UAV-based geological image acquisition and processing methods for mining areas

Abstract

This invention discloses a method for acquiring and processing geological images of mining areas based on unmanned aerial vehicles (UAVs), comprising: acquiring several first images of a region to be detected; scaling the first images by a scaling factor A to obtain several second images; binarizing the second images and obtaining the overlapping regions of any two adjacent second images according to the position coordinate system and a feature point matching algorithm; stitching them together to obtain a third image; arranging the positions of the several second images before binarization based on the stitched third image, removing overlapping regions, and obtaining a fourth image; obtaining the fourth images corresponding to several regions to be detected and splitting them into a training set and a validation set; marking roads and fault lines in the training set; building a convolutional neural network; importing the marked training set into the convolutional neural network for training to obtain a trained convolutional neural network; validating the trained convolutional neural network using the validation set and outputting a geological image of the mining area with roads and fault lines.

Description

Technical Field

[0001] This invention relates to the field of geological disaster assessment technology, and in particular to a method for acquiring and processing geological images of mining areas based on unmanned aerial vehicles (UAVs). Background Technology

[0002] In this technology, the mining area refers to the geological area after the removal of the original soil and vegetation. Because the vegetation and topsoil of this area have been removed, it is prone to geological disasters such as landslides during the rainy season. Therefore, reliable image acquisition of the mining area's geology is crucial for subsequent reliable analysis. Currently, there is no reliable method for acquiring and processing geological images of mining areas; existing technologies still rely on manual marking, which is inefficient and prone to errors, depending solely on the identification skills of professionals.

[0003] Therefore, there is an urgent need to propose a simple, accurate, and reliable method for acquiring and processing geological images of mining areas based on unmanned aerial vehicles (UAVs). Summary of the Invention

[0004] To address the aforementioned problems, the present invention aims to provide a method for acquiring and processing geological images of mining areas based on unmanned aerial vehicles (UAVs). The technical solution adopted by the present invention is as follows:

[0005] A method for acquiring and processing geological images of mining areas based on unmanned aerial vehicles (UAVs) includes the following steps:

[0006] A location coordinate system is established based on the geographical location of the mining area to be detected, and several first images of the area to be detected are collected using a drone at a fixed cruising altitude.

[0007] Obtain the actual size of the first image captured by the drone at its cruising altitude and the size of the first image itself, and calculate the scaling ratio A; use the scaling ratio A to scale the first image proportionally to obtain several second images;

[0008] Binarize any scaled second image and obtain the overlapping area of ​​any two adjacent second images based on the position coordinate system and a feature point matching algorithm.

[0009] Remove the overlapping areas of two adjacent second images and stitch them together; traverse any adjacent second images to obtain the stitched third image;

[0010] Based on the stitched third image, the positions of several second images before binarization are arranged, overlapping areas are removed, and a fourth image is obtained.

[0011] Obtain fourth images corresponding to several regions to be detected, and split them into training and validation sets; then label the roads and fault lines in the training set;

[0012] Build a convolutional neural network;

[0013] The labeled training set is imported into the convolutional neural network for training, resulting in the trained convolutional neural network.

[0014] The trained convolutional neural network was validated using a validation set, and geological images of the mining area with roads and fault lines were output.

[0015] Further, the step of obtaining the overlapping region of any two adjacent second images based on the position coordinate system and using a feature point matching algorithm includes:

[0016] Based on the position coordinates, obtain any two adjacent second images and calculate several connected components of the two adjacent second images after binarization.

[0017] The connected regions within two adjacent second images are traversed, and the same overlapping regions are obtained using a feature point matching algorithm.

[0018] Further, before traversing the connected components within two adjacent second images, any connected component is scaled; the scaling of any connected component includes the following steps:

[0019] Find the center point of any second image;

[0020] Obtain the geometric center of any connected component in the second image;

[0021] Based on the drone's fixed cruising altitude, the length of the line connecting the center point of the image to the geometric center of the connected domain, and the scaling factor A, the angle θ between the line L1 connecting the drone's shooting point in the second image and the center point of the second image and the line L2 connecting the drone's shooting point to the geometric center of the connected domain is obtained.

[0022] Based on the included angle θ, orthographic projection is performed on any connected region using the orthographic projection method;

[0023] The overlapping regions are obtained by using the connected components after orthographic projection.

[0024] Preferably, the ratio of the training set to the validation set is 7:3.

[0025] Furthermore, obtaining the connected components includes the following steps:

[0026] Obtain the second image after binarization;

[0027] Find the edge pixel feature points where the pixel value changes abruptly in the second image, connect the edge pixel feature points, and obtain the bounding box region of the connected region.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] (1) This invention cleverly acquires several images of the area to be detected, and performs image stitching based on connected component and feature point matching algorithms to obtain a complete image of the area to be detected. This invention utilizes a convolutional neural network to reliably identify the complete image to obtain roads and fault lines, providing a basic guarantee for subsequent landslide monitoring.

[0030] (2) The present invention cleverly uses the scaling ratio A of the actual size of the first image captured by the UAV at the cruising altitude and the size of the first image itself, and combines the orthographic projection of the connected domain. That is, for any connected domain, after its orthographic projection, it is equivalent to the shooting point of the UAV being located directly above the connected domain after the orthographic projection, which ensures that the feature point matching in the later stage is more accurate, so as to obtain an accurate and reliable overlapping area. In this way, the image stitching can be guaranteed to be accurate and reliable, which is suitable for geological image acquisition in large mining areas.

[0031] In summary, this invention has the advantages of simple logic and high accuracy and reliability, and has high practical and promotional value in the field of geological disaster assessment technology. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope of protection. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a schematic diagram of drone data collection according to the present invention.

[0034] Figure 2 This is a logic flowchart of the present invention.

[0035] Figure 3 This is the output image of the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this application clearer, the present invention will be further described below with reference to the accompanying drawings and embodiments. The embodiments of the present invention include, but are not limited to, the following embodiments. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0037] In this embodiment, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0038] The terms "first" and "second," etc., used in the specification and claims of this embodiment are used to distinguish different objects, not to describe a specific order of objects. For example, "first target object" and "second target object," etc., are used to distinguish different target objects, not to describe a specific order of target objects.

[0039] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0040] In the description of the embodiments in this application, unless otherwise stated, "multiple" means two or more. For example, multiple processing units means two or more processing units; multiple systems means two or more systems.

[0041] like Figures 1 to 3 As shown in the figure, this embodiment provides a method for acquiring and processing geological images of mining areas based on unmanned aerial vehicles (UAVs). The method involves continuously acquiring images of the area to be detected using UAVs and then stitching them together. Roads and fault lines are marked on the stitched images to facilitate reliable identification using convolutional neural networks in the later stages.

[0042] Specifically, the UAV-based geological image acquisition and processing method for mining areas in this embodiment includes the following steps:

[0043] The first step is to establish a coordinate system based on the geographical location of the mining area to be inspected, and then use a drone to collect several first images of the area under inspection at a fixed cruising altitude. Since several first images of the mining area to be inspected are required, and labeling any one of these images would be extremely labor-intensive, it is necessary to stitch the obtained images together.

[0044] The second step involves obtaining the actual dimensions of the first image captured by the drone at its cruising altitude, along with the dimensions of the first image itself, and calculating the scaling factor A. Then, the first image is scaled proportionally using scaling factor A to obtain several second images. Here, the scaled images and scaling factors are determined to establish the correspondence between the actual scene and the first image. Here, the actual dimensions of the scene corresponding to the first image are the length and width of the scene shown in the captured first image.

[0045] The third step involves binarizing any scaled second image and using a feature point matching algorithm based on the position coordinate system to obtain the overlapping region between any two adjacent second images. This overlapping region is then used for stitching; therefore, obtaining this overlapping region involves the following steps:

[0046] (i) Obtain any two adjacent second images based on their position coordinates, and calculate several connected components of the two adjacent second images after binarization. Before traversing the connected components within the two adjacent second images, scale any connected component; the scaling of any connected component includes the following steps:

[0047] (1) Find the center point of any second image;

[0048] (2) Obtain the geometric center of any connected region in the second image;

[0049] (3) Based on the drone's fixed cruising altitude, the length of the line connecting the center point of the image to the geometric center of the connected domain, and the scaling factor A, calculate the angle θ between the line L1 (i.e., from T to O1) connecting the drone's shooting point and the center point of the second image, and the line L2 (i.e., from T to b1) connecting the drone's shooting point and the geometric center of the connected domain. For example... Figure 1 As shown, the second image can be virtually represented as B. The width between points a and c represents the widest distance of the actual scene captured by the drone at a constant cruising altitude. The ratio of this distance to the image width (a1 to c1) is the scaling factor A. Point O represents the actual point corresponding to the center point of the second image; b represents the actual point corresponding to the geometric center of the connected component. Therefore, the line connecting the center point of the second image and the geometric center of the connected component (O1 to b1) is L3. Since the second image is scaled down relative to the actual scene, the distance between points O and b in the actual scene is A times the distance between O1 and b1. According to the principle of similar triangles, the angle between lines L1 and L2 remains constant. In this embodiment, by scaling the distance between O1 and b1 in the second image by A times and then performing an arctangent calculation with the distance at the constant cruising altitude, the angle θ between lines L1 and L2 can be obtained.

[0050] (4) Orthographic projection is performed on any connected component based on the included angle θ. Since a connected component is an irregular or regular region, the connected component of point b obtained from an image taken directly above point O differs from the connected component of point b obtained from an image taken directly above point b. That is, at time t, the UAV takes an image directly above point O, and at time t1, it takes an image directly above point b. Due to the different angles, the same connected component in the two sets of images differs. However, in reality, the corresponding connected components in the two images are the same. Therefore, in this embodiment, the connected components with the same angle can be obtained using the orthographic projection method, so as to facilitate accurate and reliable matching later.

[0051] (5) Use the connected components after orthographic projection to find the overlapping regions.

[0052] In this embodiment, obtaining the connected components includes the following steps:

[0053] (1) Obtain the second image after binarization;

[0054] (2) Find the edge pixel feature points of pixel value change in the second image, connect the edge pixel feature points, and obtain the bounding box region of the connected region.

[0055] (ii) Traverse the connected regions within two adjacent second images and use a feature point matching algorithm to find the same overlapping regions.

[0056] The fourth step is to remove the overlapping areas of two adjacent second images and then stitch them together; traverse any adjacent second images to obtain the stitched third image.

[0057] Fifth, after stitching, the third image is a binarized image, which needs to be restored. Since traditional binarization restoration algorithms are quite complex, this embodiment uses the simplest stitching replacement, that is, arranging the positions of several second images before binarization based on the stitched third image, removing overlapping areas, and obtaining the fourth image.

[0058] The sixth step is to obtain the fourth image corresponding to several regions to be detected, and split them into training set and validation set; and to label the roads and fault lines in the training set.

[0059] Step 7: Build a convolutional neural network.

[0060] The eighth step is to import the labeled training set into the convolutional neural network for training, thereby obtaining the trained convolutional neural network.

[0061] The ninth step is to validate the trained convolutional neural network using a validation set and output a geological image of the mining area with roads and fault lines. In this embodiment, taking a specific mining area as an example, the lines in the image represent roads, and the terraced shapes represent fault lines.

[0062] The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any changes made based on the design principles of the present invention, or any non-creative modifications made thereon, shall fall within the scope of protection of the present invention.

Claims

1. A method for acquiring and processing geological images of mining areas based on unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: A location coordinate system is established based on the geographical location of the mining area to be detected, and several first images of the area to be detected are collected using a drone at a fixed cruising altitude. Obtain the actual size of the first image captured by the drone at its cruising altitude and the size of the first image itself, and calculate the scaling ratio A. The first image is scaled proportionally using scaling factor A to obtain several second images; Binarize any scaled second image and obtain the overlapping area of ​​any two adjacent second images based on the position coordinate system and a feature point matching algorithm. Remove the overlapping areas of two adjacent second images and stitch them together; traverse any adjacent second images to obtain the stitched third image; Based on the stitched third image, the positions of several second images before binarization are arranged, overlapping areas are removed, and a fourth image is obtained. Obtain fourth images corresponding to several regions to be detected, and split them into training and validation sets; then label the roads and fault lines in the training set; Build a convolutional neural network; The labeled training set is imported into the convolutional neural network for training, resulting in the trained convolutional neural network. The trained convolutional neural network was validated using a validation set, and geological images of the mining area with roads and fault lines were output. The step of obtaining the overlapping region of any two adjacent second images based on the position coordinate system and using a feature point matching algorithm includes: Based on the position coordinates, obtain any two adjacent second images and calculate several connected components of the two adjacent second images after binarization. Traverse the connected components within two adjacent second images and use a feature point matching algorithm to find the same overlapping regions; Before traversing the connected components within two adjacent second images, any connected component is scaled; the scaling of any connected component includes the following steps: Find the center point of any second image; Obtain the geometric center of any connected region in the second image; Based on the drone's fixed cruising altitude, the length of the line connecting the center point of the image to the geometric center of the connected domain, and the scaling factor A, the angle θ between the line L1 connecting the drone's shooting point in the second image and the center point of the second image and the line L2 connecting the drone's shooting point to the geometric center of the connected domain is obtained. Based on the included angle θ, orthographic projection is performed on any connected region using the orthographic projection method; The overlapping regions are obtained by using the connected components after orthographic projection.

2. The method for acquiring and processing geological images of mining areas based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The ratio of the training set to the validation set is 7:

3.

3. The method for acquiring and processing geological images of mining areas based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, Obtaining the connected components includes the following steps: Obtain the second image after binarization; Find the edge pixel feature points where the pixel value changes abruptly in the second image, connect the edge pixel feature points, and obtain the bounding box region of the connected region.

Images