CN115527130A - Grassland pest mouse density investigation method and intelligent evaluation system - Google Patents
Grassland pest mouse density investigation method and intelligent evaluation system Download PDFInfo
- Publication number
- CN115527130A CN115527130A CN202211148118.XA CN202211148118A CN115527130A CN 115527130 A CN115527130 A CN 115527130A CN 202211148118 A CN202211148118 A CN 202211148118A CN 115527130 A CN115527130 A CN 115527130A
- Authority
- CN
- China
- Prior art keywords
- mouse
- rat
- trace
- density
- grassland
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 60
- 241000607479 Yersinia pestis Species 0.000 title claims abstract description 31
- 238000011156 evaluation Methods 0.000 title claims abstract description 17
- 238000011835 investigation Methods 0.000 title claims description 36
- 238000012544 monitoring process Methods 0.000 claims abstract description 20
- 230000000007 visual effect Effects 0.000 claims abstract description 16
- 238000005070 sampling Methods 0.000 claims abstract description 15
- 230000000694 effects Effects 0.000 claims abstract description 11
- 238000004445 quantitative analysis Methods 0.000 claims abstract description 6
- 238000004364 calculation method Methods 0.000 claims abstract description 4
- 241000700159 Rattus Species 0.000 claims description 156
- 241000699666 Mus <mouse, genus> Species 0.000 claims description 120
- 238000001514 detection method Methods 0.000 claims description 40
- 238000012549 training Methods 0.000 claims description 22
- 238000013135 deep learning Methods 0.000 claims description 21
- 241000699670 Mus sp. Species 0.000 claims description 20
- 238000006243 chemical reaction Methods 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 6
- 241000289702 Peramelidae Species 0.000 claims description 4
- 238000005520 cutting process Methods 0.000 claims description 4
- 238000011002 quantification Methods 0.000 claims description 4
- 244000025254 Cannabis sativa Species 0.000 claims description 3
- 230000000903 blocking effect Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 108700013565 rat Tnfsf10 Proteins 0.000 claims description 3
- 238000005286 illumination Methods 0.000 claims description 2
- 235000019994 cava Nutrition 0.000 claims 1
- 238000001739 density measurement Methods 0.000 claims 1
- 230000006378 damage Effects 0.000 abstract description 21
- 238000004458 analytical method Methods 0.000 abstract description 4
- 230000009467 reduction Effects 0.000 abstract description 4
- 241000589562 Brucella Species 0.000 description 16
- 238000005516 engineering process Methods 0.000 description 12
- 101100369993 Mus musculus Tnfsf10 gene Proteins 0.000 description 8
- 238000000605 extraction Methods 0.000 description 7
- 238000002372 labelling Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 239000002689 soil Substances 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 241000283923 Marmota monax Species 0.000 description 2
- 241001465754 Metazoa Species 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000010835 comparative analysis Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 241000589567 Brucella abortus Species 0.000 description 1
- 208000035473 Communicable disease Diseases 0.000 description 1
- 241000699800 Cricetinae Species 0.000 description 1
- 241000666039 Lasiopodomys brandtii Species 0.000 description 1
- 241000699703 Microtus pennsylvanicus Species 0.000 description 1
- 206010035148 Plague Diseases 0.000 description 1
- 241000283984 Rodentia Species 0.000 description 1
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 239000011324 bead Substances 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000009395 breeding Methods 0.000 description 1
- 230000001488 breeding effect Effects 0.000 description 1
- 229940056450 brucella abortus Drugs 0.000 description 1
- 239000004106 carminic acid Substances 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 244000144972 livestock Species 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 238000001926 trapping method Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/17—Terrestrial scenes taken from planes or by drones
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/10—Image acquisition
- G06V10/16—Image acquisition using multiple overlapping images; Image stitching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Remote Sensing (AREA)
- Image Processing (AREA)
Abstract
The invention relates to the technical field of pest mouse density survey, in particular to a grassland pest mouse density survey method and an intelligent evaluation system. According to the method, the harmful mouse activity trace image is collected by the unmanned aerial vehicle, preprocessed and input into different models preset in the intelligent evaluation system for automatic calculation according to scenes and requirements, and then harmful mouse activity trace information is output. And then, converting the number of the pest mouse activity traces into the mouse density by combining ground sampling survey data, thereby obtaining the mouse density estimation result in the monitoring area and realizing the automatic identification and quantitative analysis of the mouse density. And meanwhile, position reduction is carried out on the map according to the longitude and latitude information of the rat activity trace, so that space presentation, dynamic analysis, rat condition visual display and intelligent monitoring of rat damage in the region are realized. The invention realizes the intelligence, the automation and the high efficiency of the rat density survey, can more intuitively and accurately acquire the rat situation occurrence level, and has stronger practicability and wider application scene.
Description
Technical Field
The invention relates to the technical field of pest mouse density survey, in particular to a grassland pest mouse density survey method and an intelligent evaluation system.
Background
The rat damage is one of the major biological disasters of the grassland, which not only damages and eats the pasture to influence the animal husbandry and the ecological safety, but also can transmit various virulent infectious diseases such as plague and the like, and seriously threatens the life safety of people. The conventional rat density monitoring carried out in spring and autumn every year is the research and judgment basis for accurate assessment and scientific management of the grassland rat damage, but the traditional method is time-consuming and labor-consuming, limited in survey range, irregular in professional background of practitioners and the like, so that the survey precision is restricted, the occurrence level of the rat situation is difficult to accurately and timely reflect, and the development trend is presumed. The existing mouse density survey mode is highly dependent on manual work and is usually used in a certain area (0.25 hm) 2 ) The number of valid rat holes or mounds in the prototype was used as the relative rat density indicator. Therefore, the investigation data can only be counted manually, the investigation proof is lacked, and the authenticity and the accuracy of the data are difficult to guarantee.
Chinese patent document No. CN111881728A discloses a grassland rodent pest monitoring method based on low-altitude remote sensing. Utilize unmanned aerial vehicle to acquire the aerial photograph in experimental area, rethread Pix4D mapper platform is handled the aerial photograph, obtains 4 low latitude remote sensing orthographic images, including spring zokor rat region image, summer zokor region image, spring woodchuck rat region image, summer woodchuck rat region image. On the basis of fully mastering the ground surface characteristics of different rats, 4 methods of gray threshold segmentation, color texture optimization and rule-based object-oriented and BP neural network are used for extracting the ground surface rat damage information one by one, then the extraction precision is evaluated by adopting a dual evaluation standard of space precision and quantitative precision, the advantages of the methods are summarized, and then the optimal extraction method of the different rat damage information in the images in different seasons is obtained through comparative analysis.
Chinese patent document No. CN112033378A discloses a method for surveying the number of zokors in meadow grassland based on unmanned aerial vehicle aerial photography. The technical method comprises the steps that at the beginning of investigation, a ruler with the length of 2m is placed in a sample, and marked with striking red objects at intervals of 1 m; or placing a striking red object on each side and corner for marking so as to indicate the range of the unmanned aerial vehicle aerial image; determining the flight height of the unmanned aerial vehicle and acquiring an image; and setting related graphic analysis parameters to perform image preprocessing and calculating the number of mouse hills.
The above method has the following disadvantages: the only patent application scope of carrying out the monitoring of grassland rat damage based on unmanned aerial vehicle is limited at present, only can use at local environment, can realize single rat damage discernment under the partly simple scene, but can not satisfy the investigation of many rats under the complicated vegetation environment. The prior art can only extract the information of the rat trace of the pest rat, and the information of the rat trace is not directly quantized into rat density data. In the prior art, the manual work is reduced to a certain extent only in a pest mouse data acquisition mode (using unmanned aerial vehicle for aerial photography) and a mouse trace information extraction method (using a traditional machine learning method), but an integrated and automatic mouse damage investigation flow and an operation platform are not really formed. The monitoring method provided by the prior art has single function and applicable scene, and the investigation demand difference under different scenes is not considered.
Disclosure of Invention
Aiming at the problems in the background technology, a grassland bandicoot density investigation method and an intelligent evaluation system are provided. The invention realizes the intelligence, the automation and the high efficiency of the rat density survey, can more intuitively and accurately reflect the rat situation occurrence level, and has stronger practicability and wider application scene.
The invention provides an intelligent evaluation system for grassland pest mouse density survey, which comprises an unmanned aerial vehicle image acquisition subsystem, a server, a deep learning target monitoring algorithm and an intelligent evaluation software platform.
The unmanned aerial vehicle image acquisition subsystem is used for acquiring mouse track information, such as pictures or videos of mouse holes, hole groups, bald spots and mouse hills, and a data set is formed by making the aerial data and used as a training sample of an intelligent recognition algorithm. The server is used for storing pictures, setting a rat trace standard database and training and operating the algorithm model. The deep learning target monitoring algorithm extracts the rat trace in the image or the video, automatically counts the number and the position information of the rat trace, and outputs a result. The intelligent evaluation software platform integrates and assembles the functions into a whole, images acquired by an unmanned aerial vehicle in actual investigation are led into a computer end, and a pre-written algorithm model and a mouse density measuring and calculating method are intelligently selected through automatic matching with images in a standard mouse trace library in a server, so that mouse trace information is extracted, and a mouse density quantification result is directly output. The invention also provides a grassland rat density investigation method, which comprises the following steps: adopting an unmanned aerial vehicle to automatically acquire a rat trace image, and inputting the preprocessed image into an intelligent evaluation software platform; according to survey scales, flight heights, specific seasons and grassland types, intelligently recommending appropriate deep learning target detection algorithms under different scenes, identifying the rat trace, and automatically extracting the number and accurate position information of the rat trace; outputting the estimation result of the mouse density of the region by software through a mouse density conversion method, and carrying out automatic quantitative analysis on the mouse density in the region; and carrying out position reduction on the map according to the latitude and longitude information of the rat trace, and carrying out visual display and intelligent monitoring on the rat trace in the area.
Preferably, the method comprises the following specific steps:
s1, acquiring aerial images of an area to be surveyed in different seasons, different scales and different grassland types by using an unmanned aerial vehicle: in a typical distribution area of the mouse trail, using an unmanned aerial vehicle to collect pictures or videos of mouse holes, hole groups, bald spots and mouse hills of the mouse in the area to be investigated, and using the pictures or videos as training samples of an intelligent recognition algorithm; the pictures are also used for subsequently making a data set and establishing a rat trace standard database;
s2, carrying out ground manual investigation: setting sampling recipes in different grassland types, acquiring the number of typical mouse tracks such as mouse holes, mouse hills and the like in the recipes by adopting a hole-plugging and hole-stealing method for ground mice and a hole-opening and hole-plugging counting method for underground mice in different seasons, acquiring the number of pest mice by adopting a trapping method, and storing basic data for determining the corresponding relationship between the mouse tracks and the mouse density;
s3, preprocessing aerial images of the unmanned aerial vehicle to obtain an artificial visual interpretation result: preprocessing the unmanned aerial vehicle aerial image acquired in the S1, and labeling a mouse hole, a hole group, bald spots and a mouse hill in a scene to obtain an artificial visual interpretation result; constructing data sets of different rat trace types based on the labeling files, and establishing a multi-scene rat trace standard database;
s4, establishing an optimal deep learning rat trace automatic detection model: based on the acquired unmanned aerial vehicle image, carrying out mouse trace information extraction by using a deep learning target detection algorithm, acquiring the number, spatial position and picture area of mouse holes, hole groups, bald spots and mouse hills, and preparing for carrying out next mouse density statistics;
s5, determining a mouse density estimation method: because the density of the rats is related to the activity trace of the pest rats, the number of the pest rats and the number of actually measured total holes (soil dunes and hole groups) are obtained by carrying out ground manual investigation and counting in a sampling sample prescription, and the population density of the pest rats in the sampling range is obtained through conversion;
s6, assembling a software system: embedding the optimal model determined in the S4 and the mouse density estimation method determined in the S5 into a developed software system;
s7, carrying out grassland pest mouse density survey: a software system calls a rat trace standard database, an algorithm processor in the server matches the rat trace in the image with the standard, an optimal algorithm model and a rat density measuring and calculating method are automatically selected, and a rat trace statistical count, a positioning result and a rat density quantitative value are output; the obtained data and the system are independently learned, and the analyzed and output result data are uploaded to a server through a wireless network for further processing.
Preferably, in the specific operation of S1, the aerial photography month, the survey scale and the sampling range need to be determined, and the unmanned aerial vehicle image acquisition parameters are set.
Preferably, the specific step of S3 includes:
s31, checking pictures;
s32, cutting the picture;
s33, expert marking;
s34, manufacturing a data set;
and S35, establishing a multi-scene rat trace standard database.
Preferably, in S4, if the user inputs video data, the video data needs to be analyzed to obtain video frame data, then continuous frames of the video are spliced into a complete map picture by using a panorama splicing algorithm, then the map picture is cut, then mouse trace detection is performed on the cut small picture, and the detection result is counted to obtain mouse trace statistical data in the video.
Preferably, the specific step of S4 includes:
s41, training typical rat trace recognition models in different seasons;
s42, training a rat trace feature recognition model under a large scale;
s43, evaluating model precision;
and S44, determining the optimal recognition model in different scenes.
Preferably, the estimation content in S5 comprises mouse density estimation based on the detection results of the intelligent algorithm of mouse trail groups, bald spots and mouse hillock holes.
Compared with the prior art, the invention has the following beneficial technical effects:
according to the invention, the unmanned aerial vehicle is used for collecting the mouse trail images, preprocessing the mouse trail images, inputting the mouse trail images into different models preset in the intelligent evaluation system for automatic calculation according to scenes and requirements, and further outputting the mouse trail number and position identification results. And outputting the estimation result of the mouse density of the region by using the number of the mouse traces and a mouse density conversion method, thereby realizing the automatic quantitative analysis of the mouse density in the region. And meanwhile, position reduction is carried out on a map according to latitude and longitude information of the rat trace, so that dynamic analysis of rat damage occurrence in the area, visual display of rat conditions and intelligent monitoring are realized.
The automatic extraction method of the mouse track information used by the invention is a deep learning algorithm, and is more intelligent, automatic and efficient in identifying the target in the image compared with the artificial visual interpretation and the traditional machine learning method. The invention utilizes the programming technology to embed the conversion method from the rat trace information to the rat density into a software system, and can reflect the rat situation more intuitively and quantitatively. The invention integrates the intelligent rat damage detection model and the rat density quantification method into a software system by comprehensively utilizing a database technology, a communication technology, a programming technology and a GIS technology, and really forms an integrated and automatic rat damage investigation flow and an operation platform. According to the method, an intelligent recommendation function is started in software, different optimal model suggestions are provided according to the characteristics of the mice with different scales of the pest mice and different investigation requirements, the practicability is higher, and the application scene is wider.
Detailed Description
Example one
The invention provides an intelligent evaluation system for grassland bandicoot density survey, which comprises an unmanned aerial vehicle image acquisition subsystem, a server, a deep learning target monitoring algorithm and an intelligent evaluation software platform.
The unmanned aerial vehicle image acquisition subsystem is used for acquiring pictures or videos of mouse tracks such as mouse holes, cave groups, bald spots and mouse dunes, and making a data set by using the acquired aerial data to serve as a training sample of an intelligent recognition algorithm. The server is used for storing pictures, setting a rat trace standard database and training and operating the algorithm model. The deep learning target monitoring algorithm extracts the mouse trace in the image or the video, automatically counts the number and the position information of the mouse trace, and outputs a result. The intelligent evaluation software platform integrates the functions, the pictures collected by the unmanned aerial vehicle are led in at the computer end, and the mouse density quantization result is directly output through the written algorithm model and the mouse density measuring and calculating method.
Example two
The invention also provides a grassland rat density investigation method, which comprises the following steps: adopting an unmanned aerial vehicle to automatically acquire a rat trace image, and inputting the preprocessed image into an intelligent evaluation software platform; according to survey scales, flight heights, specific seasons and grassland types, intelligently selecting appropriate deep learning target detection algorithms under different scenes, identifying the rat trace, and automatically extracting the number and accurate position information of the rat trace; outputting the estimation result of the mouse density of the region by software through a mouse density conversion method, and carrying out automatic identification and quantitative analysis on the mouse density in the region; and carrying out position restoration on a map according to the latitude and longitude information of the rat hole or the tunnel group, and carrying out visual display and intelligent monitoring on the rat trace in the area.
EXAMPLE III
The Dongzhu Mucun flag is located in northeast of the union of SilingGuo, 44 degrees 41 'N-46 degrees 10' N, 115 degrees 10 'E-120 degrees 07' E, has high and low geodesic north, is inclined from east to west, and has an altitude of 800-1500 meters. Belongs to continental climate in the northern temperate zone, the cold air is controlled to be large by Mongolia high pressure in winter, and the water and the heat are in the same period in summer. The annual precipitation is about 300 mm, mainly concentrated in 6-8 months. The area is a high-incidence area of the mice damage of the brucella, and the mice damage occurs every year, accompanied by drought and ecological deterioration.
The Brucella (Lasiopodomys brandtii) belongs to small-sized phytophagous rats, and lives in groups with a cave as a unit. It is one of the main pest rats in temperate zone hay grassland in China, and when the occurrence number is high, not only can graze the pasture to cause great reduction of the livestock carrying capacity of the grassland, but also can accelerate the vegetation degradation and desertification. Due to mass propagation of the brucella in summer and clustering in winter, a large amount of rat holes and hole throwing soil is generated in the digging activity, the damage to vegetation in a grassland is serious, and the yield of the pasture is reduced. The vegetation damaged by the high-haired area of the brucella can be gradually restored to the original appearance in 3-5 years under the closed condition, and the development of the animal husbandry is seriously influenced.
In particular, the brucella abortus, as a species of rat living in camphole series, exhibits different rat trace characteristics at different scales. On a small scale, the mouse hole opening is taken as an active characteristic when the mouse is seen in a short distance; on a large scale, particularly in the autumn food storage period, the movable holes are seen from a long distance to present the characteristic of a gathering hole group, and clear runways are connected among the holes.
In this embodiment, a method for investigating the density of grassland pest rats in the second embodiment is further developed by taking a field brookfield as an investigation object and taking an eastern bead mooqin flag as an investigation place, and the method specifically includes the following steps:
s1, acquiring aerial images of a region to be surveyed under multiple scenes of different seasons, different scales, different grassland types and the like by using an unmanned aerial vehicle: in a typical distribution area of the Brucella rat trail, using an unmanned aerial vehicle to collect pictures or videos of the Brucella rat hole and the hole group in different seasons, different heights, different vegetation conditions, different daily periods and illumination conditions of the area to be investigated, and using the pictures or videos as training samples of an intelligent recognition algorithm; the pictures are also used for subsequently making a data set and establishing a rat trace standard database;
s2, carrying out ground manual investigation: in a selected sampling sample prescription of 0.25 hectare (50 m multiplied by 50 m), synchronously carrying out ground total hole quantity investigation and hole group quantity investigation in different seasons (spring, summer and autumn), continuously arranging and clamping for 48 hours by adopting a hole blocking method, counting the number of the mice in the sample prescription, and making basic data storage for determining the corresponding relation between the mouse trace and the mouse density; the purpose of this step is to set up the mouse trace and mouse density corresponding relation in different seasons, after obtaining the conversion coefficient of mouse density in every season, but a small amount of investigation in the normal intelligent investigation monitoring work of the density of mice of the brookfield field in the future, as correcting and using;
s3, preprocessing aerial images of the unmanned aerial vehicle to obtain an artificial visual interpretation result: preprocessing the aerial images of the unmanned aerial vehicle collected in the S1, and labeling rat holes or hole groups in the scene to obtain an artificial visual interpretation result; constructing a rat hole and hole cluster data set based on the labeled file, and establishing a multi-scene rat trace standard database;
s4, establishing an optimal deep learning rat trace automatic detection model: based on the acquired unmanned aerial vehicle image, extracting the rat trace information of the brucella field by using a deep learning target detection algorithm, acquiring the rat trace information under different scales, such as the number, the spatial position, the picture area and the like of rat holes and hole groups, and preparing for counting the rat density in the next step; particularly, for rat hole identification, the distribution conditions of the rat holes and the growth conditions of surrounding vegetation are different in different seasons, and possibly applicable algorithms are different; particularly, for mouse species living in the camp system, the acquisition heights of mouse trace information under different scales are different, and applicable models and conversion methods may also be different. The invention selects 6 deep learning algorithms to compare the rat hole and the tunnel group and selects the high-precision and high-robustness rat hole and tunnel group identification algorithm suitable for different scenes. The key core technology of intelligent, automatic and rapid extraction and detection of the rat trace is solved, and theoretical and technical references are provided for intelligent monitoring of the density of the mice in the Berkovich field of the Leione of the Heilong of the Silvery;
s5, determining a mouse density estimation method: because the density of the mice is related to the activity trace of the pest mice, the number of the brucella and the number of actually measured total holes (hole groups) are obtained by carrying out ground manual investigation and counting in a sampling sample prescription, and the population density of the brucella in the sampling range is obtained through conversion;
s6, assembling a software system: embedding the optimal model determined in the S4 and the mouse density estimation method determined in the S5 into a developed software system;
s7, performing grassland pest mouse density survey: the software system calls a mouse trace standard database, an algorithm processor in the server matches the mouse trace in the image with the standard, and outputs a mouse trace statistical counting and positioning result; the obtained data and the system are independently learned, the identification precision is improved to more than 90%, and the analyzed and output result data are uploaded to a server through a wireless network for further processing.
Further, the specific operation of S1 includes:
s11, determining aerial photography month
The selection of the aerial photographing time should comprehensively consider the breeding characteristics of the bandicoot and the grassland vegetation condition. The mouse densities of 4 months to 5 months (spring), 7 months to 8 months (summer) and 10 months to 11 months (autumn) of each year respectively represent the valley value (before reproduction), the peak value (after reproduction) and the overwintering density of the population number in one year, and the mouse densities of 4 months and 10 months also respectively represent the mouse densities of a first main pest period and a second main pest period in one year. In addition, the meadow mouse trace sampling points cover different grassland types of typical grassland and semi-desert grassland, and are influenced by temperature and rainfall, the growth conditions of grassland vegetation are different between months, and the heights, colors and the like of the grasslands in different months have certain differences, so that the image quality obtained by aerial photography at different times is different from the information reflected by photos.
Furthermore, the brucella does not hibernate, has the habit of storing food in autumn, and starts to store the grain in late 8 months or early 9 months. When the grains are stored, the buchner can clean an old warehouse or dig a new warehouse, the movable hole has the characteristic of an aggregated hole group, and new loose soil, moldy grass chips and the like begin to appear on the hole group. The runway becomes more clear and recognizable at the moment because of frequent grass storage and hole repair.
Therefore, aiming at the purpose of survey and monitoring of the brinell mouse trail in a research area, unmanned aerial vehicle image shooting and ground actual measurement data collection of the brinell mouse portal are carried out in 4 months (spring), 8 months (summer) and 11 months (autumn) in 2020; in addition, in 2021, unmanned aerial vehicle images of the brookfield rat hole groups were taken in 11 months (autumn).
S12, determining survey scale
The artificial measurement shows that the opening of the Bermuda field mouse of the Siniri Guo union is about 4-6cm, and the size of the hole group is from dozens of centimeters to several meters. Different investigation scales are adopted in consideration of the size difference of the target.
The aerial images of the rat hole openings in spring, summer and autumn are collected, the size is small, and the aerial height is low. The relative flight height shot by the unmanned aerial vehicle has positive correlation with the image width and the image resolution. The image resolution also depends on the parameters of the unmanned aerial vehicle-mounted camera. Comprehensively considering the image width, the image resolution, the size of the target object and the flying efficiency, determining a flying height interval of 10m-100m, and carrying out target object shooting experiments with different height gradients in the interval at intervals of 10 m. By contrast, the optimal acquisition height of the hamstring entrance aerial photo based on the Olympus M.Zuiko 45mm/1.8lens and the Zernia braziliana mount Zen X5S pan-tilt camera of the unmanned aerial vehicle of Wu Jiang 2 is determined to be 15 meters.
The aerial photography image of the autumn cave group is collected, the size is relatively large, and the aerial photography height is high. Shooting of the cave group is carried out by using a Xinjiang M300RTK unmanned aerial vehicle carrying a P1 pan-tilt camera, a shooting height interval is planned to be 100-400M, and different shooting height gradients at intervals of 50M are set. After comparison, the optimum height of the brucella warrior cave group aerial photography based on the equipment is determined to be 300 meters.
Furthermore, the density of the mice before winter represented by the hole group determines the base number and the occurrence position of the population number in the second year, and is very important data. Due to the fact that the area of the hole group is large, the aerial photography height is increased, the investigation scale is pushed up, and the investigation efficiency can be greatly improved. The investigation method for automatically identifying and monitoring the rat trails by taking the hole groups as basic investigation units, taking the unmanned aerial vehicle as an investigation platform and taking a deep learning algorithm for the hole-series rats belongs to an innovative method with practical significance.
S13, determining the sampling range
The size of the ground acquisition range mainly depends on the flight height and the unmanned aerial vehicle battery performance. From the previous step, the fly heights at different survey scales have been determined. In the invention, the battery performance of Dajiang Wu 2 can support the longest safe flight time of 25 minutes when Zen Si X5S is mounted, and the maximum safe flight time of Dajiang M300RTK P1 is 45 minutes. In order to guarantee flight safety, in actual flight, once flight control software prompts that the battery power is insufficient, return operation is carried out. Comprehensively considering the above conditions, the unmanned aerial vehicle image acquisition range of the rat hole research area is determined to be 0.25 hectare, and the unmanned aerial vehicle image acquisition range of the hole group is 1 square kilometer.
S14, setting image acquisition parameters of unmanned aerial vehicle
Unmanned aerial vehicle gathers the parameter and includes: aerial photography height, flight area, route planning, flight speed, photographing interval, overlapping rate, camera exposure parameters and the like. And the air line planning is carried out by using a DJI GS Pro ground station professional edition, and the flight parameters such as the flight direction, the height, the photographing interval, the overlapping rate and the like of the unmanned aerial vehicle and the camera exposure parameters are set by using DJI GO 4 flight control software. Generally, the course overlap setting is greater than 75% and the side overlap is greater than 60%. Under the grassland scene, the ground condition similarity is large, and in order to ensure the data shooting effect, the course overlapping rate is set to be 80%, and the side direction overlapping rate is set to be 70%. It is worth explaining that in the course planning, in order to improve the flight efficiency, the flight route is designed by considering the actual shape of the acquisition area as much as possible. And the rest flight parameters and the shooting parameters are set according to the on-site situation through comparative analysis. And after parameter setting is finished, carrying out automatic acquisition on the unmanned aerial vehicle images.
This part is the essential step of unmanned aerial vehicle image acquisition, nevertheless to different collection equipment and application scene, sets for differently. The specific parameter set value is not the key point of the invention, so that the proper parameter setting is adopted.
Further, the specific step of S3 includes:
s31, picture checking: all aerial images are manually checked, and scene pictures with unclear shooting, repeated shooting and poor effect are avoided;
s32, cutting the picture: cutting the whole picture into a plurality of pictures with the spatial resolution of about 500 multiplied by 500 pixels by using a programming language python to prepare for the subsequent data set production;
s33, marking (manual visual interpretation): marking the checked scene pictures, using an image marking tool LabelImg to sequentially frame and select all rat holes or hole groups in the scene by using a rectangular frame, and storing 4 values of coordinates (w, y) at the upper left corner of the rectangular frame and the length and width (w, h) of the rectangular frame into an appointed file, wherein each picture corresponds to one marking file; asking a rat injury expert to assist in identification of confusable ground objects which cannot be judged to be rat traces in the labeling process; the image labeling results are sorted and summarized and used as manual visual interpretation results, and the subsequent intelligent algorithm detection results can be compared;
s34, data set production: selecting a part of pictures containing rat hole and hole group targets in the label file, and respectively making and forming a rat hole data set and a hole group data set; the data sets are divided into a training set, a testing set and a verification set corresponding to each data set in the step S4 of establishing an algorithm model for training optimization and performance evaluation of the algorithm model for rat hole target detection and hole group target detection;
s35, establishing a multi-scene rat trace standard database: according to the air slide acquisition flow and the image preprocessing flow, the grassland mouse damage standard pictures in different seasons, different scales and different grassland types are collected by using an unmanned aerial vehicle, massive data are processed by adopting a Hadoop large data distributed system, small image files are segmented and merged by hipi, and the data are compressed and stored by using an HDFS high-efficiency storage system to form a Brinell field mouse standard mouse damage image library.
Further, in S4, if the user inputs video data, the video data needs to be analyzed to obtain video frame data, then continuous frames of the video are spliced into a complete map picture through a panorama splicing algorithm, the map picture is cut out in a format of 500 × 500, and the deficient people are supplemented with 0 pixel, then the cut small picture is subjected to rat hole or hole group detection, and the detection result is counted to obtain statistical data of the rat hole or hole group in the video.
Further, the specific step of S4 includes:
s41, training a typical rat trace recognition model in different seasons: selecting 6 common deep learning target detection models: fast R-CNN, RFCN, cascade-RCNN, SSD, retinaNet and Yolov4 to determine the best model for realizing the detection of the rat hole of the hamsters. And (3) applying a programming language python to the rat-hole data sets established in the S2, and randomly selecting 80% of pictures in each data set as a training set for training an algorithm model. 10% of the parameters are used as a verification set and used for adjusting and optimizing the model; the remaining 10% was used as a test set to verify model accuracy. Using python, the rat-hole dataset was trained for 6 models, and the test set was used to examine the model results.
S42, training a rat trace feature recognition model under a large scale: after the aerial photography height is increased, the movable holes of the Brucella mouse are characterized by hole groups. The method comprises the steps that cave group detection is carried out on an aerial image, the method still belongs to the field of small target detection in a deep learning algorithm, a target identification scene is possibly similar to the rat hole identification scene, and a model with the best comprehensive performance in rat hole identification can be selected for carrying out cave group identification; if the scene of the hole group identification in the image is different from the rat hole identification, a new algorithm model is established aiming at the hole group identification. In this example, the model with the best overall performance in the above steps can be used for identifying the hole group. On model training, similarly, 80% of the pictures of the hole cluster data set are selected as a training set, 10% are selected as a verification set, and 10% are selected as a test set. Model training was performed using python and the test set was used to verify model results.
S43, model accuracy evaluation: and comparing the result of identifying the number and the position of the rat hole with the result obtained by artificial visual interpretation of the unmanned aerial vehicle image in S3 through an algorithm, and evaluating the accuracy of algorithm identification. The calculation accuracy, robustness and efficiency of the following 3 index evaluation models are selected: AP (Average Precision), F1-score (F1 score) and FPS (frames per second). The results of the evaluation of the average accuracy of the 6 rat hole identification models are shown in table 1.
TABLE 1 average accuracy evaluation results of different rat hole identification models
S44, determining the optimal recognition models in different scenes: the automatic detection of the brucella field rat trail can be divided into two scenes of a rat hole and a hole group according to the investigation scale. For the automatic detection of rat holes, the results are shown in table 1: from the viewpoint of rat hole detection efficiency, the SSD and the YOLOv4 are the models with the fastest speed, and the model operation speed can reach 29.14 frames/second and 10.62 frames/second of FPS respectively. From the precision of rat hole detection, faster R-CNN achieved the highest accuracy and robustness with an average AP value of 0.905 and an F1-score value of 0.861, followed by YOLOv4 with an average AP value of 0.872 and an F1-score value of 0.852. In summary, in practical application, if only the precision is considered, the SSD can be selected for the selection of the rat hole detection model; if only pursuing detection precision, can choose fast-CNN; YOLOv4 may be the best choice if it is desired to compromise efficiency and detection.
For the automatic detection of the cave communities, because the identification of the rat cave and the cave communities in the aerial images under the grassland background belongs to the category of small target detection in the image processing method, the YOLOv4 with better comprehensive performance in the scene is selected for carrying out the automatic detection.
Further, the estimation content in S5 includes mouse density estimation based on the detection result of the rat hole intelligent algorithm and mouse density estimation based on the detection result of the hole group intelligent algorithm.
Considering the seasonal variation of the population of the Brucella and the number of rat holes, determining the hole coefficients of spring, summer and autumn by seasons:
the hole coefficient = the number of brucella/the measured total number of holes.
Based on the number of the rat holes (hole groups) identified by the unmanned aerial vehicle image and the algorithm, determining the density of the Brucella rats according to the following formula by adopting the corresponding relation between the hole opening coefficient and the number of the hole groups and the number of the harmful rats:
rat density = total number of holes x hole coefficient/area,
or
Rat density = number of pest rats in the hole group/area of the area.
The working principle of the invention is as follows: according to the invention, the mouse trail images are acquired by the unmanned aerial vehicle, preprocessed, and then automatically calculated by inputting different preset models in the intelligent evaluation system according to scenes and requirements, so that the number and position recognition results of mouse holes or hole groups are output. And outputting the estimation result of the regional mouse density by using the number of the mouse holes or the hole groups through a mouse density conversion method, thereby realizing the automatic identification and quantitative analysis of the regional mouse density. And meanwhile, position restoration is carried out on the map according to the longitude and latitude information of the rat hole or the tunnel group, so that dynamic analysis of rat damage occurrence in the area, visual display of rat conditions and intelligent monitoring are realized.
The automatic extraction method of the mouse track information used by the invention is a deep learning algorithm, and is more intelligent, automatic and efficient compared with the artificial visual interpretation and the traditional machine learning method for identifying the target in the image. The invention utilizes the programming technology to embed the conversion method from the rat trace information to the rat density into a software system, and can reflect the rat situation more intuitively and quantitatively. The invention comprehensively utilizes database technology, communication technology, programming technology and GIS technology to integrate the intelligent rat damage detection model and the rat density quantification method into a software system, thereby really forming an integrated and automatic rat damage investigation flow and an operation platform. According to the method, an intelligent recommendation function is started in software, different optimal model suggestions are provided according to the characteristics of mice with different scales of harmful mice and different investigation requirements, and the method is high in practicability and wide in application scene.
While the embodiments of the present invention have been described in detail, the present invention is not limited thereto, and various changes can be made without departing from the gist of the present invention within the knowledge of those skilled in the art.
Claims (8)
1. The grassland pest mouse density survey intelligent evaluation system is characterized by comprising an unmanned aerial vehicle image acquisition subsystem, a server, a deep learning target monitoring algorithm and an intelligent evaluation software platform;
the unmanned aerial vehicle image acquisition subsystem is used for acquiring information of the mouse-injurious activity trace, shooting pictures or videos, and making a data set by using aerial data to serve as a training sample of an intelligent recognition algorithm;
the server is used for storing pictures, setting a rat trace standard database and training and operating an algorithm model;
extracting the mouse trace in the image or the video by a deep learning target monitoring algorithm, automatically counting the number and the position information of the mouse trace, and outputting a result;
the intelligent evaluation software platform integrates and assembles the functions into a whole, the pictures acquired by the actual survey unmanned aerial vehicle are led in at the computer end, and the pre-written algorithm model and the mouse density measurement and calculation method are intelligently selected through automatic matching with the images in the standard mouse trace library in the server, so that the mouse trace information is extracted and the mouse density quantification result is directly output.
2. A grassland rat density investigation method comprising the intelligent evaluation system of claim 1, characterized in that the steps comprise: adopting an unmanned aerial vehicle to automatically acquire a rat trace image, and inputting the preprocessed image into an intelligent evaluation software platform; according to survey scales, flight heights, specific seasons and grassland types, intelligently recommending appropriate deep learning target detection algorithms in different scenes, identifying the rat trails, and automatically extracting the number and the accurate position information of the rat trails; outputting the estimation result of the mouse density of the region by software through a mouse density conversion method, and carrying out automatic quantitative analysis on the mouse density in the region; and (4) position restoration is carried out on a map according to the latitude and longitude information of the rat trace, and the rat trace in the area is visually displayed and intelligently monitored.
3. The method for investigating the density of grassland mouse according to claim 2, comprising the following steps:
s1, acquiring aerial images of an area to be investigated under different scenes by using an unmanned aerial vehicle: in an area with frequently distributed rat tracks, acquiring pictures or videos of typical rat tracks of pest rats and rat tracks in a large scale in different seasons, different heights, different vegetation conditions, different time periods of days and illumination conditions of the area to be investigated by using an unmanned aerial vehicle, and taking the pictures or videos as training samples of an intelligent recognition algorithm; the pictures are also used for subsequently making a data set and establishing a rat trace standard database;
s2, carrying out ground manual investigation: setting sampling squares in different grassland types, adopting a hole blocking and stealing method and a hole group counting method for ground mice and adopting a rat hill holing and hole blocking counting method for underground mice to obtain the typical mouse track quantity of rat holes, rat hills and the like in the sampling squares in different seasons, and obtaining the pest mouse quantity by adopting a trapping-out method to store basic data for determining the corresponding relation between the mouse track and the mouse density;
s3, preprocessing aerial images of the unmanned aerial vehicle to obtain an artificial visual interpretation result: preprocessing the aerial images of the unmanned aerial vehicle collected in the S1, and marking the rat trace in the scene to obtain an artificial visual interpretation result; establishing data sets of different mouse trace types based on the annotation files, establishing multi-scene mouse trace standard databases of different seasons, different scales, different grass types and the like, and automatically matching the image input types in the software system to perform intelligent algorithm recommendation;
s4, establishing an optimal deep learning rat trail automatic detection model: based on the acquired unmanned aerial vehicle image, by combining and utilizing the characteristics of a typical rat trace and a large-scale rat trace, extracting rat trace information by adopting a deep learning target detection algorithm, and acquiring the number, the spatial position and the picture area of the rat trace to prepare for next rat density statistics;
s5, determining a mouse density estimation method: because the mouse density is related to the mouse trace, the relation between the number of the pest mice and the number of the mouse trace is obtained by carrying out ground artificial investigation counting in a sampling sample plot, and the pest mouse population density in a sampling range is obtained through conversion;
s6, assembling a software system: embedding the optimal model determined in the S4 and the mouse density estimation method determined in the S5 into a developed software system;
s7, performing grassland pest mouse density survey: the software system calls a mouse trace standard database, an algorithm processor in the server matches the mouse trace in the image with the standard, and outputs a mouse trace statistical counting and positioning result; the obtained data and the system are independently learned, and the analyzed and output result data are uploaded to a server through a wireless network for further processing.
4. The grassland bandicoot density survey method according to claim 3, wherein S1 requires determining aerial photography month, survey scale, sampling range and setting unmanned aerial vehicle image acquisition parameters in specific operation.
5. The method for investigating the density of grassland mouse mice according to claim 3, wherein the step S3 comprises:
s31, checking pictures;
s32, cutting the picture;
s33, expert marking;
s34, making a data set;
and S35, establishing a multi-scene rat trace standard database.
6. The method for investigating the density of mouse and mouse in grassland according to claim 3, wherein in S4, if the user inputs video data, the video data is analyzed to obtain video frame data, then continuous frames of the video are spliced into a complete map picture by a panoramic splicing algorithm, then the map picture is cut, then mouse trace detection is carried out on the cut map, and the detection result is counted to obtain the statistical data of mouse trace in the video.
7. The method for investigating the density of grassland rat mice according to claim 3, wherein the specific step of S4 comprises:
s41, training typical rat trace recognition models in different seasons;
s42, training a rat trace feature recognition model under a large scale;
s43, evaluating the model precision;
and S44, determining the optimal recognition model in different scenes.
8. The method according to claim 3, wherein the estimation content in S5 comprises mouse density estimation based on the detection results of the intelligent algorithm of rat caves, holes, bald spots and rat dunes and mouse density estimation based on the detection results of the intelligent algorithm of holes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211148118.XA CN115527130A (en) | 2022-09-20 | 2022-09-20 | Grassland pest mouse density investigation method and intelligent evaluation system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211148118.XA CN115527130A (en) | 2022-09-20 | 2022-09-20 | Grassland pest mouse density investigation method and intelligent evaluation system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115527130A true CN115527130A (en) | 2022-12-27 |
Family
ID=84698921
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211148118.XA Pending CN115527130A (en) | 2022-09-20 | 2022-09-20 | Grassland pest mouse density investigation method and intelligent evaluation system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115527130A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116778334A (en) * | 2023-06-28 | 2023-09-19 | 中国农业大学 | Quantitative large-scale space grassland mouse entrance density prediction method and system |
| CN117593766A (en) * | 2023-12-28 | 2024-02-23 | 天科院环境科技发展（天津）有限公司 | Investigation method for wild animal population number based on unmanned aerial vehicle shooting image processing |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104320610A (en) * | 2014-10-09 | 2015-01-28 | 韩立亮 | Intelligent recognition system for small mammals (rats) |
| CN108537130A (en) * | 2018-03-15 | 2018-09-14 | 甘肃农业大学 | A kind of Myospalax baileyi and Ochotona curzoniae based on miniature drone technology endanger monitoring method |
| CN111881728A (en) * | 2020-06-16 | 2020-11-03 | 四川师范大学 | Grassland mouse damage monitoring method based on low-altitude remote sensing |
-
2022
- 2022-09-20 CN CN202211148118.XA patent/CN115527130A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104320610A (en) * | 2014-10-09 | 2015-01-28 | 韩立亮 | Intelligent recognition system for small mammals (rats) |
| CN108537130A (en) * | 2018-03-15 | 2018-09-14 | 甘肃农业大学 | A kind of Myospalax baileyi and Ochotona curzoniae based on miniature drone technology endanger monitoring method |
| CN111881728A (en) * | 2020-06-16 | 2020-11-03 | 四川师范大学 | Grassland mouse damage monitoring method based on low-altitude remote sensing |
Non-Patent Citations (1)
| Title |
|---|
| 熊瑞东: "基于低空遥感的若尔盖高寒草地鼠害程度估测研究" * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116778334A (en) * | 2023-06-28 | 2023-09-19 | 中国农业大学 | Quantitative large-scale space grassland mouse entrance density prediction method and system |
| CN117593766A (en) * | 2023-12-28 | 2024-02-23 | 天科院环境科技发展（天津）有限公司 | Investigation method for wild animal population number based on unmanned aerial vehicle shooting image processing |
| CN117593766B (en) * | 2023-12-28 | 2024-04-02 | 天科院环境科技发展（天津）有限公司 | Investigation method for wild animal population number based on unmanned aerial vehicle shooting image processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109840549B (en) | Method and device for identifying plant diseases and insect pests | |
| CN115527130A (en) | Grassland pest mouse density investigation method and intelligent evaluation system | |
| CN109325431B (en) | Method and device for detecting vegetation coverage in feeding path of grassland grazing sheep | |
| US20100040260A1 (en) | Remote sensing and probabilistic sampling based method for determining the carbon dioxide volume of a forest | |
| CN111815014A (en) | Crop yield prediction method and system based on unmanned aerial vehicle low-altitude remote sensing information | |
| Solvin et al. | Use of UAV photogrammetric data in forest genetic trials: measuring tree height, growth, and phenology in Norway spruce (Picea abies L. Karst.) | |
| CN113221765A (en) | Vegetation phenological period extraction method based on digital camera image effective pixels | |
| CN115331124A (en) | Invasive plant biomass estimation method based on unmanned aerial vehicle remote sensing | |
| CN112906734A (en) | Intelligent livestock breeding method and device, computer equipment and storage medium | |
| CN112163639A (en) | Crop lodging classification method based on height distribution characteristic vector | |
| CN114596509A (en) | Machine vision-based rice seedling leaf age period identification method | |
| CN113610009A (en) | Flood disaster unmanned aerial vehicle image information extraction system | |
| CN117079130B (en) | Intelligent information management method and system based on mangrove habitat | |
| CN109523509B (en) | Method and device for detecting heading stage of wheat and electronic equipment | |
| CN111191543A (en) | Rape yield estimation method | |
| CN115314851A (en) | Agricultural information management platform based on big data platform | |
| CN115294448A (en) | Method, device and system for counting seedlings in crop breeding community | |
| Slaymaker | Using georeferenced large-scale aerial videography as a surrogate for ground validation data | |
| Ozcan et al. | Milkweed (Asclepias syriaca) plant detection using mobile cameras | |
| CN110674453B (en) | Method and system for acquiring digital image of cotton leaf cluster index | |
| CN113553897A (en) | Crop identification method based on unmanned aerial vehicle and YOLOv3 model | |
| CN112489008A (en) | Reinforcing steel bar identification system, reinforcing steel bar identification method and reinforcing steel bar counting and acceptance system | |
| CN116403125B (en) | Grassland caterpillar suitable living area dividing method, system and terminal based on unmanned aerial vehicle aerial photography | |
| Young et al. | Crop hail damage: insurance loss assessment using remote sensing | |
| CN114743108B (en) | Grassland mouse condition identification and quantification method and mouse condition recorder |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20221227 |
|
| RJ01 | Rejection of invention patent application after publication |