows:

[0006] Step S1, sample production raw data and standardized format, including historically accumulated drone orthophoto data and land disturbance vector data. The drone orthophoto is an orthophoto that has undergone radiation correction, geometric correction, and image stitching, and is in tif format; the land disturbance vector is the data of vectorizing the land disturbance area on the drone orthophoto, and is in shp format, including multiple land disturbance surface elements. The sample production raw data is expressed as follows:

[0007]

[0008] in, It is a collection of raw data, including drone orthophoto data and the land disturbance vector ; It is tif format data; land disturbance vector Is a face vector, including multiple face features , is shp format data; is the number of elements in the land disturbance vector;

[0009] Step S2: determine the land disturbance sample size and convert the land disturbance vector , perform buffer analysis and obtain buffer vector ; Create the minimum boundary vector based on the buffer vector and obtain the minimum boundary vector , the minimum boundary vector range is the sample size;

[0010] Step S3, sample image production, using the minimum boundary vector , cut out sample images from the drone orthophoto ;

[0011] Step S4: Sample labeling, land disturbance vector and the minimum boundary vector After erasing the intersecting part, get the erasure vector , add an attribute column and assign a value of 0; change the land disturbance vector Add an attribute column and assign a value of 255; erase the vector and the land disturbance vector After merging, get the label vector ; The label vector Converted into raster data, which is the sample label ;

[0012] Step S5: land disturbance intelligent interpretation sample, sample image and sample labels One-to-one correspondence, that is, forming intelligent interpretation samples of land disturbance .

[0013] In one embodiment of the present invention, the sample preparation raw data and normalized format in step S1 are as follows:

[0014] Power grid line projects span multiple provinces and regions, and the construction period is generally 3 to 7 years. Every year, the land disturbance on site must be checked to see if it exceeds the designed disturbance area, so as to control the impact of on-site construction on the environment. Therefore, a large number of historical projects have accumulated verification data, including drone orthophotos, land disturbance vectors, and other auxiliary data. Drone orthophotos span multiple years, multiple phases, and multiple regions; land disturbance vectors are mainly land disturbance areas formed with each tower construction point in the line project as the center.

[0015] Before the widespread application of intelligent interpretation, power grid projects often used drone orthophotos to monitor land disturbance at construction sites. Through visual interpretation + manual vectorization, the area of ​​land disturbance on site was obtained, thus accumulating a large amount of drone orthophotos + land disturbance vector data. The construction of large-scale intelligent interpretation samples is the most basic part of intelligent target recognition. The production of traditional samples is mainly based on image generation samples, and then the interpretation targets are manually marked on the sample images to finally form interpretation target samples. In this process, manual marking will cost a lot of manpower and resources. Therefore, considering the requirements for the production of interpretation samples, combined with the historical accumulation of data from drone monitoring of power grid projects, the data that can be used to make samples are sorted out, including: drone orthophotos and land disturbance vector data.

[0016] In one embodiment of the present invention, the method for determining the land disturbance sample size in step S2 is as follows:

[0017] The intelligent interpretation model of images is trained based on sample blocks of limited size. If the sample size is too small, the model's perception field of view will be limited, which is not conducive to image recognition. If the sample size is too large, the hardware requirements will be high, which will affect the speed of model training. Considering that the disturbance area during the construction of the power grid line is around the tower construction area and is distributed in a point-like manner on the power grid line corridor, the land disturbance sample is produced with the tower area disturbance as the center, and the sample contains as much background area as possible.

[0018] Each vector element in the land disturbance vector As the center, construct a buffer vector; then generate a minimum boundary vector based on each element of the buffer vector. The size of each element in the minimum boundary vector is the size of the corresponding sample. The specific method steps are as follows:

[0019] Step S2.1, using the land disturbance vector Determine buffer radius, land disturbance vector There are multiple elements in , in order to generate samples for each of the surface elements, the radius of the buffer zone is calculated using the surface element with the largest area, which can ensure that the land disturbance vector The disturbed area corresponding to each surface element in can be made into a sample, and the sample includes the surrounding background area in addition to the disturbed area. The area element with the largest area in the image is taken as a circle, and the buffer radius is obtained according to the circle area calculation formula. The specific formula is as follows:

[0020]

[0021] in, is the buffer radius; is the land disturbance vector No. face elements area; is the maximum value function; It is pi.

[0022] Step S2.2: Use the ArcGIS software buffer tool to calculate the land disturbance vector Each polygon feature in Perform buffer analysis to obtain buffer vector ;

[0023]

[0024] in, is the buffer function; is the land disturbance vector Area features The buffer vector obtained after making the buffer The area elements, and the land disturbance vector Area features One to one correspondence; It is for Land disturbance vector polygon features Perform buffer analysis and obtain buffer vector Area features ; Buffer vector Number of area elements and land disturbance vector The number of polygon features is equal.

[0025] Step S2.3, use the Minimum Bounding Geometry tool in ArcGIS software, set the parameter Geometry Type = ENVELOPE, and generate the buffer vector The minimum boundary vector , the minimum boundary vector Each polygon feature in is the size of the corresponding sample:

[0026]

[0027] in, is the buffer vector The minimum boundary vector, the minimum boundary vector The area element in is the sample size; is the minimum boundary geometry function; is the minimum boundary vector The area features, which are buffer vectors The face elements The minimum boundary of the land disturbance vector Area features One-to-one correspondence is the sample size; is the buffer vector The face elements Perform minimum boundary geometry analysis to obtain the minimum boundary vector Area features ; Minimum boundary vector The number of surface elements and land disturbance vector in The number of polygon features is equal.

[0028] In one embodiment of the present invention, the sample image production method in step S3 is as follows:

[0029] Sample image from drone orthophoto , is an image containing land disturbance areas. Using ArcGIS Split Raster tool, combined with the minimum boundary vector , in the drone orthophoto Sample images can be cropped in batches , the specific formula is as follows:

[0030]

[0031] in, is the sample image; is the partition function; It is sample images, by the minimum boundary vector Area features Orthophotos from drones Clipping acquisition, and land disturbance vector Area features One-to-one correspondence; sample images Quantity and Land Disturbance Vector The number of polygon features is equal.

[0032] In one embodiment of the present invention, the sample label preparation method in step S4 is as follows:

[0033] The sample label is a binary image with the same size as the sample image. Each pixel on the sample label has only two values: the land disturbance area has a value of 255, and the other areas have a value of 0. The sample label is made based on the land disturbance vector and the minimum boundary vector. The land disturbance vector corresponds to the 255 area. After erasing the area where the minimum boundary vector and the land disturbance vector intersect, the erased vector is obtained. After merging the erased vector and the land disturbance vector, it is converted into raster data to obtain the sample label. The specific steps are as follows:

[0034] Step S4.1: Use the ArcGIS Erase tool to erase the minimum boundary vector Vector with land disturbance The intersecting parts are erased to obtain the erasing vector;

[0035]

[0036] in, is the erase vector, is the minimum boundary vector Erase land disturbance vector vector after the region; is the erase function; It is erase vector elements, and the land disturbance vector Area features and the minimum boundary vector Area features One-to-one correspondence; erase vector The number of surface elements and land disturbance vector The number of surface elements is equal and corresponds one to one.

[0037] Step S4.2: Use ArcGIS Add Field tool to add the land disturbance vector and erase vector Add a column field label, land disturbance vector Add field label=255, Vector add field label=0;

[0038] Step S4.3: Use the ArcGIS Merge tool to merge the land disturbance vector and erase vector Merge to obtain label vector :

[0039]

[0040] in, is the label vector, is the land disturbance vector and erase vector Merged vector; label vector The number of surface elements is the land disturbance vector Twice the number of polygon features.

[0041] Step S4.4: Use the ArcGIS vector-to-raster tool Feature to Raster to convert the label vector Convert to sample labels , its grid value takes the value of the field label;

[0042]

[0043] in, is the sample label, which is a binary image; It is a vector to raster function; It is sample labels, which are represented by the label vector Elements Convert to raster generation, The pixel value of the corresponding area is 255. The pixel value of the corresponding area is 0; the sample label The number of sample labels and the land disturbance vector The number of polygon features is equal.

[0044] In one embodiment of the present invention, the land disturbance intelligent interpretation sample of step S5 is as follows:

[0045] A standard intelligent interpretation sample consists of a sample image and a sample label, and the sample image and the sample label correspond one to one, and the sample label is a binary image. The specific representation is as follows:

[0046] in, It is a sample of intelligent interpretation of land disturbance and a sample image and sample labels A collection of sample images and sample labels, one-to-one correspondence; It is a sample of intelligent interpretation of land disturbance The samples; intelligent interpretation samples of land disturbance The number of samples in is equal to the land disturbance vector The number of polygon features.

[0047] The technical solution of the present invention includes at least the following beneficial effects:

[0048] Compared with the common manual annotation method for existing intelligent interpretation samples, this method is based on the data accumulated from historical projects, and designs a sample production method based on drone orthophotos and land disturbance vectors, which saves a lot of time and cost for manual re-annotation of land disturbance samples. The intelligent interpretation sample production method based on drone orthophotos and land disturbance vectors can quickly convert historical accumulated land disturbance data into land disturbance samples for training land disturbance interpretation models; the method for determining the size of land disturbance samples ensures accurate extraction of sample images from drone orthophotos, avoiding the traditional method of sequentially cutting images and generating redundant samples without land disturbance targets; the method for generating sample images can realize batch accurate cutting of sample images from drone orthophotos.

[0049] Compared with the traditional manual standard sample production method, the present invention reduces the manual input of labeling, significantly improves the efficiency and automation of drone sample image production, and achieves the goal of quickly realizing sample production by using historical accumulated data. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The embodiments of the present invention will be described below with reference to the accompanying drawings, wherein:

[0051] Figure 1 It is a flow chart for making intelligent interpretation samples based on UAV orthophotos and land disturbance vectors;

[0052] Figure 2 It is a partial map of the embodiment in which the historical drone orthophoto and the land disturbance vector data are superimposed and displayed;

[0053] Figure 3 is based on a land disturbance vector element The process for determining appropriate sample sizes;

[0054] Figure 4 is based on Crop the corresponding sample image according to the corresponding sample size;

[0055] Figure 5 is based on a land disturbance vector element Generate sample labels with corresponding minimum boundary vectors;

[0056] Figure 6 This is a land disturbance intelligent interpretation sample, including sample images and sample labels. DETAILED DESCRIPTION

[0057] A method for producing intelligent interpretation samples of land disturbance in power grid line engineering based on ArcGIS is proposed according to the industry demand for intelligent interpretation of land disturbance in the construction process of power grid line engineering, based on the accumulated drone orthophoto data and land disturbance vector data of historical engineering, which can realize the rapid production of samples required for interpretation model training and save the repeated workload of manual annotation. In order to demonstrate the technical scheme and advantages of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiment of the present invention.

[0058] like Figures 1 to 6 The specific solution process of the embodiment includes the following steps:

[0059] Step 1: Sort out the drone orthophotos and land disturbance vector data accumulated from historical land disturbance inspections. Figure 2 The local map of the superposition of drone orthophoto and land disturbance vector (including 1 land disturbance vector surface element) is shown. The land disturbance vector has a total of 223 surface elements, of which the largest element has an area of ​​1,174 square meters. It can be inferred that the buffer zone radius is 20 meters;

[0060] Step 2: Based on the buffer tool Buffer of ArcGIS, the land disturbance vector is first subjected to buffer analysis, with a buffer radius of R = 20 meters, to obtain the buffer vector; then the ArcGIS minimum bounding geometry tool Minimum Bounding Geometry is used, with the geometry parameter Geometry Type in the tool set to ENVELOPE, to generate the minimum bounding vector of the buffer vector; Figure 3 It is one of the land disturbance vector surface features , whose corresponding buffer vector element is , the corresponding minimum boundary vector element is , It is the land disturbance surface element The sample size of

[0061] Step 3: Use ArcGIS Split Raster tool to batch crop the drone orthophotos in combination with the minimum boundary vector to generate sample images. Figure 4 Shows the minimum bounding vector features After cropping the drone orthophoto, generate a sample image ;

[0062] Step 4: Use the ArcGIS Erase tool to erase the part where the minimum boundary vector and the land disturbance vector intersect to obtain the erased vector; use the ArcGIS Add Field tool to add a field label to the land disturbance vector and the erased vector. The label value of the land disturbance vector is 255, and the label value of the erased vector is 0; use the ArcGIS Merge tool to merge the land disturbance vector and the erased vector to obtain a merged vector, i.e., a label vector. The label vector attribute table contains a label field. Figure 5 Shows the minimum bounding vector features Erase land disturbance polygon features After that, the erase vector element is obtained , the minimum bounding vector element and land disturbance surface elements After adding the label field, merge to generate merged vector features.

[0063] Step 5: Use the ArcGIS vector-to-raster tool Feature to Raster to convert the merged vector into a label image; Figure 5 Shows how one of the elements in the merged vector is converted to a sample label In the sample label, the pixel value of the land disturbance area is 255, and the pixel value of other areas is 0.

[0064] Through the above steps, the final sample image and sample label form a land disturbance sample, such as Figure 6 As shown, the sample image and sample label have the same size and pixel size, and the format is tif. The sample label is a binary image. The corresponding pixel value of the land disturbance area is 255, and the pixel value of other areas is 0. The sample can be converted into any other sample format for intelligent interpretation.

[0065] This method is mainly based on the demand for quantitative monitoring of land disturbance during the construction of power grid projects. Traditional visual interpretation consumes a lot of manpower, so intelligent interpretation is needed to improve the efficiency of interpretation. Sample production is the most basic step to achieve intelligent interpretation. Traditional sample production requires manual labeling, which takes a long time. Combined with the historical data accumulated in the project, a method for producing land disturbance samples is proposed, which can generate sample images in batches, saving the traditional large amount of manual labeling workload and improving the training process of the intelligent interpretation model.

Claims

1. A method for preparing intelligent interpretation samples of land disturbance in power line engineering based on ArcGIS, characterized in that: It specifically includes the following steps: Step S1, obtaining the original data and standardized format of sample production, including the historically accumulated UAV orthophoto data and land disturbance vector data. The UAV orthophoto is an orthophoto that has undergone radiation correction, geometric correction, and image stitching, and is in tif format; the land disturbance vector is the data of vectorizing the land disturbance area on the UAV orthophoto, and is in shp format, including multiple land disturbance surface elements. The original data of sample production is expressed as follows: Data={UAV image ,Vector polygon } Vector polygon ={V1,V2,...,V n } Among them, Data is the original data set, including UAV orthophoto data UAV image and land disturbance vector Vector polygon ;UAV image It is tif format data; land disturbance vector Vector polygon Is a surface vector, including multiple surface elements V n , is the shp format data; n is the number of surface elements in the land disturbance vector; Step S2, determine the land disturbance sample size, and transform the land disturbance vector Vector polygon , perform buffer analysis and obtain the buffer vector Buffer_Vector; based on the buffer vector, make the minimum boundary vector and obtain the minimum boundary vector Envelope_Vector. The minimum boundary vector range is the sample size; Step S3, making a sample image, using the minimum boundary vector Envelope_Vector to cut out a sample image Sample_image from the drone orthophoto; Step S4, create sample labels and label the land disturbance vector Vector polygon After erasing the part that intersects with the minimum boundary vector Envelope_Vector, get the erase vector Erase_Vector, add an attribute column to it, and assign a value of 0; polygon Add an attribute column and assign a value of 255; Erase_Vector and land disturbance vector Vector polygon After merging, obtain the label vector Label_Vector; convert the label vector Label_Vector into raster data, and the raster data is the sample label Sample_label; Step S5, matching the sample image Sample_image and the sample label Sample_label one by one, thus forming a land disturbance intelligent interpretation sample Sample; The step S2 specifically includes the following steps: Step S2.1, using the land disturbance vector Vector polygon Determine the buffer zone radius and the land disturbance vector Vector polygon There are multiple surface elements in , in order to generate samples for each of the surface elements, the buffer radius is calculated using the surface element with the largest area, which can ensure that the land disturbance vector Vector polygon The disturbance area corresponding to each surface element in can be made into a sample, and the sample includes the surrounding background area in addition to the disturbance area. polygon The area element with the largest area in is taken as a circle, and the buffer radius is obtained according to the circle area calculation formula. The specific formula is as follows: Where R is the buffer radius; is the land disturbance vector Vector polygon The nth face element V n is the area of ​​; max is the maximum value function; π is the circumference of a circle; Step S2.2: Use the buffer tool in ArcGIS software to calculate the land disturbance vector Vector polygon Each face feature V in n Perform buffer analysis to obtain the buffer vector Buffer_Vector; Buffer_Vector=Buffer(Vector polygon ) ={Buffer(V1),Buffer(V2),……,Buffer(V n )} ={Buffer_V1,Buffer_V2,……,Buffer_V n } Among them, Buffer is the buffer function; Buffer_V n is the land disturbance vector Vector polygon The area element V n The nth face element in the buffer vector Buffer_Vector obtained after the buffer is made is the same as the land disturbance vector Vector polygon The area element V n One-to-one correspondence; Buffer(V n ) is the nth land disturbance vector Vector polygon The area element V n Perform buffer analysis and obtain the area element Buffer_V of the buffer vector Buffer_Vector n ; Buffer vector Buffer_Vector number of surface elements and land disturbance vector Vector polygon The number of surface elements is equal; Step S2.3, using the Minimum Bounding Geometry tool of ArcGIS software, set the parameter Geometry Type = ENVELOPE, and generate the minimum boundary vector Envelope_Vector of the buffer vector Buffer_Vector. Each face element in the minimum boundary vector Envelope_Vector is the size of the corresponding sample: Envelope_Vector=MBG(Buffer_Vector) ={MBG(Buffer_V1),MBG(Buffer_V2),……,MBG(Buffer_V n )} ={Envelope_V1,Enveloppe_V2,……,Envelope_V n } Where Envelope_Vector is the minimum boundary vector of the buffer vector Buffer_Vector, and the surface element in the minimum boundary vector Envelope_Vector is the sample size; MBG is the minimum boundary geometry function; Envelope_V n It is the nth face element in the minimum boundary vector Envelope_Vector and the nth face element Buffer_V in the buffer vector Buffer_Vector n The minimum boundary of the land disturbance vector Vector polygon The area element V n One-to-one correspondence is the sample size; MBG (Buffer_V n ) is the nth face element Buffer_V in the buffer vector Buffer_Vector n Perform minimum boundary geometry analysis to obtain the surface element Envelope_V of the minimum boundary vector Envelope_Vector n ; The number of surface elements in the minimum boundary vector Envelope_Vector and the land disturbance vector Vector polygon The number of polygon features is equal.

2. According to claim 1, a method for preparing samples of land disturbance intelligent interpretation of power grid line engineering based on ArcGIS is characterized in that: The specific method of step S3 is as follows: Using ArcGIS Split Raster tool, combined with the minimum boundary vector EnVelope_Vector, the UAV orthophoto image The sample images Sample_image can be cropped in batches. The specific formula is as follows: Among them, Sample_image is the sample image; SplitR is the segmentation function; is the nth sample image, consisting of the surface element Envelope_V of the minimum boundary vector Envelope_Vector n Orthophotos from UAV image Crop acquisition and land disturbance vector Vector polygon The area element V n One-to-one correspondence; the number of sample images Sample_image and the land disturbance vector Vector polygon The number of polygon features is equal.

3. According to claim 2, a method for preparing samples of land disturbance intelligent interpretation of power grid line engineering based on ArcGIS is characterized in that: The step S4 specifically comprises the following steps: Step S4.1: Use the ArcGIS Erase tool to merge the minimum boundary vector Envelope_Vector with the land disturbance vector Vector polygon The intersecting parts are erased to obtain the erase vector Erase_Vector. The specific formula is as follows: Erase_Vector=Erase(Envelope_Vector,Vector polygon ) ={Erase1,Erase2,...,Erase n } Among them, Erase_Vector is the erase vector, which is the minimum boundary vector Envelope_Vector erase land disturbance vector Vector polygon The vector after the region; Erase is the erase function; Erase n is the nth erasure vector element, and the land disturbance vector Vector polygon The area element V n and the area feature Envelope_V with the minimum boundary vector Envelope_Vector n One-to-one correspondence; the number of surface elements of the erase vector Erase_Vector and the land disturbance vector Vector polygon The number of area elements is equal and one-to-one corresponding; Step S4.2, use ArcGIS Add Field tool to add land disturbance vector Vector polygon Add a column field label to Erase_Vector and a land disturbance vector Vector polygon Add field label=255, Erase_Vector vector adds field label=0; Step S4.3, using the ArcGIS Merge tool, merge the land disturbance vector Vector polygon Merge it with the erase vector Erase_Vector to get the label vector Label_Vector: Label_Vector=Merge(Erase_Vector,Vector polygon ) ={{Erase1,V1),{Erase2,V2},...,{Erase n ,V n }} Among them, Label_Vector is the label vector, which is the land disturbance vector Vector polygon The vector after merging with the erase vector Erase_Vector; the number of face elements of the label vector Label_Vector is the land disturbance vector Vector polygon twice the number of area elements; Step S4.4, using the ArcGIS vector-to-raster tool Feature to Raster, convert the label vector Label_Vector into a sample label Sample_label, whose raster value is the value of the field label; Among them, Sample_label is the sample label, which is a binary image; FTR is the vector-to-raster function; is the nth sample label, which is the element of the label vector Label_Vector {Erase n , V n } is converted to grid generation, V n The pixel value of the corresponding area is 255, Erase n The pixel value of the corresponding area is 0; the number of sample labels of the sample label Sample_label and the land disturbance vector Vector polygon The number of polygon features is equal.

4. The method for preparing samples of land disturbance intelligent interpretation of power grid line engineering based on ArcGIS according to claim 2 or claim 3, characterized in that: The specific contents of step S5 are as follows: A standard intelligent interpretation sample consists of a sample image and a sample label, and the sample image and the sample label correspond one to one. The sample label is a binary image, which is specifically expressed as follows: Among them, Sample is the land disturbance intelligent interpretation sample, which is a collection of sample images Sample_image and sample labels Sample_label. The sample images and sample labels correspond one to one. is the nth sample in the land disturbance intelligent interpretation sample Sample; the number of samples in the land disturbance intelligent interpretation sample Sample is equal to the land disturbance vector Vector polygon The number of polygon features.

Images