Method and device for identifying and positioning slope dangerous rocks based on drone inspection

Abstract

The present application discloses a method and device for identifying and positioning slope dangerous rocks based on UAV inspection. Among them, the method for identifying and positioning slope dangerous rocks based on UAV inspection includes: determining whether there is an overlapping area among multiple target images of a target area obtained by the UAV; in response to there being an overlapping area among multiple target images, cross-verifying the target images with the overlapping area based on the target points in the overlapping area; correcting the target images based on the results of the cross-verification; and determining the dangerous rock body risk parameters of the target area based on the corrected target images. In this way, the image quality and content accuracy are improved, the accuracy of anomaly recognition based on images can be enhanced, and the monitoring convenience and accuracy are improved.

Description

Technical Field

[0001] This application relates to the technical field of disaster monitoring, and particularly to a method and device for identifying and positioning slope dangerous rocks based on drone inspections. Background Art

[0002] For the disaster monitoring and prevention and control of high-altitude terrains such as slope dangerous rocks, currently, generally three methods of manual investigation, remote sensing image interpretation, and local sensor monitoring are relied on. However, these methods have the defects of low monitoring accuracy and great difficulty in sensor layout. Summary of the Invention

[0003] An embodiment of this application provides a method and device for identifying and positioning slope dangerous rocks based on drone inspections to solve the technical problem of poor geological anomaly monitoring effect.

[0004] According to the first aspect of the embodiment of this application, a method for identifying and positioning slope dangerous rocks based on drone inspections is provided. The method includes: determining whether there is an overlapping area among multiple target images of a target area obtained by a drone; in response to there being an overlapping area among multiple target images, cross-verifying the target images with the overlapping area based on target points in the overlapping area; correcting the target images based on the result of the cross-verification; and determining the risk parameters of dangerous rock bodies in the target area based on the corrected target images.

[0005] Further, before cross-verifying the target images with the overlapping area based on target points in the overlapping area in response to there being an overlapping area among multiple target images, the method further includes: designing a convolutional kernel of an instance segmentation convolutional neural network based on a rock texture direction filter; designing a loss function of the instance segmentation convolutional neural network based on a rock texture direction consistency constraint term; optimizing the instance segmentation convolutional neural network based on the convolutional kernel and the loss function; and the cross-verifying the target images with the overlapping area based on target points in the overlapping area includes: processing the target images with the overlapping area based on the instance segmentation convolutional neural network to cross-verify the target images with the overlapping area.

[0006] Further, the designing a loss function of the instance segmentation convolutional neural network based on a rock texture direction consistency constraint term includes: determining the product of the rock texture direction consistency constraint term and a predetermined balance coefficient; and adding the product to the segmentation loss term of the instance segmentation convolutional neural network to obtain the total loss term of the loss function of the instance segmentation convolutional neural network.

[0007] Further, the cross-verification of the target image with an overlapping region based on the target points in the overlapping region includes: processing the target image with the overlapping region based on the instance segmentation convolutional neural network and the attention mechanism to obtain an instance segmentation mask corresponding to the target image; determining the target weights of the target points in the overlapping region in multiple target images with the overlapping region; the target weights are related to the image quality and / or confidence of the target image; performing weighted average calculation based on the target weights and the instance segmentation mask corresponding to the target image to obtain the result of the cross-verification of the target image.

[0008] Further, the determination of the dangerous rock mass risk parameters of the target region based on the corrected target image includes: determining the coordinate data of at least one dangerous rock mass in the target region based on the corrected target image; determining the surface area of the dangerous rock mass based on the coordinate data of the dangerous rock mass; determining the slope distribution field of the target region through a digital elevation model (DEM); determining the slope at the location of the dangerous rock mass based on the slope distribution field; determining the dangerous rock mass risk parameters of the target region based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass.

[0009] Further, after determining the dangerous rock mass risk parameters of the target region based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass, the method further includes: determining the temperature difference parameter of the dangerous rock mass in the target region based on the infrared image of the target region acquired by the UAV within a predetermined time period; in response to the temperature difference parameter exceeding a preset range, adjusting the dangerous rock mass risk parameters of the target region based on the temperature difference parameter.

[0010] Further, the determination of the dangerous rock mass risk parameters of the target region based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass includes: calculating the instability probability of the dangerous rock mass based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass; generating a spatial distribution map of the dangerous rock mass in the target region through a kernel density estimation algorithm; determining the distribution density of the dangerous rock mass based on the spatial distribution map; performing weighted calculation based on the instability probability and the distribution density to determine the dangerous rock mass risk parameters in the target region.

[0011] Further, the calculation of the instability probability of the dangerous rock mass based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass includes: obtaining the microseismic signal intensity parameter and the current rainfall parameter inside the dangerous rock mass; performing weighted calculation on the microseismic signal intensity parameter and the current rainfall parameter based on a predetermined weight coefficient to obtain an environmental parameter; calculating the instability probability of the dangerous rock mass based on the surface area, the slope, the mechanical parameters corresponding to the dangerous rock mass, and the environmental parameter.

[0012] According to a second aspect of the embodiments of the present application, there is also provided a slope dangerous rock identification and positioning device based on drone inspection, including: a determination unit configured to determine whether there is an overlapping area among multiple target images of a target area obtained by the drone; a verification unit configured to, in response to there being an overlapping area among multiple target images, perform cross-verification on the target images with the overlapping area based on the target points in the overlapping area; a correction unit configured to correct the target images based on the results of the cross-verification; and a processing unit configured to determine the dangerous rock body risk parameters of the target area based on the corrected target images.

[0013] According to a third aspect of the embodiments of the present application, there is also provided an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory is configured to store a computer program; and the processor is configured to implement the method according to any one of the foregoing first aspects when executing the computer program.

[0014] A slope dangerous rock identification and positioning method proposed by the embodiments of the present application determines whether there is an overlapping area among multiple target images of a target area obtained by the drone; in response to there being an overlapping area among multiple target images, performs cross-verification on the target images with the overlapping area based on the target points in the overlapping area; corrects the target images based on the results of the cross-verification; and determines the dangerous rock body risk parameters of the target area based on the corrected target images. In this way, by using drone inspection to obtain survey images, it not only eliminates the difficulty of manually deploying sensors, but also enables convenient and efficient image acquisition. The images collected by the drone flying at low altitude have higher clarity and accuracy. On this basis, cross-verifying multiple overlapping images according to the common points in the overlapping area can avoid accidental errors of single images or abnormal monitoring errors caused by insufficient clarity, reduce the risk of local distortion or misjudgment, enhance the overall identification effect, and improve the robustness in complex environments. Description of the Drawings

[0015] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0016] Figure 1 is a hardware structure block diagram of a computer according to an embodiment of the present application;

[0017] Figure 2 is a flowchart of a slope dangerous rock identification and positioning method based on drone inspection according to an embodiment of the present application;

[0018] Figure 3 It is a flowchart of a method for identifying and positioning slope dangerous rocks based on UAV inspection according to an embodiment of the present application;

[0019] Figure 4 It is a schematic diagram of a device for identifying and positioning slope dangerous rocks based on UAV inspection according to an embodiment of the present application;

[0020] Figure 5 It is a structural diagram of an electronic device according to an embodiment of the present application. Detailed implementation manners

[0021] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application. It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0023] The method embodiment provided in the first embodiment of the present application can be executed on a mobile phone, a computer, a tablet or a similar computing device. Taking running on a computer as an example, Figure 1 It is a hardware structure block diagram of a computer according to an embodiment of the present application. As Figure 1 shown, the computer may include one or more ( Figure 1 only one is shown in Figure 1The structure shown is only schematic and does not limit the structure of the above computer. For example, the computer may further include more or fewer components than those shown in Figure 1 or have a different configuration from that shown in Figure 1 .

[0024] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to a method for identifying and positioning slope dangerous rocks based on UAV inspection in the embodiments of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, implements the above method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely provided with respect to the processor 102, and these remote memories can be connected to the computer through a network. Examples of the above network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and combinations thereof.

[0025] The transmission device 106 is used to receive or send data via a network. Specific examples of the above network may include the wireless network provided by the communication provider of the computer. In one instance, the transmission device 106 includes a network adapter (Network Interface Controller, abbreviated as NIC), which can be connected to other network devices through a base station and thus can communicate with the Internet. In one instance, the transmission device 106 can be a radio frequency (Radio Frequency, abbreviated as RF) module, which is used to communicate with the Internet wirelessly.

[0026] In this embodiment, a method for identifying and positioning slope dangerous rocks based on UAV inspection is provided. Figure 2 is a flowchart of a method for identifying and positioning slope dangerous rocks based on UAV inspection according to an embodiment of the present application. As Figure 2 shown, the process includes the following steps:

[0027] S10: Determine whether there is an overlapping area in multiple target images obtained by the UAV for the target area;

[0028] S20: In response to there being an overlapping area in multiple target images, perform cross-verification on the target images with overlapping areas based on the target points in the overlapping area;

[0029] S30: Correct the target images based on the results of the cross-verification;

[0030] S40: Determine the risk parameters of the dangerous rock mass in the target area based on the corrected target images.

[0031] In this embodiment, the dangerous rock mass can refer to the rock mass located at the high position of the slope. The risk parameters of the dangerous rock mass can include at least one of the following: the falling risk probability of the dangerous rock mass, the falling risk level of the dangerous rock mass, the falling risk distribution of the dangerous rock mass, and the coordinate data of the dangerous rock mass with falling risk, etc. The method described in this embodiment can be applied to positions such as mine slopes and tunnel openings. By using an unmanned aerial vehicle (UAV) to obtain multiple target images of the target area, it can be obtained by shooting with at least one camera mounted on the UAV. Among them, the camera can be inclined on the UAV, that is, the camera forms a predetermined angle with the horizontal plane, such as 30 degrees or 45 degrees, etc., so as to provide a wider-angle and more comprehensive image as the basis for anomaly monitoring.

[0032] In one embodiment, before step S10, the method may further include: using the UAV to obtain the target image and the infrared image of the target area. For example, the UAV can carry a camera for shooting the target image and a thermal infrared imager for shooting the infrared image.

[0033] In one embodiment, one target image can refer to one frame of the target image. That multiple target images have an overlapping area means that there is an overlap between two or more frames of the target images. For example, there is an overlap between two adjacent frames of images. The target point in the overlapping area can refer to a dangerous rock mass point in the overlapping area. For example, the dangerous rock mass can refer to a dangerous rock mass, a crack or a depression, etc. Here, the target point is included in multiple target images in the same overlapping area.

[0034] In one embodiment, based on the target point in the overlapping area, cross-verifying the target images with the overlapping area may include: cross-verifying the target images with the overlapping area based on the position of the target point in the overlapping area. For example, based on the position of the target point in the overlapping area can refer to based on the position of the target point in the overlapping area and the weight of the target point in each target image.

[0035] Here, the target point can refer to a dangerous rock mass point, etc.

[0036] In one embodiment, correcting the target image based on the result of the cross-verification may include: correcting the target image with a clarity lower than a predetermined threshold based on the result of the cross-verification. For example, correcting the overlapping area in multiple target images based on the result of the cross-verification. For example, according to the result of the cross-verification corresponding to each overlapping area, correcting the target image with the overlapping area, such as the target image with a clarity lower than the predetermined threshold.

[0037] In one embodiment, determining the risk parameters of the dangerous rock mass in the target area based on the corrected target image may include: determining the instability probability of the dangerous rock mass in the target area based on the corrected target image; and determining the risk parameters of the dangerous rock mass in the target area based on the instability probability.

[0038] In one embodiment, determining the risk parameters of the dangerous rock mass in the target area based on the instability probability may include: comparing the instability probability with multiple predetermined intervals to determine the risk level corresponding to the predetermined interval where the instability probability is located as the risk level of the dangerous rock mass in the target area. Wherein, the risk level of the dangerous rock mass may correspond to the dangerous rock mass, and each dangerous rock mass corresponds to a risk level of the dangerous rock mass.

[0039] In one embodiment, determining the risk parameters of the dangerous rock mass in the target area based on the instability probability may include: comparing the instability probability with multiple predetermined intervals to determine the risk level corresponding to the predetermined interval where the instability probability is located as the risk parameters of the dangerous rock mass in the target area; generating the risk distribution of all the dangerous rock masses in the target area based on the risk parameters of the dangerous rock mass. Here, the risk distribution may refer to a risk distribution map, etc.

[0040] Exemplarily, based on the instability probability P and the multiple predetermined intervals indicated by the risk level division standard of the dangerous rock mass, the dangerous rock mass is divided into four levels: low risk (P < 0.3), medium risk (0.3 ≤ P < 0.6), high risk (0.6 ≤ P < 0.9), and extremely high risk (P ≥ 0.9). Generating the risk distribution of all the dangerous rock masses in the target area based on the risk level of the dangerous rock mass may refer to spatially mapping the dangerous rock masses of different risk levels in the three-dimensional real-scene model in combination with the spatial positioning result to form a risk distribution map of the dangerous rock mass.

[0041] In one embodiment, determining the risk parameters of the dangerous rock mass in the target area based on the instability probability may include: performing a weighted average calculation based on the instability probability and the distribution density of the dangerous rock mass to determine the risk parameters of the dangerous rock mass in the target area.

[0042] In this way, by using a drone to obtain survey images, not only the difficulty of manually deploying sensors is eliminated, but also the image acquisition is convenient and efficient, and the clarity and accuracy of the images collected by the drone flying at low altitude are higher. On this basis, cross-verifying multiple overlapping images according to the common points in the overlapping areas can avoid the accidental errors of single images or abnormal monitoring errors caused by insufficient clarity, reduce the risk of local distortion or misjudgment, enhance the overall recognition effect, and improve the robustness in complex environments.

[0043] In some embodiments, before step S20, the method may further include:

[0044] Designing the convolution kernel of the instance segmentation convolutional neural network based on the rock texture direction filter;

[0045] Design a loss function for the instance segmentation convolutional neural network based on the rock texture direction consistency constraint term;

[0046] Optimize the instance segmentation convolutional neural network based on the convolutional kernel and the loss function.

[0047] Here, the above steps can also be executed before step S10. The instance segmentation convolutional neural network is used to process the target image with overlapping regions to perform cross-validation on the target image with overlapping regions, which may mean that the instance segmentation convolutional neural network is used to process the target image with overlapping regions to obtain an instance segmentation mask corresponding to the target image, and this instance segmentation mask can be used to perform cross-validation on the target image with overlapping regions.

[0048] In one embodiment, the cross-validation of the target image with overlapping regions based on the target points in the overlapping regions includes: processing the target image with overlapping regions based on the instance segmentation convolutional neural network to perform cross-validation on the target image with overlapping regions.

[0049] In one embodiment, the rock texture direction filter is a filter based on rock texture direction data. Designing the convolutional kernel of the instance segmentation convolutional neural network based on the rock texture direction filter may mean using the rock texture direction filter as the convolutional kernel of the instance segmentation convolutional neural network.

[0050] Here, the rock texture direction filter can be used as a learnable convolutional kernel to enhance the local features related to the texture direction in the image through convolution operations. This process can help the instance segmentation convolutional neural network focus on the dangerous rock areas with directional structures, thereby improving the perception ability of complex rock forms. At the same time, the instance segmentation convolutional neural network can automatically learn the spatial distribution of these texture features through training and reflect them in instance segmentation.

[0051] In one embodiment, the rock texture direction consistency constraint term can characterize the consistency of the rock texture direction, mainly used to measure the degree of coincidence between the boundary information output by the instance segmentation convolutional neural network and the rock texture direction. Here, the "degree of coincidence" refers to the similarity between the boundary information and the rock texture direction. If the texture direction of the rock has a certain regularity, the loss function designed based on this constraint term can make the boundary information output by the instance segmentation convolutional neural network match the rock texture direction.

[0052] In one embodiment, a rock texture direction consistency constraint term is introduced into the loss function. The purpose of this constraint term is to further improve the segmentation accuracy, especially the boundary recognition ability in complex backgrounds, by guiding the output segmentation result, i.e., the boundary information, to be consistent with the texture direction in the original image. The loss function is used to measure the gap between the output of the instance segmentation convolutional neural network and the ground truth result, and the instance segmentation convolutional neural network can optimize its prediction performance by minimizing the loss function.

[0053] In some embodiments, designing the loss function of the instance segmentation convolutional neural network based on the rock texture direction consistency constraint term includes:

[0054] Determining the product of the rock texture direction consistency constraint term and a predetermined balance coefficient;

[0055] Adding the product to the segmentation loss term of the instance segmentation convolutional neural network to obtain the total loss term of the loss function of the instance segmentation convolutional neural network.

[0056] Here, the segmentation loss term can be used in the segmentation process of the instance segmentation mask output by the instance segmentation convolutional neural network. The instance segmentation mask output by the instance segmentation convolutional neural network may include boundary information, where the boundary information may refer to the boundary contour of the dangerous rock mass, etc.

[0057] The loss function can be expressed as:

[0058]

[0059] Where, is the total loss term, is the traditional segmentation loss term, which is used to optimize the difference between the segmentation effect and the ground truth label; and is the rock texture direction consistency constraint term. By reasonably selecting the predetermined balance coefficient , the sensitivity of the model to the morphological characteristics of the dangerous rock can be further enhanced on the basis of ensuring the segmentation accuracy.

[0060] In this way, by introducing the rock texture direction filter and the rock texture direction consistency constraint term, optimizing the convolutional kernel and the loss function, the process of the instance segmentation convolutional neural network obtaining the instance segmentation mask can better match the texture characteristics of the dangerous rock mass, so that when performing cross-validation based on the instance segmentation mask, the target image can be more accurately corrected, improving the image accuracy and the matching with the characteristics of the dangerous rock mass itself.

[0061] In some embodiments, as Figure 3 shown, the step S20 may include:

[0062] S21: In response to the existence of overlapping regions in multiple target images, process the target images with overlapping regions based on an instance segmentation convolutional neural network and an attention mechanism to obtain an instance segmentation mask corresponding to the target images;

[0063] S22: Determine the target weights of the target points in the overlapping regions in the multiple target images with overlapping regions; the target weights are related to the image quality and / or confidence of the target images;

[0064] S23: Perform weighted average calculation based on the target weights and the instance segmentation mask corresponding to the target images to obtain the cross-validation result of the target images.

[0065] Here, processing the target images with overlapping regions may refer to identifying the boundary contours of dangerous rock masses in the target images with overlapping regions. The instance segmentation mask may carry or represent the boundary information of the dangerous rock masses. The target points may be a dangerous rock mass point in the overlapping region of the target images.

[0066] In one embodiment, the instance segmentation mask corresponding to the target image may refer to the instance segmentation mask of the dangerous rock mass in the target image. Processing the target images with overlapping regions based on an instance segmentation convolutional neural network and an attention mechanism may include: inputting the target images with overlapping regions into the instance segmentation convolutional neural network and processing the target images in combination with the attention mechanism. Among them, processing the target images in combination with the attention mechanism may include: performing an attention mechanism processing operation and a multi-scale feature fusion operation on the target images in combination with the convolutional layer weights, the fully connected layer weights, and the bias parameters. Here, the fully connected layer weights may be the shared weights of the fully connected layer.

[0067] For example, for the input target image I, the processing process of processing the target images with overlapping regions based on an instance segmentation convolutional neural network and an attention mechanism can be expressed by the following formula:

[0068]

[0069] Among them, represents the convolution operation, is the weight of the convolutional layer, is the attention mechanism processing operation, is the multi-scale feature fusion operation, and are the fully connected layer weights and the bias parameters respectively, is the activation function, is the output instance segmentation mask.

[0070] Thus, introducing an attention mechanism into the basic convolutional feature extraction network can effectively capture fine-grained local feature information, while fusing multi-scale semantic features to enhance the network model's perception ability of targets at different scales. On this basis, by introducing weights and weighted calculations, the overlapping other target images can be corrected in combination with the effectiveness of the target point position in the target image, improving the content recognition accuracy of the target image.

[0071] In one embodiment, determining the target weight of the target point position in the overlapping region in multiple target images with overlapping regions may refer to determining the target weight of the target point position in each target image among multiple target images with overlapping regions. The target weight is related to the image quality and / or confidence of the target image, such as being positively correlated. The image quality may include at least one of image clarity, pixel ratio, brightness, and contrast. For example, the higher the image quality of the target image, the higher the target weight of the target point position in this target image, or the higher the confidence of the target image, the higher the target weight of the target point position in this target image.

[0072] Exemplarily, the confidence can characterize at least one of the clarity, importance, number of dangerous rock masses, clarity of the target point position in the target image, and distance of the target point position from the center point of the target image of a target image.

[0073] In one embodiment, performing weighted average calculation based on the target weight and the instance segmentation mask corresponding to the target image may include: performing weighted average calculation on the instance segmentation mask corresponding to the target image based on the target weight and the number of target images with the overlapping region.

[0074] Exemplarily, for the instance segmentation mask of the th frame of video, denotes the instance segmentation mask corresponding to the target point position, that is, the data corresponding to the target point position in the instance segmentation mask. In the overlapping region, the final fusion result

[0075]

[0076] where is the target weight of the th frame of target image at the target point position and

[0077] In this way, through the weights corresponding to the common target points in different target images, the image quality and confidence of different target images can be comprehensively considered. After weighted average calculation, images with overlap and higher quality or better confidence can be used to cross-validate other overlapping images, and the reliability of the corrected other images in the overlapping area is higher, improving the accuracy of identifying geological anomaly risks based on image content.

[0078] In some embodiments, step S10 may include:

[0079] Based on the number of data points in each target image obtained by the drone for the target area, as well as the rotation parameters and displacement parameters of image registration, perform registration optimization on the target images;

[0080] Determine whether there are overlapping areas in the multiple target images after registration optimization.

[0081] In one embodiment, based on the number of data points in each target image obtained by the drone for the target area, as well as the rotation parameters and displacement parameters of image registration, performing registration optimization on the target images may include: based on the number of data points in each target image obtained by the drone for the target area, as well as the rotation parameters and displacement parameters of image registration, perform distortion correction, exposure equalization, and registration optimization on the target images.

[0082] Here, registration optimization may include image stitching and registration optimization.

[0083] Exemplarily, the optimization objectives of image stitching and registration can be expressed by the following formula:

[0084]

[0085] Wherein, is the target point obtained after optimization, is the original data point in the target image, and are the rotation parameters and displacement parameters of image registration, is the total number of data points in the target image.

[0086] In one embodiment, determining whether there are overlapping areas in the multiple target images after registration optimization may include: based on the target images after registration optimization, perform image content comparison to determine whether there are overlapping areas in the multiple target images.

[0087] In this way, according to the rotation parameters, displacement parameters, and the number of data points in the image, the images can be registered more accurately and efficiently, improving the quality of the target images and avoiding errors in subsequent anomaly risk identification caused by image errors.

[0088] In some embodiments, determining the risk parameters of the dangerous rock mass in the target area based on the corrected target image may include:

[0089] Determining the coordinate data of at least one dangerous rock mass in the target area based on the corrected target image;

[0090] Determining the surface area of the dangerous rock mass based on the coordinate data of the dangerous rock mass;

[0091] Determining the slope distribution field of the target area through a digital elevation model (DEM);

[0092] Determining the slope at the location where the dangerous rock mass is located based on the slope distribution field;

[0093] Determining the risk parameters of the dangerous rock mass in the target area based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass.

[0094] In one embodiment, determining the coordinate data of at least one dangerous rock mass in the target area based on the corrected target image may include:

[0095] Mapping the corrected target image into a three-dimensional model based on the internal parameter matrix, rotation matrix, and translation vector of the camera that captured the target image;

[0096] Determining the coordinate data of at least one dangerous rock mass in the target area based on the three-dimensional model and the digital elevation model (DEM).

[0097] In one embodiment, mapping the corrected target image into a three-dimensional model based on the internal parameter matrix, rotation matrix, and translation vector of the camera that captured the target image may include: constructing an inverse perspective projection transformation model based on the internal parameter matrix, rotation matrix, and translation vector of the camera that captured the target image; mapping the corrected target image into a three-dimensional model based on the inverse perspective projection transformation model.

[0098] Here, based on the internal parameter matrix, rotation matrix, and translation vector of the camera that captured the target image may include: based on the internal parameter matrix, rotation matrix, translation vector, Beidou positioning data of the unmanned aerial vehicle, and the image pose information of the target image.

[0099] In one embodiment, mapping the corrected target image into a three-dimensional model may refer to mapping the instance segmentation mask corresponding to the corrected target image into a three-dimensional model.

[0100] Exemplarily, the mapping process may be expressed as: ,

[0101] Wherein, $p$ is a point in the two-dimensional target image, $P$ is a point in the three-dimensional model, $K$ is the camera intrinsic matrix, and $R$ and $T$ are the camera rotation matrix and translation vector, respectively.

[0102] In one embodiment, the coordinate data of the dangerous rock mass may include the coordinate data of the centroid of the dangerous rock mass. For example, determining the coordinate data of at least one dangerous rock mass in the target area based on the three-dimensional model and the digital elevation model may refer to performing pixel-by-pixel spatial overlay based on the vertices of the three-dimensional model and the Digital Elevation Model (DEM), and solving the coordinate data of the center point (i.e., the centroid) of at least one dangerous rock mass based on the centroid weighting algorithm. Among them, the coordinate data may refer to geographical coordinates, which may include longitude, latitude, and elevation.

[0103] Exemplarily, set the regional boundary points of each dangerous rock mass as The regional boundary points can be determined based on the boundary information output by the instance segmentation convolutional neural network, where is the coordinate of a certain vertex in three-dimensional space, then the centroid coordinate data of the dangerous rock mass can be obtained by weighted average calculation:

[0104]

[0105] Among them, is the weight associated with each vertex . The obtained through centroid calculation is the centroid coordinate data of the dangerous rock mass.

[0106] In one embodiment, determining the dangerous rock mass risk parameter of the target area based on the coordinate data of the dangerous rock mass may include: determining the dangerous rock mass risk parameter of the target area based on the coordinate data of the dangerous rock mass and the distribution of the dangerous rock mass in the target image. For example, the above steps may include: determining the surface area of the dangerous rock mass based on the coordinate data of the dangerous rock mass; determining the slope distribution field of the target area through the digital elevation model DEM; determining the slope at the location of the dangerous rock mass based on the slope distribution field; and determining the dangerous rock mass risk parameter of the target area based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass.

[0107] In this way, by performing inverse perspective projection transformation mapping on the two-dimensional target image in combination with camera data to obtain a three-dimensional geographical model, and then jointly obtaining the coordinate data of the dangerous rock mass, such as the dangerous rock mass, with the DEM, the recognition result in the two-dimensional image can be accurately mapped to the surface of the three-dimensional real scene model, achieving sub-pixel registration accuracy and improving the recognition accuracy.

[0108] In some embodiments, after determining the risk parameters of the unstable rock mass in the target area based on the surface area, the slope, and the mechanical parameters corresponding to the unstable rock mass, the method further includes:

[0109] Determining the temperature difference parameter of the unstable rock mass in the target area based on the infrared image of the target area acquired by the unmanned aerial vehicle within a predetermined period;

[0110] In response to the temperature difference parameter exceeding a preset range, adjusting the risk parameters of the unstable rock mass in the target area based on the temperature difference parameter.

[0111] Here, the temperature difference parameter may refer to the temperature difference of the unstable rock mass within a predetermined period. For example, the predetermined period may be one day, and the temperature difference parameter may refer to the day-night temperature difference within one day, etc. The temperature difference parameter can characterize the local cooling caused by fissure water seepage or the temperature rise difference of the rock mass caused by sunlight exposure.

[0112] In one embodiment, the preset range may characterize the normal temperature difference range. For example, the preset range may be 0 to 20 degrees, etc.

[0113] In this way, the thermal infrared image of the unmanned aerial vehicle can be used to detect the temperature anomaly on the surface of the rock mass, such as the local cooling caused by fissure water seepage or the temperature rise difference of the rock mass caused by sunlight exposure. If a certain rock mass shows an anomaly in the day-night temperature difference monitoring, it may indicate that its internal fissures are relatively developed and rich in water, and it should be marked as a high-risk state, thereby further improving the accuracy of the risk parameters of the unstable rock mass.

[0114] In one embodiment, determining the surface area of the unstable rock mass based on the coordinate data of the unstable rock mass may refer to determining the surface area of the unstable rock mass based on the coordinate data of the unstable rock mass by the triangulation method. The coordinate data may include boundary point coordinate data.

[0115] In one embodiment, determining the surface area of the unstable rock mass based on the coordinate data of the unstable rock mass may include: determining the surface area of the unstable rock mass by the triangulation method based on the boundary point coordinate data of the unstable rock mass and the number of boundary points included in the unstable rock mass. For example, based on the boundary point coordinate data of the unstable rock mass, the triangulation method is used to estimate its surface area , and the calculation formula is as follows:

[0116]

[0117] Wherein, represents the vector from the th boundary point to the th boundary point, represents the vector from the th boundary point to the th boundary point, and n is the number of boundary points.

[0118] In one embodiment, determining the slope distribution field of the target area through a digital elevation model (DEM) may refer to determining the slope distribution field of the target area based on the elevation data corresponding to the DEM. For example, based on the DEM elevation data , the slope distribution field is obtained through gradient calculation :

[0119]

[0120] wherein and are partial derivatives of the elevation data in the and directions respectively

[0121] In one embodiment, based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass, determining the dangerous rock mass risk parameter of the target area may include:

[0122] Calculating the instability probability of the dangerous rock mass based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass;

[0123] Generating a spatial distribution map of the dangerous rock mass in the target area through a kernel density estimation algorithm;

[0124] Determining the distribution density of the dangerous rock mass based on the spatial distribution map;

[0125] Performing weighted calculation based on the instability probability and the distribution density to determine the dangerous rock mass risk parameter in the target area.

[0126] Here, the mechanical parameters may include at least one of parameters such as the friction coefficient, cohesion, and gravitational potential energy of the dangerous rock mass. The instability probability may characterize the probability of events such as the dangerous rock mass falling, breaking, collapsing, or caving in, which cause geological anomalies. For example, the instability sliding probability value can be expressed as:

[0127]

[0128] where PE represents gravitational potential energy, represents the slope, represents the surface area of the dangerous rock mass, represents the friction coefficient, and c represents cohesion.

[0129] In one embodiment, calculating the instability probability of the dangerous rock mass based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass may include: calculating, through a dynamic environment factor, based on the surface area, the slope, the mechanical parameters corresponding to the dangerous rock mass, the rainfall infiltration coefficient, and the ground motion response coefficient, to obtain the instability probability of the dangerous rock mass.

[0130] Among them, the dynamic environment factor can be used to characterize the participation degree of the rainfall infiltration coefficient and the ground motion response coefficient in the calculation, such as characterizing weights, etc. The dynamic environment factor can refer to an environment factor adjusted based on current weather data. For example, the rainfall amount of the current weather data is proportional to the rainfall infiltration coefficient. By constructing a risk assessment framework driven by the dynamic environment factor, the rock mass saturation softening effect caused by rainfall infiltration or the rock mass dynamic instability probability caused by seismic loads is quantified, so as to establish a more timely grading early warning mechanism.

[0131] In one embodiment, calculating the instability probability of the dangerous rock mass based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass may include:

[0132] Obtaining the microseismic signal intensity parameter inside the dangerous rock mass and the current rainfall parameter;

[0133] Performing weighted calculation on the microseismic signal intensity parameter and the current rainfall parameter based on a predetermined weight coefficient to obtain an environment parameter;

[0134] Calculating the instability probability of the dangerous rock mass based on the surface area, the slope, the mechanical parameters corresponding to the dangerous rock mass, and the environment parameter.

[0135] Here, obtaining the microseismic signal intensity parameter inside the dangerous rock mass and the current rainfall parameter may include: obtaining the microseismic signal intensity parameter inside the dangerous rock mass through a vibration sensor, and obtaining the current rainfall parameter. Among them, before step S10, the method may further include: dropping vibration sensors to at least one dangerous rock mass in the target area through a drone, and the vibration sensor can be a small seismic sensor, etc.

[0136] In one embodiment, the microseismic signal intensity parameter may be the ground motion response coefficient, which can characterize the microseismic intensity occurring inside the dangerous rock mass caused by ground motion. The current rainfall parameter may be the rainfall infiltration coefficient, which can characterize the current rainfall amount, such as the infiltration amount of the current rainfall into the dangerous rock mass, etc.

[0137] In one embodiment, the instability probability of the dangerous rock mass can be expressed as:

[0138]

[0139] Among them, represents gravitational potential energy represents slope represents the surface area of the unstable rock mass represents the coefficient of friction represents cohesion represents the current rainfall parameter, i.e., rainfall represents the microseismic signal intensity parameter, γ, and are respectively the predetermined weight coefficients corresponding to the current rainfall parameter and the microseismic signal intensity parameter

[0140] In one embodiment, the predetermined weight coefficients may include a first weight coefficient and a second weight coefficient. The current rainfall parameter corresponds to the first weight coefficient, and the second weight coefficient corresponds to the microseismic signal intensity parameter. The values of the first weight coefficient and the second weight coefficient may be determined based on the historical monitoring data of the target area. For example, if it is determined based on the historical monitoring data that the number or frequency of earthquakes in the target area is higher than a preset value, then the second weight coefficient is greater than the first weight coefficient; or, if it is determined based on the historical monitoring data that the daily average rainfall in the target area is higher than a preset value, then the first weight coefficient is greater than the second weight coefficient, etc

[0141] In one embodiment, based on the weighted calculation of the instability probability and the distribution density, determining the unstable rock mass risk parameter in the target area may include: performing a weighted calculation based on the instability probability and the distribution density of all unstable rock masses in the target area

[0142] For example, the weighted calculation may obtain the regional risk value corresponding to the target area. Determining the unstable rock mass risk parameter in the target area may be to determine the unstable rock mass risk parameter in the target area based on the regional risk value

[0143] Exemplarily, the regional risk value may be expressed as:

[0144]

[0145] wherein is the regional risk value represents the coordinate data at the instability probability of the unstable rock mass is the distribution density of the unstable rock mass and are the weight coefficients

[0146] In one embodiment, the weight coefficients corresponding to the distribution density and the probability of instability can be preset fixed values, or they can be flexibly adjusted according to the distribution density and the probability of instability. For example, the first weight coefficient corresponding to the probability of instability can be positively correlated with at least one of the number of dangerous rock masses, the average probability of instability of the dangerous rock masses in multiple target images, the clarity of the target image, etc. For example, the second weight coefficient corresponding to the distribution density can be positively correlated with the distribution density and the average number of dangerous rock masses at the same height. Among them, the average number of dangerous rock masses at the same height can be determined by the elevation data corresponding to the DEM.

[0147] In this way, by combining multiple spatial models such as the slope distribution field, the spatial distribution map, and the DEM, the distribution of dangerous rock masses and the detailed terrain data in the target area can be accurately characterized in all directions, so that the instability probability of each dangerous rock mass and the abnormal risk of the target area can be calculated comprehensively and accurately, further improving the recognition accuracy. On this basis, small seismic sensors are deployed by drones to monitor the weak vibrations inside the dangerous rocks in real time and obtain the microseismic signal intensity. In addition, a rainfall index can be introduced to design and determine the risk parameters of dangerous rock masses more accurately and flexibly according to the rainfall situation and the vibration situation.

[0148] In one embodiment, after determining the risk parameters of the dangerous rock masses in the target area, the method may further include: matching a treatment plan based on the risk parameters of the dangerous rock masses; projecting the treatment plan onto the target image through augmented reality technology.

[0149] Here, the treatment plan can be a reinforcement treatment plan. For high-risk and extremely high-risk dangerous rocks, based on their spatial geometric parameters and local terrain features, and according to the pre-established geological risk assessment and engineering treatment knowledge base, the optimal reinforcement treatment plan is automatically matched and recommended, such as anchor cable reinforcement, installation of active protection nets, or direct removal of potential dangerous rock masses. Then, using augmented reality (AR) technology, the recommended treatment plan is intuitively superimposed on the aerial real-scene images taken by drones and clearly presented to on-site engineering personnel in a visual way.

[0150] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc), and includes several instructions for causing a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present application.

[0151] In this embodiment, a slope dangerous rock identification and positioning device based on UAV inspection is also provided to implement the above-mentioned embodiments and preferred implementation manners, which have been described and will not be repeated here. As used hereinafter, the term "module" can be a combination of software and / or hardware that realizes a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0152] Figure 4 is a block diagram of a slope dangerous rock identification and positioning device based on UAV inspection according to an embodiment of the present application. As Figure 4 shown, the device includes:

[0153] A determination unit 100, configured to determine whether there is an overlapping area in multiple target images of a target area obtained by a UAV;

[0154] A verification unit 200, configured to, in response to there being overlapping areas in multiple target images, perform cross-verification on the target images with overlapping areas based on the target points in the overlapping areas;

[0155] A correction unit 300, configured to correct the target images based on the results of the cross-verification;

[0156] A processing unit 400, configured to determine the dangerous rock body risk parameters of the target area based on the corrected target images.

[0157] Further, before performing cross-verification on the target images with overlapping areas in response to there being overlapping areas in multiple target images based on the target points in the overlapping areas, the verification unit 200 is further configured to: design a convolution kernel of an instance segmentation convolutional neural network based on a rock texture direction filter; design a loss function of the instance segmentation convolutional neural network based on a rock texture direction consistency constraint term; optimize the instance segmentation convolutional neural network based on the convolution kernel and the loss function; the instance segmentation convolutional neural network is used to process the target images with overlapping areas to perform cross-verification on the target images with overlapping areas.

[0158] Further, the verification unit 200 is configured to: determine the product of the rock texture direction consistency constraint term and a predetermined balance coefficient; obtain the total loss term of the loss function of the instance segmentation convolutional neural network by adding the product to the segmentation loss term of the instance segmentation convolutional neural network.

[0159] Further, the verification unit 200 is configured to: process the target image with overlapping regions based on the instance segmentation convolutional neural network and the attention mechanism to obtain an instance segmentation mask corresponding to the target image; determine the target weight of the target points in the overlapping regions in multiple target images with overlapping regions; the target weight is related to the image quality and / or confidence of the target image; perform weighted average calculation based on the target weight and the instance segmentation mask corresponding to the target image to obtain the cross-verification result of the target image.

[0160] Further, the processing unit 400 is configured to: determine the coordinate data of at least one dangerous rock mass in the target area based on the corrected target image; determine the surface area of the dangerous rock mass based on the coordinate data of the dangerous rock mass; determine the slope distribution field of the target area through a digital elevation model (DEM); determine the slope at the location of the dangerous rock mass based on the slope distribution field; determine the dangerous rock mass risk parameter of the target area based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass.

[0161] Further, after determining the dangerous rock mass risk parameter of the target area based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass, the processing unit 400 is further configured to: determine the temperature difference parameter of the dangerous rock mass in the target area based on the infrared image of the target area acquired by the UAV within a predetermined time period; in response to the temperature difference parameter exceeding a preset range, adjust the dangerous rock mass risk parameter of the target area based on the temperature difference parameter.

[0162] Further, the processing unit 400 is configured to: calculate the instability probability of the dangerous rock mass based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass; generate a spatial distribution map of the dangerous rock mass in the target area through a kernel density estimation algorithm; determine the distribution density of the dangerous rock mass based on the spatial distribution map; perform weighted calculation based on the instability probability and the distribution density to determine the dangerous rock mass risk parameter in the target area.

[0163] Further, the processing unit 400 is configured to: obtain the microseismic signal intensity parameter and the current rainfall parameter inside the dangerous rock mass; perform weighted calculation on the microseismic signal intensity parameter and the current rainfall parameter based on a predetermined weight coefficient to obtain an environmental parameter; calculate the instability probability of the dangerous rock mass based on the surface area, the slope, the mechanical parameters corresponding to the dangerous rock mass, and the environmental parameter.

[0164] It should be noted that the above-mentioned modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to this: the above modules are all located in the same processor; or, the above-mentioned modules are separately located in different processors in any combination form.

[0165] An embodiment of the present application further provides an electronic device. Figure 5 It is a structural diagram of an electronic device according to an embodiment of the present application, as Figure 5 shown, including a processor 41, a communication interface 42, a memory 43, and a communication bus 44. Among them, the processor 41, the communication interface 42, and the memory 43 complete mutual communication through the communication bus 44. The memory 43 is used to store computer programs; when the processor 41 executes the programs stored on the memory 43, it implements the steps described in any one or more of the foregoing method embodiments.

[0166] The communication bus mentioned in the above terminal may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, only a thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0167] The communication interface is used for communication between the above terminal and other devices.

[0168] The memory may include a Random Access Memory (RAM), or may also include a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located far from the foregoing processor.

[0169] The above-mentioned processor may be a general-purpose processor, including a Central Processing Unit (CPU for short), a Network Processor (NP for short), etc.; it may also be a Digital Signal Processor (DSP for short), an Application Specific Integrated Circuit (ASIC for short), a Field-Programmable Gate Array (FPGA for short), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0170] In another embodiment provided by the present application, a computer-readable storage medium is also provided. Instructions are stored in the computer-readable storage medium. When it runs on a computer, it causes the computer to execute the method described in any one or more of the above embodiments.

[0171] In another embodiment provided by the present application, a computer program product containing instructions is also provided. When it runs on a computer, it causes the computer to execute the method described in any one or more of the above embodiments.

[0172] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that includes one or more integrated available media. The available medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a Solid State Disk (SSD)).

[0173] The above are only the preferred embodiments of the present application and are not intended to limit the protection scope of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application are all included in the protection scope of the present application.

[0174] The above is only the specific implementation manner of the present application, enabling those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for identifying and locating dangerous rocks on slopes based on drone inspection, characterized in that, Including: Determine whether there is an overlapping area among multiple target images of a target area obtained by a drone; In response to there being an overlapping area among multiple target images, based on the target points in the overlapping area, perform cross-verification on the target images with the overlapping area; Correct the target images based on the results of the cross-verification; Determine the dangerous rock mass risk parameters of the target area based on the corrected target images; Before performing, in response to there being an overlapping area among multiple target images, based on the target points in the overlapping area, performing cross-verification on the target images with the overlapping area, the method further includes: Design the convolutional kernel of the instance segmentation convolutional neural network based on the rock texture direction filter; Design the loss function of the instance segmentation convolutional neural network based on the rock texture direction consistency constraint term; Optimize the instance segmentation convolutional neural network based on the convolutional kernel and the loss function; The performing cross-verification on the target images with the overlapping area based on the target points in the overlapping area includes: Process the target images with the overlapping area based on the instance segmentation convolutional neural network to perform cross-verification on the target images with the overlapping area; The processing the target images with the overlapping area based on the instance segmentation convolutional neural network to perform cross-verification on the target images with the overlapping area includes: Process the target images with the overlapping area based on the instance segmentation convolutional neural network and the attention mechanism to obtain the instance segmentation mask corresponding to the target images; Determine the target weights of the target points in the overlapping area in multiple target images with the overlapping area; the target weights are related to the image quality and / or confidence of the target images; Perform weighted average calculation based on the target weights and the instance segmentation mask corresponding to the target images to obtain the cross-verification results of the target images.

2. The method according to claim 1, wherein The designing the loss function of the instance segmentation convolutional neural network based on the rock texture direction consistency constraint term includes: Determine the product of the rock texture direction consistency constraint term and a predetermined balance coefficient; Add the product to the segmentation loss term of the instance segmentation convolutional neural network to obtain the total loss term of the loss function of the instance segmentation convolutional neural network.

3. The method according to claim 1, characterized in that, The determining the dangerous rock mass risk parameters of the target area based on the corrected target images includes: Determine the coordinate data of at least one dangerous rock mass in the target area based on the corrected target images; Determine the surface area of the dangerous rock mass based on the coordinate data of the dangerous rock mass; Determine the slope distribution field of the target area through the digital elevation model DEM; Determine the slope of the location where the dangerous rock mass is located based on the slope distribution field; Determine the dangerous rock mass risk parameters of the target area based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass.

4. The method according to claim 3, characterized in that, After determining the dangerous rock mass risk parameters of the target area based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass, the method further includes: Determine the temperature difference parameter of the dangerous rock mass in the target area based on the infrared image of the target area obtained by the drone within a predetermined time period; In response to the temperature difference parameter exceeding a preset range, adjust the risk parameter of the dangerous rock mass in the target area based on the temperature difference parameter.

5. The method according to claim 3, characterized in that, The determining the risk parameter of the dangerous rock mass in the target area based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass includes: Calculate the instability probability of the dangerous rock mass based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass; Generate a spatial distribution map of the dangerous rock mass in the target area through a kernel density estimation algorithm; Determine the distribution density of the dangerous rock mass based on the spatial distribution map; Perform weighted calculation based on the instability probability and the distribution density to determine the risk parameter of the dangerous rock mass in the target area.

6. The method according to claim 5, characterized in that, The calculating the instability probability of the dangerous rock mass based on the surface area, the slope, and the mechanical parameters corresponding to the dangerous rock mass includes: Obtain the microseismic signal intensity parameter inside the dangerous rock mass and the current rainfall parameter; Perform weighted calculation on the microseismic signal intensity parameter and the current rainfall parameter based on a predetermined weight coefficient to obtain an environmental parameter; Calculate the instability probability of the dangerous rock mass based on the surface area, the slope, the mechanical parameters corresponding to the dangerous rock mass, and the environmental parameter.

7. A slope dangerous rock identification and positioning device based on drone inspection, characterized in that, Includes: A determining unit, configured to determine whether there is an overlapping area in multiple target images of a target area obtained by an unmanned aerial vehicle; A verification unit, configured to, in response to there being multiple target images with overlapping areas, perform cross-verification on the target images with overlapping areas based on the target points in the overlapping area; A correction unit, configured to correct the target images based on the result of the cross-verification; A processing unit, configured to determine the risk parameter of the dangerous rock mass in the target area based on the corrected target images; Before, in response to there being multiple target images with overlapping areas, performing cross-verification on the target images with overlapping areas based on the target points in the overlapping area, the verification unit is further configured to: design a convolution kernel of an instance segmentation convolutional neural network based on a rock texture direction filter; design a loss function of the instance segmentation convolutional neural network based on a rock texture direction consistency constraint term; optimize the instance segmentation convolutional neural network based on the convolution kernel and the loss function; the instance segmentation convolutional neural network is used to process the target images with overlapping areas to perform cross-verification on the target images with overlapping areas; The verification unit is configured to: process the target images with overlapping areas based on the instance segmentation convolutional neural network and an attention mechanism to obtain an instance segmentation mask corresponding to the target images; determine the target weights of the target points in the overlapping area in multiple target images with overlapping areas; The target weights are related to the image quality and / or confidence of the target images; perform weighted average calculation based on the target weights and the instance segmentation mask corresponding to the target images to obtain the result of the cross-verification of the target images.

8. An electronic device, characterized in that, Includes: A processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete mutual communication through the communication bus; The memory is used to store a computer program; the processor is used to implement the method according to any one of claims 1-6 when executing the computer program.

Images