A method and device for obtaining wildlife information based on drones

By performing classification preprocessing and similarity detection on the images collected by the drone, combined with deep learning models, the problem of inaccurate acquisition of wildlife information in the prior art is solved, and the recognition accuracy is significantly improved.

CN118522038BInactive Publication Date: 2024-11-15ACAD OF NATURAL SCI ENVIRONMENTAL TECH DEV (TIANJIN) CO LTD +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410969551.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, the drone-based wildlife information acquisition method has the problem of insufficient identification in the image acquisition stage, especially due to the different number of wildlife and the impact of occlusion.

Method used

After the images are collected by drones, the images are pre-processed, and divided into single wildlife images and multiple wildlife images. The similarity threshold is set according to the body shape of the wildlife for rough detection, combining binarization and sharpening treatment. Finally, a deep learning model is input to improve recognition accuracy.

Benefits of technology

Through highly targeted preprocessing processes and appropriate threshold settings, the accuracy of image recognition is significantly improved, laying a solid data foundation for subsequent wildlife information acquisition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118522038B_ABST
    Figure CN118522038B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of wildlife data acquisition technology, and in particular to a wildlife information acquisition method and device based on drones. In the wildlife image preprocessing stage, the present invention first classifies the image to be detected into two types: single wildlife image and multiple wildlife images, and then adopts different preprocessing processes to effectively identify and improve the pertinence and effectiveness of image processing in the preprocessing stage, laying a solid data foundation for subsequent wildlife acquisition; at the same time, when the present invention adopts similarity comparison to roughly detect the wildlife image, different thresholds are set according to the body shape characteristics of the wildlife to be detected. If the wildlife to be detected is a large wildlife, a smaller comparison threshold is set, and if the wildlife to be detected is a small wildlife, a larger comparison threshold is set, so that the classification result obtained by the similarity comparison method can be as accurate as possible.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of wild animal data acquisition, and in particular to a method and device for acquiring wild animal information based on an unmanned aerial vehicle. Background Art

[0002] As human society continues to develop, nature is being exploited without restraint by humans, and various industrial and domestic wastes are discharged into the natural environment, exceeding the self-purification capacity of the ecosystem. The natural environment has gradually deteriorated, leading to a sharp deterioration in the living environment of wild animals. In addition, the living environment of wild animals faces various threats. Wild animals are of vital importance in maintaining the balance of the natural ecosystem and maintaining the diversity of biological species.

[0003] In recent years, with the development of science and technology, the way of image acquisition has become more diverse. From the earliest handheld shooting to the later satellite-generated images, the channels and quality of image acquisition have been constantly updated and iterated. One of the most important developments is the in-depth research of unmanned aerial vehicles (UAVs), and the field of drones has developed rapidly. Due to the advantages of convenient carrying, simple operation, and flexible flight, drones are widely used in surveying and mapping, remote sensing, inspection, and landmark detection. UAVs have a unique imaging perspective, which can make up for the shortcomings of the above technologies, and can be better combined with the above technologies and give birth to more diversified application scenarios. Carrying cameras on drones can obtain images and scenes that are difficult to capture with ground cameras. With the development of science and technology, especially the optimization of algorithms in the field of vision and the improvement of hardware computing power, image detection has ushered in a new milestone. With the continuous progress of deep learning research, the combination of drones and intelligent detection of aerial animal groups has greatly reduced the cost of ecological detection while improving the convenience and accuracy of ecological detection, providing new ideas and methods for protecting wildlife.

[0004] In the prior art, when using deep learning models to identify wildlife information, the image to be detected is generally pre-processed with a fixed process and then input into the deep learning model to obtain information such as the number of wild animals. However, during the image acquisition stage, due to the different number of wild animals that may exist in the image and the occlusion of wild animals, the above solution has the effect of inaccurate recognition.

[0005] Therefore, the prior art urgently needs a method and device for obtaining wildlife information based on drones with improved accuracy. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a method and device for obtaining wildlife information based on drones with improved accuracy in view of the shortcomings of the above technical solutions.

[0007] In order to achieve the above object, according to one aspect of the present invention, a method for obtaining wildlife information based on a drone is provided, comprising the following steps:

[0008] S1: Use drones to collect images of the area to be inspected;

[0009] S2: performing a preprocessing operation on the image;

[0010] S2.1: performing a rough wildlife image detection on the image, and dividing the image into a single wildlife image and a multiple wildlife image;

[0011] S2.2: performing a first image preprocessing operation on the single wild animal image;

[0012] S2.3: performing a second image preprocessing operation on the multiple wildlife images;

[0013] S3: Input the image pre-processed by S2 into the deep learning model to obtain the quantity information of wild animals.

[0014] Preferably, in S2, the first image preprocessing is to perform binarization processing on the single wild animal image;

[0015] The second preprocessing operation is specifically:

[0016] S2.3.1: performing binarization processing on the plurality of wildlife images;

[0017] S2.3.2: Sharpening the image pre-processed in S2.3.1.

[0018] Preferably, in S1, the drone is a quad-rotor drone, and the quad-rotor drone is equipped with a communication device for realizing wireless communication with a workstation.

[0019] Preferably, in S2.1, the wild animal image is roughly detected by similarity comparison.

[0020] Preferably, S2.1 is specifically as follows: if there is an area in the image to be detected whose similarity with the standard image of wild animals is greater than a set threshold, the image to be detected is considered to be a single wild animal image; if there is more than one area in the image to be detected whose similarity with the standard image of wild animals is greater than the set threshold, the image to be detected is considered to be a multiple wild animal image.

[0021] Preferably, the size of the set threshold is determined according to the body size of the wild animal to be detected.

[0022] Preferably, if the wild animal to be detected is a large wild animal, the preset threshold is 30%; if the wild animal to be detected is a medium-sized wild animal, the preset threshold is 38%; if the wild animal to be detected is a small wild animal, the preset threshold is 47%.

[0023] Preferably, in S2.3.2, the sharpening process is specifically: using a convolution kernel to perform a convolution operation on the image preprocessed by S2.3.1, and then performing a difference operation between the image after the convolution operation and the image preprocessed by S2.3.1, so as to achieve the sharpening process.

[0024] Preferably, the deep learning model is a convolutional neural network model, the input of the convolutional neural network model is the image preprocessed by S2, and the output of the convolutional neural network model is the number of wild animals in the image.

[0025] According to another aspect of the present invention, a device for obtaining wildlife information based on a drone is provided. The device adopts the above-mentioned method for obtaining wildlife information based on a drone, and the device comprises:

[0026] UAV, used to collect images of the area to be inspected;

[0027] An image preprocessing module, used for performing a preprocessing operation on the image;

[0028] The image recognition module is used to input the image processed by the image preprocessing module into the deep learning model to obtain the quantity information of wild animals.

[0029] Based on the above technical solution, the present application provides a method and device for obtaining wildlife information based on drones, which has the following technical effects:

[0030] In the wildlife image preprocessing stage of the present invention, the images to be detected are first classified into two types: single wildlife images and multiple wildlife images. Then different preprocessing processes are adopted to effectively improve the pertinence and effectiveness of image processing in the preprocessing stage, laying a solid data foundation for subsequent wildlife acquisition.

[0031] At the same time, when the present invention uses similarity comparison to roughly detect the wild animal image, different thresholds are set according to the body shape characteristics of the wild animal to be detected. If the wild animal to be detected is a large wild animal, a smaller comparison threshold is set; if the wild animal to be detected is a small wild animal, a larger comparison threshold is set, so that the classification result obtained by the similarity comparison method is as accurate as possible. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 A flowchart of a method for obtaining wildlife information based on a drone is provided in an embodiment of the present application. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0035] The following first describes the concepts involved in this application in conjunction with the accompanying drawings. It should be noted that the following description of each concept is only to make the content of this application easier to understand and does not limit the scope of protection of this application.

[0036] Embodiment 1: Figure 1 As shown, a method for obtaining wildlife information based on a drone includes the following steps:

[0037] S1: Use drones to collect images of the area to be inspected;

[0038] Wherein, the UAV is a quad-rotor UAV, and the quad-rotor UAV is equipped with a communication device for realizing wireless communication with the workstation;

[0039] In addition, the camera is an important component of the drone for acquiring wildlife information. In this embodiment, the camera of the drone is a 10-megapixel binocular camera, which can capture high-definition images of the area to be detected, providing a good material basis for subsequent wildlife information acquisition.

[0040] It is worth noting that when the drone collects images of the area to be detected, the following requirements should be observed:

[0041] (1): Safety. UAVs should comply with laws and regulations when flying. Each time an image acquisition mission is performed, it is necessary to determine whether there is a no-fly zone in the area to be detected, and set the flight altitude to meet the low-altitude (below 500m) requirements;

[0042] (2): Stability. The drone must have excellent flight performance and remain stable during shooting to ensure that enough images can be captured.

[0043] S2: performing a preprocessing operation on the image;

[0044] In the field of image processing, the quality of image preprocessing directly affects the subsequent interpretation of image information. Therefore, during image preprocessing, this embodiment improves the existing image preprocessing method in order to obtain a more accurate information acquisition effect.

[0045] In this embodiment, the preprocessing operation performed on the image is specifically:

[0046] S2.1: performing a rough wildlife image detection on the image, and dividing the image into a single wildlife image and a multiple wildlife image;

[0047] In this embodiment, the wildlife image is roughly detected by similarity comparison;

[0048] Specifically, S2.1 is as follows: if there is a region in the image to be detected whose similarity with the wild animal standard image is greater than a set threshold, the image to be detected is considered to be a single wild animal image; if there are more than one region in the image to be detected whose similarity with the wild animal standard image is greater than the set threshold, the image to be detected is considered to be a multiple wild animal image;

[0049] Exemplarily, if the wild animal to be detected is a clouded leopard, an image of an adult clouded leopard is first obtained as a standard image of a wild animal, and then the image to be detected is compared with the standard image of a wild animal, so as to roughly classify the animal image into two types: one or more wild animals may exist in the image;

[0050] It is worth noting that in this step, a relatively simple similarity algorithm is used to identify the number of wild animals as 1 or other numbers. In essence, there is still a certain error. Therefore, this step is a rough detection result, and subsequent more precise image recognition steps are required to obtain convincing information such as the number of wild animals.

[0051] In this embodiment, the size of the set threshold is determined according to the size of the wild animal to be detected;

[0052] Specifically, if the wild animal to be detected is a large wild animal, the preset threshold is 30%; if the wild animal to be detected is a medium-sized wild animal, the preset threshold is 38%; if the wild animal to be detected is a small wild animal, the preset threshold is 47%;

[0053] Among them, if the wild animal to be detected is a large wild animal, first of all, the pixel information contained in its wild animal standard image can reflect more information that distinguishes the wild animal from the environment. In addition, when the wild animal is large in size and multiple wild animals live together, it is more likely that some wild animals will be obscured by other wild animals or the environment. For the case of small wild animals, on the one hand, there is relatively less information to distinguish it from the environment, and on the other hand, there are relatively few cases of being obscured by other wild animals or the environment. As described, this embodiment sets different thresholds according to the body size of the wild animal, in order to classify the image as much as possible.

[0054] S2.2: performing a first image preprocessing operation on the single wild animal image;

[0055] Among them, since a single wild animal image may only contain one wild animal image, a relatively simple preprocessing process is sufficient for the wild animal image; the first image preprocessing is to binarize the single wild animal image; since the binocular camera is an RGB color picture, it contains three-channel color information, but in the image recognition process, an overly complex image matrix is ​​not conducive to image recognition, so the image is converted into a single-channel binary image to simplify the preprocessing process.

[0056] The binarization process is as follows: converting the single wild animal image into a grayscale domain, and then using a pixel threshold method to set the image with a grayscale value higher than a set grayscale threshold to 1, and setting the image with a grayscale value not higher than the set grayscale threshold to 0, thereby completing the image binarization process;

[0057] S2.3: performing a second image preprocessing operation on the multiple wildlife images;

[0058] The second preprocessing operation is specifically:

[0059] S2.3.1: performing binarization processing on the plurality of wildlife images;

[0060] The binarization process is the same as the binarization process in S2.2;

[0061] S2.3.2: sharpening the image pre-processed in S2.3.1;

[0062] The sharpening process is used to enhance the detail information in the multiple wild animal images, making the multiple wild animal images clearer and more structured; specifically, the sharpening process is: using a convolution kernel to perform a convolution operation on the image pre-processed by S2.3.1, and then performing a difference operation between the image after the convolution operation and the image pre-processed by S2.3.1, thereby achieving sharpening processing;

[0063] In this step, in the preprocessing stage, this embodiment first classifies the images to be detected into two types: single wildlife images and multiple wildlife images. Then, different preprocessing processes are adopted to effectively identify the pertinence and effectiveness of the preprocessing stage for image processing, laying a solid data foundation for subsequent wildlife acquisition.

[0064] S3: inputting the image pre-processed by S2 into a deep learning model to obtain the quantity information of wild animals;

[0065] In this step, the deep learning model is a convolutional neural network model. The convolutional neural network model (CNN) is a special neural network and a typical algorithm in deep learning. The convolutional neural network model includes structures such as convolutional layer, pooling layer, activation function and fully connected layer. Among them, the convolutional layer is the core component of the convolutional neural network, which is used to extract local features of the image. The emergence of convolutional neural networks has qualitatively improved the performance of computers in fields such as image recognition, and has become one of the important research directions in the field of deep learning.

[0066] The convolution layer is the core part of the convolutional neural network model. It consists of multiple filters, each of which is a small matrix. The convolution layer extracts features by applying these filters to each local area of ​​the image preprocessed by S2. Each filter slides on the input data, calculates the convolution operation between the filter and the input data, and generates an output feature map, which is input to the next layer for further processing and analysis.

[0067] In S3, the input of the convolutional neural network model is the image preprocessed by S2, and the output of the convolutional neural network model is the number of wild animals in the image;

[0068] The convolutional neural network model requires a large amount of data set training to obtain accurate results. In this step, the Snapshot Serengeti data set is used as the training set of the convolutional neural network model. This training set is the largest image collection in the field of wildlife machine recognition, containing approximately 7.1 million images, and has been accurately labeled with high annotation accuracy.

[0069] Embodiment 2: This embodiment includes a wild animal information acquisition device based on a drone, the device adopts the wild animal information acquisition method based on a drone of embodiment 1, and the device includes:

[0070] UAV, used to collect images of the area to be inspected;

[0071] An image preprocessing module, used for performing a preprocessing operation on the image;

[0072] An image recognition module is used to input the image processed by the image preprocessing module into a deep learning model to obtain the quantity information of wild animals.

[0073] Embodiment 3: This embodiment includes a computer-readable storage medium on which a data processing program is stored. The data processing program is executed by a processor to implement the method for obtaining wildlife information based on a drone of embodiment 1.

[0074] It should be understood by those skilled in the art that the embodiments of this article can be provided as methods, devices (equipment), or computer program products. Therefore, this article can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Including but not limited to RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassette, tape, disk storage or other magnetic storage device, or any other medium that can be used to store desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that communication media generally contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0075] This document is described with reference to flowcharts and / or block diagrams of methods, apparatuses (devices), and computer program products according to the embodiments of this document. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0076] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0077] The embodiments and / or implementation methods described above are only used to illustrate the preferred embodiments and / or implementation methods for realizing the technology of the present invention, and are not intended to limit the implementation methods of the technology of the present invention in any form. Any technical personnel in this field may make slight changes or modifications to other equivalent embodiments without departing from the scope of the technical means disclosed in the content of the present invention, but they should still be regarded as technologies or embodiments that are essentially the same as the present invention.

Claims

1. A method for obtaining wildlife information based on drones. It is characterized in that The following steps are involved: S1: Use drones to collect images of the area to be inspected; S2: performing a preprocessing operation on the image; S2.1: Perform a rough detection of wildlife images on the image, and divide the image into a single wildlife image and a multiple wildlife image; in S2.1, the wildlife image is roughly detected by similarity comparison, and S2.1 is specifically as follows: if there is a region in the image to be detected whose similarity with the wildlife standard image is greater than a set threshold, the image to be detected is considered to be a single wildlife image; if there are more than one region in the image to be detected whose similarity with the wildlife standard image is greater than the set threshold, the image to be detected is considered to be a multiple wildlife image; The size of the set threshold is determined according to the size of the wild animal to be detected; if the wild animal to be detected is a large wild animal, the preset threshold is 30%; if the wild animal to be detected is a medium-sized wild animal, the preset threshold is 38%; if the wild animal to be detected is a small wild animal, the preset threshold is 47%; S2.2: performing a first image preprocessing operation on the single wild animal image; S2.3: performing a second image preprocessing operation on the multiple wild animal images; in S2, the first image preprocessing is performing a binarization operation on the single wild animal image; The second image preprocessing operation is specifically: S2.3.1: performing binarization processing on the plurality of wildlife images; S2.3.2: sharpening the image pre-processed in S2.3.1; S3: Input the image pre-processed by S2 into the deep learning model to obtain the quantity information of wild animals.

2. A method for obtaining wildlife information based on drones according to claim 1, It is characterized in that In S1, the drone is a quad-rotor drone, and the quad-rotor drone is equipped with a communication device for realizing wireless communication with a workstation.

3. A method for obtaining wildlife information based on drones according to claim 2, It is characterized in that In S2.3.2, the sharpening process is specifically: using a convolution kernel to perform a convolution operation on the image preprocessed by S2.3.1, and then performing a difference operation between the image after the convolution operation and the image preprocessed by S2.3.1, so as to achieve the sharpening process.

4. The method for obtaining wildlife information based on drone according to claim 1, It is characterized in that The deep learning model is a convolutional neural network model, the input of the convolutional neural network model is the image preprocessed by S2, and the output of the convolutional neural network model is the number of wild animals in the image.

5. A wild animal information acquisition device based on drones, It is characterized in that The device adopts the method for obtaining wildlife information based on a drone according to any one of claims 1 to 4, and the device comprises: UAV, used to collect images of the area to be inspected; An image preprocessing module, used for performing a preprocessing operation on the image; The image recognition module is used to input the image processed by the image preprocessing module into the deep learning model to obtain the quantity information of wild animals.