Geological image identification marking method based on images acquired by unmanned aerial vehicle

Abstract

The invention provides a geological image identification marking method based on images acquired by an unmanned aerial vehicle, which comprises the following steps of: (1) multi-source data acquisition: acquiring a visual image, an infrared image and / or a laser image aiming at a specific area from a database, and integrating the acquired images into a single multi-channel image; (2) multi-region type identification: carrying out probability division and identification on the multi-channel image according to region and geological types to obtain a partition probability table; (3) primary screening; (4) secondary screening; and (5) map marking. The invention further provides a computer readable storage medium. According to the method, multiple data sources can be effectively integrated through a mode of integrating multiple images into a multi-channel image and performing partition identification, the geological type identification accuracy can be greatly improved, and most environmental interference can be eliminated based on integration, so that effective benefits are provided for geological exploration, an effective geological exploration reference is provided, and the geological exploration efficiency is improved. And the consumption of manpower and material resources is reduced, and the working efficiency of geological exploration in depopulated areas is particularly improved.

Description

Technical Field

[0001] The invention relates to a geological image recognition and marking method, which belongs to a geological image recognition and marking method based on images collected by unmanned aerial vehicles. Background Art

[0002] There are many technologies in the existing technology that use drone aerial photography for geological mapping, such as the Chinese invention patent application number CN202310020153.1 that discloses a drone-based geological information collection method, system and electronic equipment, and the Chinese invention patent application number CN202210784312.0 that discloses a geological identification method and system based on remote sensing technology.

[0003] However, in the existing technology, geological mapping is generally carried out based solely on a certain image type, such as a visual image (i.e., an image taken by an ordinary camera). The data information is single, the accuracy of geological type recognition is low, and it is easily affected by various interferences. Most of them can only carry out geological appearance mapping as accurately as possible, which lacks effective help for geological exploration and makes it difficult to provide effective geological exploration reference. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides a geological image recognition and marking method based on images collected by drones. This geological image recognition and marking method based on images collected by drones can effectively integrate multiple data sources by integrating multiple images into multi-channel images and performing partitioned recognition, thereby greatly improving the accuracy of geological type recognition.

[0005] The present invention is achieved through the following technical solutions.

[0006] The present invention provides a geological image recognition and marking method based on images collected by drones, comprising the following steps:

[0007] ① Multi-source data acquisition: Obtain visual images, infrared images, and / or laser images of a specific area from the database and integrate the acquired images into a single multi-channel image;

[0008] ②Multi-region recognition type: divide the multi-channel image into regions and geological types according to probability and identify them to obtain a partition probability table;

[0009] ③ Primary screening: Screening the partition probability table based on the probability of adjacent geological types;

[0010] ④ Secondary screening: Screening the zoning probability table based on the probability of possible regional geological types;

[0011] ⑤ Map marking: Mark the partition probability table that has undergone secondary screening by region on the map and output it for display.

[0012] The visible image is obtained by normalizing a plurality of continuous images of a specific area collected by a drone through sequence image standardization.

[0013] The sequence image standardization is as follows: for multiple continuous images of the drone, one with the highest similarity to the infrared image or the laser image is selected as the standard image, and the other images are projected onto the standard image viewing angle and then spliced ​​into the standard image.

[0014] The step ② specifically includes the following steps:

[0015] A. Region division: Perform image processing and divide the multi-channel image into multiple regions according to different color rendering;

[0016] B. Region-by-region identification: Use the geological identification model to determine the probability of geological type in each region;

[0017] C. Summarize into a table: combine regions and probabilities into a partition probability table.

[0018] The image similarity is calculated using a perceptual hash algorithm.

[0019] In step A, the watershed algorithm is used to segment the image into multiple regions.

[0020] In the sequence image standardization, the infrared image or the laser image is also projected onto the standard image viewing angle.

[0021] In the step ③, the partition probability table and the reverse screening are to delete the item with the lower probability of the corresponding geological type in the partition probability table, if the probability of the geological types in adjacent areas being adjacent is lower than the threshold.

[0022] In the step ③, the partition probability table is screened by deleting the corresponding geological type probability items in the partition probability table whose probability of the regional geological type is lower than a threshold.

[0023] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned geological image recognition and marking method based on images collected by a drone.

[0024] The beneficial effects of the present invention are as follows: by integrating multiple images into multi-channel images and performing partitioned identification, multiple data sources can be effectively integrated, the accuracy of geological type identification can be greatly improved, and most environmental interference can be eliminated based on the integration, thereby providing effective assistance to geological exploration, and then providing effective geological exploration references, reducing the consumption of manpower and material resources, and being particularly beneficial to improving the work efficiency of geological exploration in uninhabited areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram of a process of at least one embodiment of the present invention. DETAILED DESCRIPTION

[0026] The technical solution of the present invention is further described below, but the scope of protection claimed is not limited to the description.

[0027] The first embodiment of the present invention relates to a geological image recognition and marking method based on images collected by drones, such as Figure 1 As shown, the following steps are included:

[0028] ① Multi-source data acquisition: Obtain visual images, infrared images, and / or laser images of a specific area from the database and integrate the acquired images into a single multi-channel image;

[0029] ②Multi-region recognition type: divide the multi-channel image into regions and geological types according to probability and identify them to obtain a partition probability table;

[0030] ③ Primary screening: Screening the partition probability table based on the probability of adjacent geological types;

[0031] ④ Secondary screening: Screening the zoning probability table based on the probability of possible regional geological types;

[0032] ⑤ Map marking: Mark the partition probability table that has undergone secondary screening by region on the map and output it for display.

[0033] Generally, the operation of integrating the acquired images into a single multi-channel image in step ① is to add a new color channel as the color channel of the infrared image or laser image on the basis of the multiple color channels (commonly three color channels) of the visible image, thereby integrating multiple color information on one image to facilitate subsequent processing. At this time, the color channels of the single multi-channel image (assuming that the color channels of the visible image are RBG) are R, G, B, R0 (infrared channel), and J (laser channel).

[0034] It is easy to understand that since the image in step ① is targeted (i.e., targeted at a specific area), the location information is attached to the image file when the image is taken, and it is easy to mark the information in the image on the map based on the location information.

[0035] The steps of the various methods above are divided only for the purpose of clear description. During implementation, they can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this patent. Adding insignificant modifications or introducing insignificant designs to the algorithm or process without changing the core design of the algorithm and process are all within the scope of protection of this patent.

[0036] The second embodiment of the present invention is essentially the same as the first embodiment, primarily with respect to the preferred embodiment of visual images. Visual images are obtained by normalizing multiple consecutive images of a specific area captured by a drone through a sequence of normalized images. Because visual images provide the richest information and are relatively abundant, a drone can typically capture multiple consecutive images in a single aerial photo, which can be sorted by timestamp to create a sequence of images.

[0037] It is not difficult to find that this embodiment is a system embodiment corresponding to the first embodiment, and this embodiment can be implemented in conjunction with the first embodiment. The relevant technical details mentioned in the first embodiment are still valid in this embodiment, and to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the first embodiment.

[0038] Furthermore, image sequence standardization involves selecting the most similar image to the infrared or laser image from multiple consecutive drone images as the standard image. The remaining images are then projected onto the standard image's viewing angle and then stitched together. When drones are taking aerial photos, information such as the shooting angle and coordinates can be included in the image file information, making it easy to perform projective transformations based on this information.

[0039] Preferably, the image similarity is calculated using a perceptual hashing algorithm. Tests have shown that when images are subject to possible rotation or scaling transformations, the perceptual hashing algorithm has the highest accuracy compared to other classic image similarity calculation methods.

[0040] Furthermore, in the standardization of sequential images, infrared or laser images are projected onto a standard image perspective. Similarly, in geological exploration, infrared and laser images are typically obtained from drones. Information such as the shooting angle and coordinates are also included in the infrared and laser image files. Therefore, similar methods can be used to project infrared and laser images.

[0041] The third embodiment of the present invention is substantially the same as the first embodiment, except for the preferred solution of step ②, which specifically includes the following steps:

[0042] A. Region division: Perform image processing and divide the multi-channel image into multiple regions according to different color rendering;

[0043] B. Region-by-region identification: Use the geological identification model to determine the probability of geological type in each region;

[0044] C. Summarize into a table: combine regions and probabilities into a partition probability table.

[0045] It is easy to understand that segmentation by color means that after the image is grayscaled, the image grayscale value is the color value, and segmentation is performed based on the image grayscale value.

[0046] Generally, the geological recognition model is an image recognition and classification model, which can directly adopt the image classification model in the existing technology such as VGGNet, and make the model output a multi-classification probability table.

[0047] Furthermore, in step A, the watershed algorithm is used to segment the image into multiple regions. Obviously, aerial images have a certain degree of blur. This blur is sometimes not entirely caused by camera accuracy. In the solution of this application, it is more caused by differences in shooting posture and the transformation operation in step ①. Therefore, when performing image segmentation, the watershed algorithm can better take into account this blur while maintaining high efficiency.

[0048] The fourth embodiment of the present invention is basically the same as the first embodiment, mainly in that, in step ③, the partition probability table and the reverse screening are performed to delete the item with a lower probability of geological type in the partition probability table, where the probability of adjacent geological types in adjacent areas is lower than a threshold.

[0049] For example, if Figure 1 The partition probability table shown in the figure is used for screening. Partitions 1 and 2 are adjacent. Among the probabilities of adjacent geological types in adjacent areas, the probability of class 1 and class 2 being adjacent is 0.02%. The threshold is set to 5%. The probability of class 1 in partition 1 is greater than the probability of class 2 in partition 2. Therefore, the probability value of class 2 in partition 2 is deleted and becomes a null value.

[0050] Furthermore, in step ③, the partition probability table is screened by deleting the corresponding geological type probability items in the partition probability table whose probabilities of the regional geological types are lower than a threshold.

[0051] For example, if Figure 1 The partition probability table shown in is screened. Among the probabilities of possible geological types in the region, the probability of class 2 is 0.1%, and the threshold is 1%. Therefore, the class 2 probability values ​​of all partitions in the partition probability table are deleted.

[0052] The fifth embodiment of the present invention relates to a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the method for identifying and marking geological images based on images collected by a drone is implemented.

[0053] That is, those skilled in the art will understand that all or part of the steps in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a program, which is stored in a storage medium and includes a number of instructions for causing a device (which may be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., various media that can store program code.

[0054] Those skilled in the art will appreciate that the above-mentioned embodiments are specific examples for implementing the present invention, and that in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present invention.

Claims

1. A geological image recognition and marking method based on images collected by unmanned aerial vehicles, characterized in that: The steps include: ① Multi-source data acquisition: Obtain visible images, infrared images and / or laser images of a specific area from the database, and integrate the acquired images into a single multi-channel image; ②Multi-zone recognition type: divide the multi-channel image into regions and geological types according to probability and identify them to obtain a partition probability table; ③ Primary screening: Screening the partition probability table based on the probability of adjacent geological types; ④ Secondary screening: Screening the zoning probability table based on the probability of possible regional geological types; ⑤ Map marking: Mark the partition probability table that has undergone secondary screening on the map by region and output it for display.

2. The geological image recognition and marking method based on unmanned aerial vehicle collected images according to claim 1, characterized in that: The visible image is obtained by normalizing a plurality of continuous images of a specific area collected by a drone through sequence image standardization.

3. The geological image recognition and marking method based on unmanned aerial vehicle collected images according to claim 2, characterized in that: The sequence image standardization is as follows: for multiple continuous images of the drone, one image with the highest similarity to the infrared image or the laser image is selected as the standard image, and the other images are projected at the standard image viewing angle and then spliced ​​into the standard image.

4. The geological image recognition and marking method based on unmanned aerial vehicle collected images according to claim 1, characterized in that: The step ② specifically includes the following steps: A. Region division: Perform image processing and divide the multi-channel image into multiple regions according to different color rendering; B. Area-by-area identification: Use the geological identification model to determine the probability of geological type in each area; C. Summarize into a table: combine regions and probabilities into a partition probability table.

5. The geological image recognition and marking method based on unmanned aerial vehicle collected images according to claim 3, characterized in that: The image similarity is calculated using a perceptual hashing algorithm.

6. The geological image recognition and marking method based on unmanned aerial vehicle collected images according to claim 4, characterized in that: In step A, the watershed algorithm is used to segment the image into multiple regions.

7. The geological image recognition and marking method based on unmanned aerial vehicle collected images according to claim 3, characterized in that: In the standardization of the sequence images, the infrared image or the laser image is also projected onto the standard image viewing angle.

8. The geological image recognition and marking method based on unmanned aerial vehicle collected images according to claim 1, characterized in that: In the step ③, the partition probability table and the reverse screening are to delete the item with a lower probability of the corresponding geological type in the partition probability table, where the probability of the adjacent geological types in the adjacent areas is lower than the threshold.

9. The geological image recognition and marking method based on unmanned aerial vehicle collected images according to claim 1, characterized in that: In the step ③, the partition probability table is screened by deleting the corresponding geological type probability items in the partition probability table whose probability of the regional possible geological type is lower than the threshold.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the geological image recognition and marking method based on images collected by a drone according to any one of claims 1 to 9 is implemented.

Images