Red fire ant nest recognition method based on UAV hyperspectral imaging

Abstract

The invention provides a method for identifying a red fire ant nest based on unmanned aerial vehicle hyperspectral imaging, which belongs to the field of pest control technology and includes the following steps: S1 obtains hyperspectral image data within a target range; S2 preprocesses the hyperspectral image data to obtain an intermediate data set; S3 extracts the target position of the red fire ant nest from the intermediate data set by a hierarchical classification method. The invention analyzes the hyperspectral image data by a hierarchical classification method, efficiently and automatically identifies the position of the red fire ant nest, is simple to operate, has strong timeliness, and can save a lot of manpower and material resources.

Description

Technical Field

[0001] The present invention relates to the technical field of pest control, and in particular to a method for identifying a red fire ant nest based on unmanned aerial vehicle hyperspectral imaging. Background Art

[0002] Red fire ants are major agricultural and medical pests originating from South America. They are a species that has invaded China in recent years. Red fire ants have complex diets, ferocious habits, rapid reproduction, and strong competitiveness. They pose a serious threat to human health, public safety, agricultural and forestry production, and the ecological environment in the invaded areas. Prevention and control measures should be taken in a timely manner. Therefore, it is of great significance to carry out rapid detection of red fire ant nests.

[0003] At present, the method of identifying red fire ant nests in the wild is generally to observe with the naked eye, first compare the color and shape of the target soil pile with the surrounding ordinary soil, then use hard objects to disturb the target soil pile, and then judge based on the characteristics of the red fire ant nest and the behavior of the red fire ants actively attacking intruders. If this method is used to conduct a red fire ant epidemic investigation, it will be time-consuming and labor-intensive. Summary of the invention

[0004] The technical problem of the present invention is to provide a method for identifying a red fire ant nest based on unmanned aerial vehicle hyperspectral imaging, which can automatically identify the red fire ant nest at a long distance and in a non-contact manner.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] The method for identifying a red fire ant nest based on unmanned aerial vehicle hyperspectral imaging includes the following steps: S1 obtains hyperspectral image data within a target range; S2 preprocesses the hyperspectral image data to obtain an intermediate data set; S3 extracts the target position of the red fire ant nest from the intermediate data set by a hierarchical classification method.

[0007] Furthermore, S3 is: extracting spectra of wet ant nest soil, dry ant nest soil, normal soil, vegetation, and tree trunks from the intermediate data through a hierarchical classification method for analysis to extract the target position.

[0008] Furthermore, the factors of the hierarchical classification method include: MNF transformation, plant senescence reflectance index and spectral angle matching method.

[0009] Further, S3 includes: S31 performing MNF transformation processing on the intermediate data set to obtain a second intermediate data set; S32 selecting a target component from the second intermediate data set, and judging the data in the second intermediate data set based on the target component and the plant aging reflectance index and the spectral angle to obtain the target position.

[0010] Further, S32 includes: S321 determines whether the target component is greater than a preset value, if so, enters S322 for calculation; if not, sets it to a null value and does not participate in the next calculation; S322 determines whether the spectral angle of the data that meets S321 is less than or equal to the preset angle, if so, enters S323 for calculation; if not, sets it to a null value and does not participate in the next calculation; S323 determines whether the plant aging reflectance index of the data that meets the conditions of S323 is greater than the preset index, if so, calibrates; if not, sets the judgment result to a null value; S324 loops through S322 and S323 until the data in the second intermediate data set that meets S321 is traversed to obtain the target position.

[0011] Furthermore, the preprocessing includes: mirror transformation, black and white frame correction, atmospheric correction and image stitching. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The present invention and its features, configurations and advantages will become more apparent by reading the detailed description of non-limiting embodiments with reference to the following drawings. The same reference numerals indicate the same parts throughout the drawings. The drawings are not drawn to scale, but rather to illustrate the subject matter of the present invention.

[0013] Figure 1 It is a schematic diagram of a method for identifying a red fire ant nest based on unmanned aerial vehicle hyperspectral imaging provided by the present invention;

[0014] Figure 2 It is a spectral reflectance curve diagram of red fire ant nest soil and surrounding objects provided by the present invention. DETAILED DESCRIPTION

[0015] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0016] With the deepening of globalization, various means of transportation have been continuously improved, transportation capacity has been greatly improved, and the Internet has been further popularized. The whole world has become more and more a close whole. Capital globalization, production globalization, trade globalization, information globalization, etc. are the main themes of today's world. Behind this upward and positive era background, it has also built a huge bridge for species invasion, so that the activity range of species is no longer limited to a corner, but spreads and spreads to the world. The invasion of alien species, as we all know, has caused a great impact on the original balance of the local ecosystem, and in severe cases, it will cause ecological disasters. However, this invasion of red fire ants is also a typical case of alien species invasion, which not only causes serious harm to our ecology, but also threatens people's life, health and safety.

[0017] In response to the invasion problem of red fire ants, the present invention adopts the DJI M600 Pro six-rotor drone. The drone has a net weight of about 4kg and a maximum load of about 10kg. The hyperspectral imager GaiaSky-mini2 of Sichuan Shuangli Hepu Technology Co., Ltd. is installed on the M600Pro drone platform. The drone is hovered in the air, and the built-in push-scan function of the hyperspectral imaging spectrometer is used to obtain hyperspectral image data of the target ground. In this embodiment, the hyperspectral image data acquisition time is 11:30 noon on October 9, 2021. The weather is clear. The flight altitude of the drone is 30m, the exposure time of the hyperspectral imager is 1.13ms, the gain is 1, the spatial resolution of the hyperspectral image data is 0.02m, the image resolution is 696*736, and the number of bands is 256.

[0018] Then, the obtained hyperspectral image data is preprocessed in SpecView software, including mirror transformation, black and white frame calibration, atmospheric correction and image stitching. Among them, the mirror transformation is to correct the influence of internal image distortion caused by built-in push-scan on stitching; the purpose of black and white frame calibration is to convert DN value (digital quantization value) into reflectance value (reflectance). The specific formula is as follows:

[0019]

[0020] In the formula, R ref is the reflectance value of the black and white corrected image, DN raw is the DN value of the original image, DN white is the white frame data of the reference board, DN dark is the camera’s system error DN value, R white is the reflectivity coefficient of the reference plate.

[0021] Considering that after the UAV flies to a certain altitude, the hyperspectral image data obtained by the hyperspectral imager may be affected by factors such as atmosphere and water vapor. In order to eliminate the influence of these factors, before the UAV takes off, a 2m*2m gray cloth calibrated by the National Institute of Metrology is placed in the shooting area. When the hyperspectral image is acquired, only one of the hyperspectral images needs to be covered with the gray cloth to achieve atmospheric correction.

[0022] The method to eliminate the influence of factors such as atmosphere and water vapor is shown in Formula 2.

[0023]

[0024] In the formula, R fixed is the image spectral reflectance after eliminating factors such as atmosphere and water vapor, R refis the image reflectance after black and white correction, R standard is the spectral reflectance of the gray cloth calibrated by the National Institute of Metrology, R grayref is the spectral reflectance of the gray cloth in the image after black and white correction.

[0025] Then, the SpecStitcher software developed by Jiangsu Shuangli Hepu Technology Co., Ltd. was used to stitch the atmospherically corrected UAV hyperspectral image data to obtain the intermediate data set.

[0026] Select the spectrum of typical objects on the stitched hyperspectral image for analysis, such as Figure 2 As shown in the figure, the reflectance waveforms of wet and dry red fire ant nest soil and ordinary soil are similar. In the near-infrared range, the spectral reflectance of wet and dry ant nest soil is higher than that of ordinary soil, and the spectral reflectance of wet ant nest soil is higher than that of dry ant nest soil. There are significant differences between the spectral reflectance curves of ant nest soil and ordinary soil and other landforms such as vegetation and tree trunks. The spectral reflectance value of tree trunks in the range of 400-1000nm is lower than that of ant nest soil and ordinary soil. Vegetation has obvious "peaks" and "valleys" in the visible light range, and obvious "edges" in the range of 700-1000nm.

[0027] For the spliced ​​UAV hyperspectral image data, a hierarchical classification method was constructed using MNF transformation (Minimum Noise Fraction Rotation, MNF Rotation), spectral angle matching method, Plant Senscence Reflectance Index (PSRI), etc. to efficiently extract the location information of red fire ant nests in UAV hyperspectral images. Figure 1 As shown, the intermediate data set is subjected to MNF transformation processing to obtain a second intermediate data set; the target component is selected from the second intermediate data set, and the data in the second intermediate data set is judged based on the target component, the plant aging reflectance index and the spectral angle to obtain the target position. Specifically, it is judged whether the target component is greater than the preset value, if so, the spectral angle is calculated; if not, it is set to a null value and does not participate in the next calculation; it is judged whether the spectral angle of the data that meets the requirements of the target component is less than or equal to the preset angle, if so, the plant aging reflectance index is calculated; if not, it is set to a null value and does not participate in the next calculation; it is judged whether the plant aging reflectance index of the data that meets the requirements of the target component and the spectral angle is greater than the preset index, if so, it is calibrated; if not, the judgment result is set to a null value; the image position that meets the requirements of the target component, the spectral angle and the plant aging reflectance index can be judged as the position of the red fire ant nest.

[0028] In this embodiment, MNF Rotation selects the fifth component MNF greater than 5 as the judgment condition. When MNF5 is greater than 5, the next calculation is entered. When it is less than or equal to 5, it is taken as a null value and does not participate in the next calculation; the spectral angle matching method selects the maximum angle of the ant nest soil spectrum as 0.13 as the judgment condition. When the spectral angle is less than or equal to 0.13, the next calculation is entered. When it is greater than 0.13, it is taken as a null value and does not participate in the next calculation; the plant aging reflectance index selects greater than 0.2 as the judgment condition. When the plant aging reflectance index is greater than 0.2, the next calculation is entered. When the plant aging reflectance index is less than or equal to 0.2, it is taken as a null value and does not participate in the next calculation. The image position that comprehensively meets the fifth component MNF greater than 5, the spectral angle less than or equal to 0.13 and the plant aging reflectance index greater than 0.2 can be judged as the red fire ant nest position.

[0029] In summary, the present invention proposes a red fire ant nest identification method based on unmanned aerial vehicle hyperspectral imaging, which can realize the long-distance identification of the spatial distribution of red fire ant nests, making the red fire ant disaster monitoring faster and more efficient, and the operation is simple and timely, which can save a lot of manpower and material resources.

[0030] The preferred embodiments of the present invention are described above; it should be understood that the present invention is not limited to the above-mentioned specific embodiments, and the devices and structures not described in detail should be understood to be implemented in a common manner in the art; any technician familiar with the art can make many possible changes and modifications without departing from the technical solution of the present invention, or modify them into equivalent embodiments with equivalent changes, which does not affect the essential content of the present invention; therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of protection of the technical solution of the present invention.

Claims

1. A method for identifying red fire ant nests based on UAV hyperspectral imaging, characterized in that: The following steps are involved: S1 obtains hyperspectral image data within the target range; S2 preprocesses the hyperspectral image data to obtain an intermediate data set; S3 extracts the target location of the red fire ant nest from the intermediate data set by a hierarchical classification method; The S3 is: Extracting the spectra of wet ant nest soil, dry ant nest soil, normal soil, vegetation, and tree trunks from the intermediate data through a hierarchical classification method for analysis to extract the target location; The factors of the hierarchical classification method include: MNF transformation, plant senescence reflectance index and spectral angle matching method; The S3 includes: S31 performs MNF transformation on the intermediate data set to obtain a second intermediate data set; S32: selecting a target component from the second intermediate data set, and judging the data in the second intermediate data set based on the target component and the plant senescence reflectance index and the spectral angle to obtain the target position; The S32 includes: S321 determines whether the target component is greater than a preset value. If so, proceed to S322 for calculation; if not, set it to a null value and do not participate in the next step of calculation; S322 determines whether the spectral angle of the data in S321 is less than or equal to the preset angle. If so, proceed to S323 for calculation; if not, set it to a null value and do not participate in the next step of calculation; S323 determines whether the plant aging reflectance index of the data meeting the condition of S322 is greater than a preset index, and if so, performs calibration; if not, sets the determination result to a null value; S324 loops through S322 and S323 until the data in the second intermediate data set that satisfies S321 is traversed to obtain the target position; MNF Rotation selects the fifth component MNF greater than 5 as the judgment condition; the spectral angle matching method selects the maximum angle of the ant nest soil spectrum as 0.13 as the judgment condition; the plant aging reflectance index selects greater than 0.2 as the judgment condition.

2. The method for identifying red fire ant nests based on unmanned aerial vehicle hyperspectral imaging according to claim 1, characterized in that: The preprocessing includes: mirror transformation, black and white frame correction, atmosphere correction and image stitching.

Images