CN112633161B - Pine wood nematode disease dead tree detection positioning method based on high-altitude holder identification - Google Patents

Abstract

The invention discloses a pine wood nematode disease dead tree detection and positioning method based on high-altitude holder identification, which comprises the following steps: determining a detection target area, acquiring the surrounding environment in the detection target area by adopting a high-altitude pan-tilt camera, and generating PTZ scanning information; positioning the photographing position of a camera according to the PTZ scanning information to generate an image to be recognized; the aerial tripod head carries out CNN model identification on the images to be identified in the capture queue Q1, and suspicious images are extracted; the high-altitude holder controls the camera to re-shoot and re-identify the corresponding position according to the suspicious image, and locks the PTZ coordinate position of the dead tree in the suspicious image; and determining longitude and latitude information of a corresponding camera according to the PTZ coordinate position of the dead tree in the suspicious image, and calculating the longitude and latitude of the geographic position of the dead tree. The beneficial effects are that: the method realizes high-altitude shooting of dead trees of pine wood nematode disease, image identification and accurate positioning, and guides direction for patrol.

Description

Pine wood nematode disease dead tree detection positioning method based on high-altitude holder identification

Technical Field

The invention relates to the technical field of forest protection, in particular to a pine wood nematode disease dead tree detection and positioning method based on high-altitude holder identification

Background

Pine wood nematode disease, also called pine wilt disease, is a destructive forest disease caused by pine wood nematodes (Bursaphelenchus xylophilus), belongs to a major foreign invasive species in China, and is listed as an object for quarantining forest plants in and out pairs in China. Since 1982, the disease is rapidly spread and spread, and the area of the disease is 7.7 ten thousand hectares in excess of 14 provinces (city and district) in China, so that a large amount of pine is dead, serious damage is caused to pine forest resources, natural landscapes and ecological environment in China, serious economic and ecological losses are caused, and only 2017 is reported, the pine wood nematode disease in China causes about 195 hundred million yuan of economic losses, wherein the direct economic losses are 35 hundred million yuan, and the indirect economic losses are 160 hundred million yuan. Pine wood nematodes become the first dangerous pests in forestry in China.

The diseased plants are found as soon as possible and cut down, so that the spreading of pine wood nematode diseases can be prevented, and the damage of the diseases to the ecological environment and the economic loss caused by the diseases can be greatly reduced. At present, the method for detecting and treating pine wood nematode disease in China usually comprises the steps of manually searching for diseased pine tree in pine forest, then carrying out treatments such as pesticide application and felling on the pine tree, but most of pine tree dead is first at the top end of the pine tree, and the pine wood nematode disease is widely spread after artificial discovery due to vegetation shielding, and many places are rare, so that the feasibility is low.

Some research institutions use unmanned aerial vehicles to patrol a forest farm, and corresponding algorithms are used for identifying the forest farm. However, the unmanned aerial vehicle has short dead time and limited patrol range, and almost all the whole processes including the processes of taking off, returning, path planning, video reading, uploading analysis and the like still need to be manually processed, so that the cost is quite high.

Thus, the prior art has the disadvantages: the technology which is low in cost, timely, recyclable and less in human participation is lacked, so that the existing pine wood nematode dead tree can be conveniently detected and positioned.

Disclosure of Invention

Aiming at the problems, the invention provides a pine wood nematode disease dead tree detection and positioning method based on high-altitude holder recognition, which utilizes a high-altitude holder and a related camera to remotely shoot forest trees, and combines an image recognition technology and the image characteristics of the pine wood nematode disease to recognize and position the dead trees, thereby being beneficial to forest maintenance personnel to timely control related diseases and preventing the pine wood nematode disease from happening in a large area.

In order to achieve the above purpose, the invention adopts the following specific technical scheme:

a pine wood nematode disease dead tree detection and positioning method based on high-altitude holder identification is characterized by comprising the following steps:

s1: determining a detection target area, acquiring the surrounding environment in the detection target area by adopting a high-altitude pan-tilt camera, and generating PTZ scanning information;

wherein P represents a horizontal angle value; t represents a vertical angle value; z represents the magnification value of the camera;

s2: positioning the photographing position of a camera according to the PTZ scanning information to generate an image to be recognized;

s3: the aerial tripod head carries out CNN model identification on the images to be identified in the capture queue Q1, and suspicious images are extracted;

s4: the high-altitude holder controls the camera to re-shoot and re-identify the corresponding position according to the suspicious image, and locks the PTZ coordinate position of the dead tree in the suspicious image;

s5: and (3) determining longitude and latitude information of a corresponding camera according to the PTZ coordinate position of the dead tree in the suspicious image obtained in the step (S4), and calculating the longitude and latitude of the geographic position of the dead tree.

Through the design, the existing forestry high-altitude fire prevention cloud deck is fully utilized, the dead pine tree caused by the pine wood nematode disease is automatically identified and positioned, any newly-added equipment is not required to be added at the front end, the problem of patrol of the dead pine tree caused by the pine wood nematode disease from the rare to the forest area of the human track can be solved, and the purposes of saving manpower and material resources and protecting ecology by using an algorithm are truly realized.

The further technical scheme is as follows: the specific content of the PTZ scanning information generated in the step S1 is as follows:

setting a start PTZ value and an end PTZ value of the high-altitude PTZ camera to be scanned, namely setting P _start 、T _start 、Z _start 、P _end 、T _end 、Z _end ；

Setting a camera shooting positioning value according to a step-by-step comparison lookup table of a Z value and a PT value between the starting PTZ value and the ending PTZ value, and controlling the camera to shoot according to the shooting positioning value; sequentially placing the photographing positioning values into a photographing positioning array aP 1;

wherein, the step-by-step comparison lookup table of the Z value and the PT value is shown in the first table;

z value range (Z) 0 -Z x )	δP	δT
			1-3	p1	t1
3-8	p2	t3
			8-12	p3	t4
12-15	p4	t5
			15-20	p5	t6
20-30	p6	t7
			Greater than 30	p7	t8

Setting a photographing positioning array aP1 as an empty array, and setting initial values of a horizontal angle value P, a vertical angle value T and a camera magnification value Z; let P ₀ ＝P _start ，T ₀ ＝T _start ，Z ₀ ＝Z _start =1; and P is taken up ₀ ＝P _start ，T ₀ ＝T _start ，Z ₀ ＝Z _start Putting the camera positioning array aP 1;

step-by-step from Z value to PT value according to current magnification Z value of cameraThe lookup table is compared to find out the corresponding horizontal stepping value delta P and vertical stepping value delta T; and obtain P _n ＝P _n-1 +δP and T _n ＝T _n-1 +δT, so that the camera is currently taking a picture; n is an integer between 1 and x;

if P _n <P _start Let P _n-1 ＝P _start Or P _n >P _end Let P _n-1 ＝P _end The method comprises the steps of carrying out a first treatment on the surface of the At this time, P _n＝ P _n-1 +δP’，δP’＝-δP；

If T<T _tart Let T _n-1 ＝T _end Or T>T _end Let T _n ＝T _start ；

And (3) grouping all horizontal angle values P, vertical angle values T and camera magnification times Z with the sequence numbers of 0 to x into a photographing positioning array aP1, and enabling the final photographing positioning array aP1 to be PTZ scanning information.

By adopting the scheme, the p and t values in the stepping comparison lookup tables of the Z value and the PT value are different for different areas, and progressive ptz scanning information is formed according to area self-adaptive setting.

The further calculation scheme is that the specific step of generating the image to be identified in the step S2 is as follows:

s21: the high-altitude holder sequentially reads PTZ values in a photographing positioning array aP1, and positions the camera to a photographing position corresponding to the designated PTZ value by using a camera API of the camera;

s22: after the camera reaches any photographing position in the photographing positioning array aP1, suspending for m seconds, grabbing an image to be identified by using the camera API, and placing the image to be identified into a capture queue Q1;

s23: controlling the camera to reach the next photographing position in the photographing positioning array aP1, and repeating the step S22;

by adopting the scheme, preliminary image acquisition of dead trees is realized, PTZ values are set through different position directions and distances, and for the same area, the PTZ values are changed by a smaller stepping value, so that acquisition of all images of the same area is realized.

The further technical scheme is that the specific steps of CNN recognition of the image to be recognized in step S3 to extract the suspicious image are as follows:

the high-altitude holder sequentially acquires images to be identified from the capture image queue Q1, and puts the images into a CNN model for feature identification to obtain an identification result; the identification result at least comprises frame selection content of the image and dead tree identification data; the dead tree identification data are as follows:

aBoxs＝[{x ₁₁ ,y ₁₁ ,x ₁₂ ,y ₁₂ },{x ₂₁ ,y ₂₁ ,x ₂₂ ,y ₂₂ },...,{x _i1 ,y _i1 ,x _i2 ,y _i2 },...]；

wherein x is _i1 ，y _i1 Representing that the identified tree i corresponds to the upper left corner coordinate of the image to be identified; x is x _i2 ,y _i2 Representing that the identified tree i corresponds to the lower right corner coordinate of the image to be identified;

if dead trees exist in the identification result of the image to be identified, the identification result is put into a dead tree array R1, and the image is a suspicious image; otherwise the image is discarded.

The dead and dead pine tree caused by pine wood nematode disease is generally represented by that the yellow tree trunk and bark of the tree leaves fall off, and in the pitching process, a plurality of disease trees or dead trees are generally accompanied, wood dust and bark appear in groups, and the characteristics can be utilized to automatically identify from far to near when the characteristic identification is carried out. And obtaining suspicious photos.

In still further technical solution, the specific steps of step S4 are: s41: the high-altitude holder sequentially acquires dead tree identification data corresponding to each suspicious image from the dead tree array R1, namely:

aBoxs＝[{x ₁₁ ,y ₁₁ ,x ₁₂ ,y ₁₂ },{x ₂₁ ,y ₂₁ ,x ₂₂ ,y ₂₂ },...,{x _i1 ,y _i1 ,x _i2 ,y _i2 },...]；

s42: the high-altitude cradle head sequentially takes the elements i { x _i1 ,y _i1 ,x _i2 ,y _i2 Use ofThe camera 3D positioning API sends a positioning command to the camera;

s43: the high-altitude cradle head controls the API capture of the camera and stores the captured image of the re-shooting;

s44: the high-altitude holder acquires a PTZ value according to the API of the camera of the current camera; the captured images and the captured images are stored in a captured image queue Q2;

s45: the aerial tripod head sequentially acquires the re-shooting snap-shot images from the re-shooting snap-shot image queue Q2, and places the re-shooting snap-shot images into a CNN model for feature recognition to obtain a re-shooting image recognition result, and places the re-shooting snap-shot images into a re-recognition dead tree array R2; the re-shooting image recognition result comprises re-shooting frame selection content and re-shooting dead tree recognition data;

s46: if dead trees exist in the identification result of the double-shot snap images, locking the PTZ coordinate positions of the dead trees in the suspicious images; otherwise, discarding dead tree identification data corresponding to the suspicious image;

s47: judging whether the dead tree identification data in the suspicious image is identified again, if so, proceeding to step S48; otherwise, identifying the next dead tree identification data in the suspicious image, and returning to the step S41;

s48: judging whether all suspicious images are recognized, if not, locking the next suspicious image, and returning to the step S41; if the identification is completed, outputting all PTZ coordinate positions to enter step S5.

By adopting the steps, the suspicious image is further confirmed, the shooting accuracy is improved by adjusting the shooting angle and the shooting magnification in a re-shooting mode, and the suspicious image is subjected to secondary screening, so that the workload of patrol personnel is reduced.

The further technical scheme is that the specific steps of the step S5 are as follows:

s51: sequentially reading PTZ coordinates in the dead tree array R2, and converting the PTZ coordinates in the dead tree array R2 to dead tree longitude and latitude values by using a geokools tool with the longitude and latitude of a camera as a coordinate origin;

s52: storing the longitude and latitude values of dead trees in a one-to-one correspondence manner in a dead tree re-identification array R2 to obtain dead tree coordinate data R3;

s53: and judging whether all the dead tree arrays R2 are converted or not, if so, outputting dead tree coordinate data R3, otherwise, returning to the step S51.

In the above steps, the target tree is precisely positioned to obtain the corresponding longitude and latitude value, and the patrol personnel can carry out directional patrol according to the data.

The invention has the beneficial effects that: the invention effectively utilizes the high-altitude tripod head to realize high-altitude shooting of dead trees of pine wood nematode disease, identifies images, accurately positions and guides the direction for patrol.

Drawings

FIG. 1 is a flow chart of the method of the present invention;

FIG. 2 is a flow chart of a high altitude pan-tilt camera;

FIG. 3 is a flowchart of the steps of calibrating and locating suspicious pictures by the aerial platform.

Detailed Description

The following describes the embodiments and working principles of the present invention in further detail with reference to the drawings.

The pine wood nematode disease dead tree detection and positioning method based on high-altitude holder recognition is carried out according to the following steps as can be seen from fig. 1:

s1: determining a detection target area, acquiring the surrounding environment in the detection target area by adopting a high-altitude pan-tilt camera, and generating PTZ scanning information;

wherein P represents a horizontal angle value; t represents a vertical angle value; z represents the magnification value of the camera;

the specific content of the PTZ scanning information generated in the step S1 is as follows:

setting a start PTZ value and an end PTZ value of the high-altitude PTZ camera to be scanned, namely setting P _start 、T _start 、Z _start 、P _end 、T _end 、Z _end ；

Setting a camera shooting positioning value according to a step-by-step comparison lookup table of a Z value and a PT value between the starting PTZ value and the ending PTZ value, and controlling the camera to shoot according to the shooting positioning value; sequentially placing the photographing positioning values into a photographing positioning array aP 1;

in this embodiment, the step-by-step comparison lookup table of the Z value and the PT value is shown in table one;

setting a photographing positioning array aP1 as an empty array, and setting initial values of a horizontal angle value P, a vertical angle value T and a camera magnification value Z; let P ₀ ＝P _start ，T ₀ ＝T _start ，Z ₀ ＝Z _start =1; and P is taken up ₀ ＝P _start ，T ₀ ＝T _start ，Z ₀ ＝Z _start Putting the camera positioning array aP 1;

according to the current magnification value Z of the camera, a corresponding horizontal stepping value delta P and a corresponding vertical stepping value delta T are found out from a stepping comparison lookup table of the Z value and the PT value; and obtain P _n ＝P _n-1 +δP and T _n ＝T _n-1 +δT, so that the camera is currently taking a picture; n is an integer between 1 and x;

if P _n <P _start Let P _n-1 ＝P _start Or P _n >P _end Let P _n-1 ＝P _end The method comprises the steps of carrying out a first treatment on the surface of the At this time, P _n＝ P _n-1 +δP’，δP’＝-δP；

If T<T _tart Let T _n-1 ＝T _end Or T>T _end Let T _n ＝T _start ；

And (3) grouping all horizontal angle values P, vertical angle values T and camera magnification times Z with the sequence numbers of 0 to x into a photographing positioning array aP1, and enabling the final photographing positioning array aP1 to be PTZ scanning information.

S2: positioning the photographing position of a camera according to the PTZ scanning information to generate an image to be recognized;

in this embodiment, referring to fig. 2, the specific steps for generating the image to be identified in step S2 are as follows:

s21: the high-altitude holder sequentially reads PTZ values in a photographing positioning array aP1, and positions the camera to a photographing position corresponding to the designated PTZ value by using a camera API of the camera;

s22: after the camera reaches any photographing position in the photographing positioning array aP1, suspending for m seconds, grabbing an image to be identified by using the camera API, and placing the image to be identified into a capture queue Q1;

s23: controlling the camera to reach the next photographing position in the photographing positioning array aP1, and repeating the step S22;

s3: the aerial tripod head carries out CNN model identification on the images to be identified in the capture queue Q1, and suspicious images are extracted;

the specific steps of performing CNN recognition on the image to be recognized in the step S3 to extract the suspicious image are as follows:

the high-altitude holder sequentially acquires images to be identified from the capture image queue Q1, and puts the images into a CNN model for feature identification to obtain an identification result; the identification result at least comprises frame selection content of the image and dead tree identification data; the dead tree identification data are as follows:

aBoxs＝[{x ₁₁ ,y ₁₁ ,x ₁₂ ,y ₁₂ },{x ₂₁ ,y ₂₁ ,x ₂₂ ,y ₂₂ },...,{x _i1 ,y _i1 ,x _i2 ,y _i2 },...]；

wherein x is _i1 ，y _i1 Representing that the identified tree i corresponds to the upper left corner coordinate of the image to be identified; x is x _i2 ,y _i2 Representing that the identified tree i corresponds to the lower right corner coordinate of the image to be identified;

if dead trees exist in the identification result of the image to be identified, the identification result is put into a dead tree array R1, and the image is a suspicious image; otherwise the image is discarded.

S4: the high-altitude holder controls the camera to re-shoot and re-identify the corresponding position according to the suspicious image, and locks the PTZ coordinate position of the dead tree in the suspicious image;

referring to fig. 3, the specific steps of step S4 are as follows:

s41: the high-altitude holder sequentially acquires dead tree identification data corresponding to each suspicious image from the dead tree array R1, namely:

aBoxs＝[{x ₁₁ ,y ₁₁ ,x ₁₂ ,y ₁₂ },{x ₂₁ ,y ₂₁ ,x ₂₂ ,y ₂₂ },...,{x _i1 ,y _i1 ,x _i2 ,y _i2 },...]；

s42: the high-altitude tripod head sequentially takes elements i { x of dead tree identification data _i1 ,y _i1 ,x _i2 ,y _i2 Using the camera's 3D positioning API, sending a positioning command to the camera;

s43: the high-altitude cradle head controls the API capture of the camera and stores the captured image of the re-shooting;

s44: the high-altitude holder acquires a PTZ value according to the API of the camera of the current camera; the captured images and the captured images are stored in a captured image queue Q2;

s45: the aerial tripod head sequentially acquires the re-shooting snap-shot images from the re-shooting snap-shot image queue Q2, and places the re-shooting snap-shot images into a CNN model for feature recognition to obtain a re-shooting image recognition result, and places the re-shooting snap-shot images into a re-recognition dead tree array R2; the re-shooting image recognition result comprises re-shooting frame selection content and re-shooting dead tree recognition data;

s46: if dead trees exist in the identification result of the double-shot snap images, locking the PTZ coordinate positions of the dead trees in the suspicious images; otherwise, discarding dead tree identification data corresponding to the suspicious image;

s47: judging whether the dead tree identification data in the suspicious image is identified again, if so, proceeding to step S48; otherwise, identifying the next dead tree identification data in the suspicious image, and returning to the step S41;

s48: judging whether all suspicious images are recognized, if not, locking the next suspicious image, and returning to the step S41; if yes, outputting all PTZ coordinate positions after identification is finished, and entering step S5.

S5: and (3) determining longitude and latitude information of a corresponding camera according to the PTZ coordinate position of the dead tree in the suspicious image obtained in the step (S4), and calculating the longitude and latitude of the geographic position of the dead tree.

In this embodiment, the specific steps of step S5 are as follows:

s51: sequentially reading PTZ coordinates in the dead tree array R2, and converting the PTZ coordinates in the dead tree array R2 to dead tree longitude and latitude values by using a geokools tool with the longitude and latitude of a camera as a coordinate origin;

s52: storing the longitude and latitude values of dead trees in a one-to-one correspondence manner in a dead tree re-identification array R2 to obtain dead tree coordinate data R3;

s53: and judging whether all the dead tree arrays R2 are converted or not, if so, outputting dead tree coordinate data R3, otherwise, returning to the step S51.

It should be noted that the above description is not intended to limit the invention, but rather the invention is not limited to the above examples, and that variations, modifications, additions or substitutions within the spirit and scope of the invention will be within the scope of the invention.

Claims (4)

1. A pine wood nematode disease dead tree detection and positioning method based on high-altitude holder recognition is characterized by comprising the following steps:

s1: determining a detection target area, acquiring the surrounding environment in the detection target area by adopting a high-altitude pan-tilt camera, and generating PTZ scanning information;

wherein P represents a horizontal angle value; t represents a vertical angle value; z represents the magnification value of the camera;

s2: positioning the photographing position of a camera according to the PTZ scanning information to generate an image to be recognized;

s3: the aerial tripod head carries out CNN model identification on the images to be identified in the capture queue Q1, and suspicious images are extracted;

s4: the high-altitude holder controls the camera to re-shoot and re-identify the corresponding position according to the suspicious image, and locks the PTZ coordinate position of the dead tree in the suspicious image;

s5: determining longitude and latitude information of a corresponding camera according to the PTZ coordinate position of the dead tree in the suspicious image obtained in the step S4, and calculating the longitude and latitude of the geographic position of the dead tree;

the specific content of the PTZ scanning information generated in the step S1 is as follows:

setting a start PTZ value and an end PTZ value of the high-altitude PTZ camera to be scanned, namely setting P _start 、T _start 、Z _start 、P _end 、T _end 、Z _end ；

Setting a camera shooting positioning value according to a step-by-step comparison lookup table of a Z value and a PT value between the starting PTZ value and the ending PTZ value, and controlling the camera to shoot according to the shooting positioning value; sequentially placing the photographing positioning values into a photographing positioning array aP 1;

wherein, the step-by-step comparison lookup table of Z value and PT value is detailed in the following table:

z value range (Z) ₀ -Z _x ）

P

T 1-3 p1 t1 3-8 p2 t3 8-12 p3 t4 12-15 p4 t5 15-20 p5 t6 20-30 p6 t7 Greater than 30 p7 t8

Setting a photographing positioning array aP1 as an empty array, and setting initial values of a horizontal angle value P, a vertical angle value T and a camera magnification value Z; let P ₀ =P _start ，T ₀ = T _start ，Z ₀ =Z _start =1; and P is taken up ₀ =P _start ，T ₀ = T _start ，Z ₀ =Z _start Putting the camera positioning array aP 1;

according to the current magnification value Z of the camera, finding out the corresponding horizontal stepping value from the stepping comparison lookup table of the Z value and the PT value

P, vertical step value->

T is a T; and obtain P _n =P _n-1 +/>

P and T _n =T _n-1 +/>

T, so that the camera takes a picture at the current position; n is an integer between 1 and x;

if P _n <P _start Let P _n-1 =P _start Or P _n >P _end Let P _n-1 =P _end The method comprises the steps of carrying out a first treatment on the surface of the At this time, P _n= P _n-1 +

P’，/>

P’=-/>

P；

If T<T _tart Let T _n-1 =T _end Or T>T _end Let T _n =T _start ；

All horizontal angle values P, vertical angle values T and camera magnification times value Z with sequence numbers of 0 to x are put into a photographing positioning array aP1 in groups, and the final photographing positioning array aP1 is made to be PTZ scanning information;

the specific steps for generating the image to be identified in the step S2 are as follows:

s21: the high-altitude holder sequentially reads PTZ values in a photographing positioning array aP1, and positions the camera to a photographing position corresponding to the designated PTZ value by using a camera API of the camera;

s22: after the camera reaches any photographing position in the photographing positioning array aP1, suspending for m seconds, grabbing an image to be identified by using the camera API, and placing the image to be identified into a capture queue Q1;

s23: and controlling the camera to reach the next photographing position in the photographing positioning array aP1, and repeating the step S22.

2. The pine wood nematode disease dead tree detection and positioning method based on high-altitude holder recognition according to claim 1, wherein the specific steps of performing CNN recognition on the image to be recognized in the step S3 to extract the suspicious image are as follows:

the high-altitude holder sequentially acquires images to be identified from the capture image queue Q1, and puts the images into a CNN model for feature identification to obtain an identification result; the identification result at least comprises frame selection content of the image and dead tree identification data; the dead tree identification data are as follows:

aBoxs=[{x ₁₁ ,y ₁₁ ,x ₁₂ ,y ₁₂ },{x ₂₁ ,y ₂₁ ,x ₂₂ ,y ₂₂ },...,{x _i1 ,y _i1 ,x _i2 ,y _i2 },...]；

wherein x is _i1 ，y _i1 Representing that the identified tree i corresponds to the upper left corner coordinate of the image to be identified; x is x _i2 ,y _i2 Representing that the identified tree i corresponds to the lower right corner coordinate of the image to be identified;

if dead trees exist in the identification result of the image to be identified, the identification result is put into a dead tree array R1, and the image is a suspicious image; otherwise the image is discarded.

3. The pine wood nematode disease dead tree detection and positioning method based on high-altitude holder recognition according to claim 2, wherein the specific steps of step S4 are as follows:

s41: the high-altitude holder sequentially acquires dead tree identification data corresponding to each suspicious image from the dead tree array R1, namely:

aBoxs=[{x ₁₁ ,y ₁₁ ,x ₁₂ ,y ₁₂ },{x ₂₁ ,y ₂₁ ,x ₂₂ ,y ₂₂ },...,{x _i1 ,y _i1 ,x _i2 ,y _i2 },...]；

s42: the high-altitude cradle head sequentially takes the elements i { x _i1 ,y _i1 ,x _i2 ,y _i2 Using the camera's 3D positioning API, sending a positioning command to the camera;

s43: the high-altitude cradle head controls the API capture of the camera and stores the captured image of the re-shooting;

s44: the high-altitude holder acquires a PTZ value according to the API of the camera of the current camera; the captured images and the captured images are stored in a captured image queue Q2;

s45: the aerial tripod head sequentially acquires the re-shooting snap-shot images from the re-shooting snap-shot image queue Q2, and places the re-shooting snap-shot images into a CNN model for feature recognition to obtain a re-shooting image recognition result, and places the re-shooting snap-shot images into a re-recognition dead tree array R2; the re-shooting image recognition result comprises re-shooting frame selection content and re-shooting dead tree recognition data;

s46: if dead trees exist in the identification result of the double-shot snap images, locking the PTZ coordinate positions of the dead trees in the suspicious images; otherwise, discarding dead tree identification data corresponding to the suspicious image;

s47: judging whether the dead tree identification data in the suspicious image is identified again, if so, proceeding to step S48; otherwise, identifying the next dead tree identification data in the suspicious image, and returning to the step S41;

s48: judging whether all suspicious images are recognized, if not, locking the next suspicious image, and returning to the step S41; if the identification is completed, outputting all PTZ coordinate positions to enter step S5.

4. The pine wood nematode disease dead tree detection and positioning method based on high-altitude holder recognition according to claim 3, wherein the specific steps of the step S5 are as follows:

s51: sequentially reading PTZ coordinates in the dead tree array R2, and converting the PTZ coordinates in the dead tree array R2 to dead tree longitude and latitude values by using a geokools tool with the longitude and latitude of a camera as a coordinate origin;

s52: storing the longitude and latitude values of dead trees in a one-to-one correspondence manner in a dead tree re-identification array R2 to obtain dead tree coordinate data R3;

s53: and judging whether all the dead tree arrays R2 are converted or not, if so, outputting dead tree coordinate data R3, otherwise, returning to the step S51.

Images