CN104881865A - Forest disease and pest monitoring and early warning method and system based on unmanned plane image analysis - Google Patents

Abstract

The invention discloses a forest disease and pest monitoring and early warning method based on unmanned plane image analysis. The method includes the steps of: S1: shooting a forest area image through a camera carried on an unmanned plane; S2: performing early-stage preprocessing on an image collected by an image collection unit, so as to improve image data and enhance image features; S3: performing image segmentation on the forest area image through an improved fuzzy C-mean algorithm to determine the position of a forest area in the image; S4: performing segmentation on the forest area image through a mark watershed algorithm based on a mixed template to determine the position of a pest plague area are in the image; and S5: on the basis of determining the position of the pest plague area, and in combination with ground survey data, grading the pest plague degree of the pest plague area. The forest disease and pest monitoring and early warning method uses the unmanned plane to carry the camera to shoot data such as pictures of the specified forest area, realizes functions of positioning of the disease and pest area of the forest area, disease and pest grading and early warning and the like, and can satisfy requirements of timely, comprehensive and efficient monitoring and early warning of a disease and pest situation of the forest area.

Description

Based on forest pest and disease monitoring method for early warning and the system thereof of unmanned plane graphical analysis

Technical field

The present invention relates to forest protection and monitoring technique field, particularly a kind of method for early warning that forest disease and pest region and degree are monitored and system thereof.

Background technology

Forest disease and pest causes huge harm to Forest Health and production of forestry, the direct economic loss that China causes due to forest disease and pest every year and indirect ecology lost revenue heaviness, therefore scientifically monitor and the sound development of control forest disease and pest to China's forestry and ecologic environment is most important effectively.

Traditional Forest Pest Calamity monitoring method mainly takes woodland or the method such as field fixed point monitoring and random searching, and the usual at substantial human and material resources of this method and time, the result of gained is also difficult to grasp the condition of a disaster comprehensively, usually misses the best control time, causes huge loss.Want comprehensively, accurately, promptly monitoring management forest disease and pest must rely on advanced space technology means.From the thirties in 20th century, various countries have carried out the research of Aeronautics and Astronautics remote sensing monitoring plant pest feasibility test in succession, so far achieved many challenging achievements, application remote sensing technology carries out to forest disease and pest and diseases and pests of agronomic crop the hot issue that dynamic monitoring has become Recent study.

But the existing monitoring technology for forest disease and pest and diseases and pests of agronomic crop often also exists human and material resources cost consumption amount greatly, and monitoring periods is long, the accurate not drawback of analytical structure, cannot meet the actual demand in forest zone.Therefore, how to utilize advanced data monitoring and image processing techniques to monitor rapidly and accurately and early warning forest zone disease and pest situation, become those skilled in the art's problem demanding prompt solution.

Summary of the invention

The object of the invention is to solve the defect that existing forest zone pest and disease monitoring dynamics is inadequate, accuracy rate is not high, a kind of forest pest and disease monitoring method for early warning based on unmanned plane graphical analysis and system thereof are provided.

The present invention provide firstly a kind of forest pest and disease monitoring method for early warning based on unmanned plane graphical analysis, comprises the following steps:

S1: by the video camera shooting forest zone image that unmanned plane carries;

S2: carry out pre-service in early stage to the image that described image acquisition units collects, to improve view data, strengthens characteristics of image;

S3: by the Fuzzy C-Means Algorithm improved, Iamge Segmentation is carried out to forest zone image, determine the position in forest zone in image;

S4: by the Based On Method of Labeling Watershed Algorithm based on hybrid template to forest zone Image Segmentation Using, determine the position in plague of insects region in image;

S5: on the basis determining plague of insects regional location, the plague of insects degree of combined ground enquiry data to described plague of insects region carries out classification.

According to forest pest and disease monitoring method for early warning provided by the invention, the mode coming and going scanning by unmanned plane course line in step S1 takes image, the data overlap degree 70% between image.

According to forest pest and disease monitoring method for early warning provided by the invention, the Fuzzy C-Means Algorithm improved described in step S3 comprises the following steps:

A1: clustering initialization

For the limited data set X={x of n vector x i composition ₁, x ₂, x ₃x _n, wherein n is natural number, given initial cluster center set V={v ₀, v ₁v _n-1, primary iteration number of times k=0, clusters number is c (1<c<n), Weighting exponent m (m>0), maximum iteration time T, end condition threshold epsilon;

A2: the subordinated-degree matrix U asking for X ^(k)={ u _ij ^(k), wherein i, j are natural number, u _ijfor being subordinate to angle value;

For arbitrary natural number i and r, work as d _ir ^(k)during >0, wherein d _irfor Euclidean distance is estimated;

and r,

Carrying out stretch processing to being subordinate to angle value, obtaining the membership function stretched:

$y=\left\{\begin{array}{cc}0& x\&le;0.2\\ \frac{3x-1}{2}& 0.2<x<1\\ 1& x>1\end{array}\right.$

Wherein x is for being subordinate to angle value u _ik;

A3: the cluster centre set V asking for renewal ^(k+1)

$\&ForAll;j,{v_j}^{(k+1)}=\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}{\left(y^{\left(k\right)}\right)}^m{x}_i/\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}{\left(y^{\left(k\right)}\right)}^m$

A4: judge cluster termination condition

If || V ^(k)-V ^(k+1)|| < ε or k>T, then stop, otherwise make k=k+1, turns to step a2.

According to forest pest and disease monitoring method for early warning provided by the invention, the described Based On Method of Labeling Watershed Algorithm based on hybrid template is on the basis of Based On Method of Labeling Watershed Algorithm, utilize field character matrix plate operator and improve template operator composition hybrid template, carried out the gradient information of pixel in computed image by the average solving both; Wherein said field character matrix plate operator is:

$M_1=\left[\begin{array}{ccccc}\frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}\\ 0& -p& -p& -p& 0\\ \frac{(p^2-p)}{2}& -p& 8& -p& \frac{(p^2-p)}{2}\\ 0& -p& -p& -p& 0\\ \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{\text{2}}\end{array}\right]$

Described improvement template operator is:

$M_2=\left[\begin{array}{ccccc}-1& 0& -2& 0& 1\\ 0& -3& -4& 3& 0\\ -2& -4& 0& 4& 2\\ 0& -3& 4& 3& 0\\ -1& 0& 2& 0& 1\end{array}\right]$

Calculate pixel Grad:

$G=\frac{1}{2}{M}_{1}A+\frac{1}{2}{M}_{2}A$

In above formula, p is differential order, and A is the pixel matrix of original image.

According to forest pest and disease monitoring method for early warning provided by the invention, step S4 comprises:

Calculate the Grad of each pixel based on hybrid template and sort;

Carry out image to spread unchecked, marker image vegetarian refreshments, obtain preliminary classification result;

Color space conversion, to LUV, upgrades pixel mark, obtains secondary classification image and combined region color average;

Pixel constant for label information is labeled as watershed divide, as Iamge Segmentation border.

In addition, present invention also offers a kind of forest pest and disease monitoring early warning system based on unmanned plane graphical analysis, comprising:

Image acquisition units, for the video camera shooting forest zone image by unmanned plane carries;

Image pre-processing unit, carries out pre-service in early stage for the image collected described image acquisition units, to improve view data, strengthens characteristics of image;

Forest zone positioning unit, carries out Iamge Segmentation by the Fuzzy C-Means Algorithm improved to forest zone image, determines the position in forest zone in image;

Plague of insects positioning unit, by the Based On Method of Labeling Watershed Algorithm based on hybrid template to forest zone Image Segmentation Using, determines the position in plague of insects region in image;

Plague of insects stage unit, on the basis determining plague of insects regional location, the plague of insects degree of combined ground enquiry data to described plague of insects region carries out classification.

According to forest pest and disease monitoring early warning system provided by the invention, the Fuzzy C-Means Algorithm of described improvement comprises the following steps:

A1: clustering initialization

For the limited data set X={x of n vector x i composition ₁, x ₂, x ₃x _n, wherein n is natural number, given initial cluster center set V={v ₀, v ₁v _n-1, primary iteration number of times k=0, clusters number is c (1<c<n), Weighting exponent m (m>0), maximum iteration time T, end condition threshold epsilon;

A2: the subordinated-degree matrix U asking for X ^(k)={ u _ij ^(k), wherein i, j are natural number, u _ijfor being subordinate to angle value

For arbitrary natural number i and r, work as d _ir ^(k)during >0, wherein d _irfor Euclidean distance is estimated;

and r,

Carrying out stretch processing to being subordinate to angle value, obtaining the membership function stretched:

$y=\left\{\begin{array}{cc}0& x\&le;0.2\\ \frac{3x-1}{2}& 0.2<x<1\\ 1& x>1\end{array}\right.$

Wherein x is for being subordinate to angle value u _ik;

A3: the cluster centre set V asking for renewal ^(k+1)

$\&ForAll;j,{v_j}^{(k+1)}=\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}{\left(y^{\left(k\right)}\right)}^m{x}_i/\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}{\left(y^{\left(k\right)}\right)}^m$

A4: judge cluster termination condition

If || V ^(k)-V ^(k+1)|| < ε or k>T, then stop, otherwise make k=k+1, turns to step a2.

According to forest pest and disease monitoring early warning system provided by the invention, the described Based On Method of Labeling Watershed Algorithm based on hybrid template is on the basis of Based On Method of Labeling Watershed Algorithm, utilize field character matrix plate operator and improve template operator composition hybrid template, carried out the gradient information of pixel in computed image by the average solving both; Wherein

Described field character matrix plate operator is:

$M_1=\left[\begin{array}{ccccc}\frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}\\ 0& -p& -p& -p& 0\\ \frac{(p^2-p)}{2}& -p& 8& -p& \frac{(p^2-p)}{2}\\ 0& -p& -p& -p& 0\\ \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{\text{2}}\end{array}\right]$

Described improvement template operator is:

$M_2=\left[\begin{array}{ccccc}-1& 0& -2& 0& 1\\ 0& -3& -4& 3& 0\\ -2& -4& 0& 4& 2\\ 0& -3& 4& 3& 0\\ -1& 0& 2& 0& 1\end{array}\right]$

Calculate pixel Grad:

$G=\frac{1}{2}{M}_{1}A+\frac{1}{2}{M}_{2}A$

In above formula, p is differential order, and A is the pixel matrix of original image.

According to forest pest and disease monitoring early warning system provided by the invention, the described Based On Method of Labeling Watershed Algorithm based on hybrid template comprises the steps:

Calculate the Grad of each pixel based on hybrid template and sort;

Carry out image to spread unchecked, marker image vegetarian refreshments, obtain preliminary classification result;

Color space conversion, to LUV, upgrades pixel mark, obtains secondary classification image and combined region color average;

Pixel constant for label information is labeled as watershed divide, as Iamge Segmentation border.

According to forest pest and disease monitoring early warning system provided by the invention, it is characterized in that, described plague of insects stage unit also comprises:

Plague of insects rating database, records the plague of insects level data of the regional obtained by manual research mode;

Prewarning unit, for sending warning message when plague of insects grade exceedes threshold value.

Compared with prior art, beneficial effect of the present invention is:

1, taken photo by plane by unmanned plane and gather forest zone image, can quick obtaining multi-spatial scale, heterogeneous to ground observation data, the image gathered has the advantage of high-resolution, large scale, small size, high Up-to-date state, and unmanned plane structure is simple, use cost is low, meets an urgent need in burst thing, have very large effect in early warning etc.;

2, the present invention is on the basis of FCM algorithm, propose a kind of Fuzzy C-Means Algorithm (referred to as MFCM) of improvement, namely stretch processing is carried out to the degree of membership of each sample, increase the difference between edge samples, thus obtain image segmentation result more accurately;

3, the present invention utilizes field character matrix plate operator and improves template operator composition hybrid template, is carried out the gradient information of pixel in computed image by the average solving both; While reduction algorithm complex, enhance the neighborhood information of pixel, thus obtain the marginal information in multiple goal region;

4, automatic monitoring result combines with ground artificial enquiry data by the present invention, effectively can avoid the generation of misjudgment phenomenon, greatly improves validity and the degree of accuracy of testing result.

Accompanying drawing explanation

Fig. 1 be unmanned plane of the present invention take in 200m high-altitude, certain forest zone just take the photograph image;

Fig. 2 is the general view that unmanned plane of the present invention is taken in certain forest zone;

Fig. 3 is the course line scanning route schematic diagram of unmanned plane of the present invention in certain forest zone;

Fig. 4 is the process flow diagram that the present invention processes forest zone image;

Fig. 5 is the forest zone classification results that the present invention obtains according to the FCM Algorithms (MFCM) improved;

Fig. 6 is the process flow diagram of the mark Algorithm of Watershed Image Segmentation that the present invention is based on hybrid template;

Fig. 7 be the present invention is based on hybrid template mark Algorithm of Watershed Image Segmentation to the segmentation result of disaster-stricken Seabuckthorn Forest ground image;

Fig. 8 is the hardware composition schematic diagram of forest pest and disease monitoring early warning system of the present invention;

Fig. 9 is the software module block diagram of forest pest and disease monitoring early warning system of the present invention.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, be clearly and completely described the technical scheme in the embodiment of the present invention, obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not paying the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.

The invention discloses a kind of forest pest and disease monitoring method for early warning based on unmanned plane graphical analysis, first be the image utilizing the equipment shootings such as UAV flight's video camera, GPS, GPS registering instrument to specify forest zone, and then by background system, the forest zone image obtained is processed, contrasts, judged, finally determine forest zone plague of insects intensity grade.

GPS on unmanned plane and GPS registering instrument are for locating the forest zone geographical location information of current image shot, and the image of camera acquisition comprises just takes the photograph image (as Fig. 1), general view picture (as Fig. 2) i.e. course line scan image Three models.Wherein, just taking the photograph image can select from the different shooting height of 20m to 200m (shooting interval is 5 meters or 10 meters) as required; General view picture is for characterizing the overall picture in whole region; Course line scan image then can obtain view data more accurately, is 30m course line, certain forest land scanning route schematic diagram as shown in Figure 3, by the scanning of this reciprocation type course line, has the degree of overlapping of 70% between the image photographed, ensure not omit any information.

Fig. 4 processes Aerial Images thus finally determines the process flow diagram of plague of insects grade.Mainly comprise following step:

(1) Image semantic classification

Preliminary pretreatment work is carried out to the image captured by unmanned plane, comprises the geometric correction of imagery, image registration, image mosaic and image enhaucament etc.By carrying out preliminary process to original image, can view data be improved, strengthening characteristics of image, for the steps such as follow-up image characteristics extraction are prepared.

(2) position, forest zone is extracted

Forest zone image zooming-out mainly relies on image Segmentation Technology to realize, and what native system adopted is a kind of Fuzzy C-Means Algorithm (referred to as MFCM) of improvement).Clustering algorithm is a kind of basic analytical approach in image recognition.It is by one group of physics or abstract object, according to certain clustering criteria, it is classified, make sample in class similar as far as possible, between class, the different as far as possible MFCM of sample improves the new algorithm obtained on the basis of FCM algorithm, FuzzyC-Means (FCM algorithm) is a kind of clustering algorithm based on fuzzy set theory, and key step is as follows:

A1: clustering initialization: given initial cluster center V={v ₀, v ₁..., v _n-1, primary iteration number of times k=0, clusters number c, Weighting exponent m (0<m), maximum iteration time T, end condition threshold epsilon.

A2: ask for U ^(k):

r, when time, ${\mathrm{\&mu;}}_{\mathrm{ij}}^{\left(k\right)}=1/\underset{r=1}{\overset{c}{\mathrm{\&Sigma;}}}\left[{\left(\frac{d_{\mathrm{ij}}^{\left(k\right)}}{d_{\mathrm{ir}}^{\left(k\right)}}\right)}^{\frac{2}{m-1}}\right]$

and r,

A3: ask for V ^(k+1):

$\&ForAll;j,v_j^{(k+1)}=\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}{\left({\mathrm{\&mu;}}_{\mathrm{ij}}^{\left(k\right)}\right)}^m{x}_i/\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}{\left({\mathrm{\&mu;}}_{\mathrm{ij}}^{\left(k\right)}\right)}^m$

A4: judge cluster termination condition:

If || V ^(k)-V ^(k+1)|| < ε or k>T, then stop, otherwise make k=k+1, turns to step a2.

The main deficiency of FCM algorithm be due to FCM algorithm to the isolated point of data or noise spot more responsive, so this algorithm has good image segmentation to without the image of making an uproar or signal to noise ratio (S/N ratio) is quite high.Thus, we propose a kind of Fuzzy C-Means Algorithm (referred to as MFCM) of improvement, namely carry out stretch processing to the degree of membership of each sample, increase the difference between edge samples, to obtain image segmentation result more accurately.The membership function of MFCM algorithm is defined as follows:

$y=\left\{\begin{array}{cc}0& x\&le;0.2\\ \frac{3x-1}{2}& 0.2<x<1\\ 1& x>1\end{array}\right.$

Wherein, x represents that FCM's is subordinate to angle value μ ^ik, y represents the membership function μ ' of MFCM ^ik.

Classify its result as shown in Figure 5 according to the forest zone of MFCM algorithm, forest zone well makes a distinction with surrounding physical environment by this algorithm.Because the texture in the image of forest zone is comparatively complicated, its values of fractal dimension is higher, and we are by position, forest zone in maximum values of fractal dimension determination image.

(3) insect pest district differentiates extraction

The present invention adopts the mark Algorithm of Watershed Image Segmentation (being called for short WTM-M-Watershed) based on hybrid template in this step.WTM-M-Watershed algorithm main thought is on the basis of Based On Method of Labeling Watershed Algorithm, utilizes field character matrix plate operator and improves template operator composition hybrid template, carried out the gradient information of pixel in computed image by the average solving both.While reduction algorithm complex, strengthen the neighborhood information of pixel, thus obtain the marginal information in multiple goal region in image.

Wherein character matrix plate operator in field is:

$M_1=\left[\begin{array}{ccccc}\frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}\\ 0& -p& -p& -p& 0\\ \frac{(p^2-p)}{2}& -p& 8& -p& \frac{(p^2-p)}{2}\\ 0& -p& -p& -p& 0\\ \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{\text{2}}\end{array}\right]$

Described improvement template operator is:

$M_2=\left[\begin{array}{ccccc}-1& 0& -2& 0& 1\\ 0& -3& -4& 3& 0\\ -2& -4& 0& 4& 2\\ 0& -3& 4& 3& 0\\ -1& 0& 2& 0& 1\end{array}\right]$

Calculate pixel Grad:

$G=\frac{1}{2}{M}_{1}A+\frac{1}{2}{M}_{2}A$

In above formula, p is differential order, and its empirical value is taken as 0.2 usually; A is the pixel matrix of original image.

Based on the mark Algorithm of Watershed Image Segmentation of hybrid template overall flow figure as shown in Figure 6, its key step is as follows:

S1 ': input picture, is designated as I ₀, duplicating image obtains I _c, gray processing obtains gray level image I _g; Utilize hybrid template computed image I _cin the Grad of each pixel, pixel identical for Grad is divided into similar (totally 256 classes), calculates number of pixels in each Grad, it is sorted.And the mark value of all pixels is initialized as-1, define marking image.

S2 ': overbank process is carried out to image, utilize the correspondence markings value of all pixels in matrix access graph picture, by contrasting the mark value of current pixel point and its four neighborhoods pixel, determine the label information of pixel, pixel identical for label is classified as same class, and obtains the preliminary classification information of image with this.

S3 ': in order to image is classified further, color space conversion is carried out to image, first the color space of image is transferred to LUV space by rgb space, merge the region of part discreet region and mutual colourity difference less (being less than 0.2%) in image according to the LUV colouring information of pixel, the pixel of merging upgrades label.Scan according to the label information of the pixel after renewal subsequently, obtain the Images Classification information after upgrading.Draw the up-to-date color average of the current region after merging simultaneously, finally image is converted to RGB color space again by LUV color space.

S4 ': color scanning present image, finds the pixel of not modified label information (label is still-1), it is marked with particular color, be labeled as watershed divide, output image.

Based on hybrid template mark Algorithm of Watershed Image Segmentation to the segmentation result of disaster-stricken Seabuckthorn Forest ground image as shown in Figure 7, as can be seen from the figure, WTM-M-Watershed algorithm has extracted disaster-stricken sea-buckthorn plant region (highlight regions) exactly.

(4) insect pest district the condition of a disaster classification

The present invention is extracting on the basis of devastated exactly, classification can be carried out to plague of insects degree further combined with ground investigation data, as: damage degree < 2 is healthy, 2≤damage degree < 4 is slightly disaster-stricken, 4≤damage degree < 7 is that moderate is disaster-stricken, and damage degree >=7 are that severe is disaster-stricken; If plague of insects degree exceedes threshold value, send warning information and report to the police, remind staff to carry out site disposal in time.

Except above-mentioned forest pest and disease monitoring method for early warning, the invention allows for a kind of forest pest and disease monitoring early warning system based on unmanned plane image analysis technology.Its hardware composition schematic diagram as shown in Figure 8.

Be arranged on the geographical location information in the forest zone of GPS on unmanned plane and GPS registering instrument location current image shot, and the video camera shooting forest zone image by unmanned plane carries, by image pick-up card, image digitazation be sent to computing machine; Then adopt image analysis technology to carry out forest zone extraction and analysis to gathered image, to judge in the image of forest zone whether ill affected area, if had, to carry out classification judgement to it, and send early warning information to computer terminal.

Fig. 9 is the function structure chart of forest pest and disease monitoring early warning system software section of the present invention.As shown in Figure 9, this monitor and early warning system comprises:

Image acquisition units, for the video camera shooting forest zone image by unmanned plane carries;

Image pre-processing unit, carries out pre-service in early stage for the image collected described image acquisition units, to improve view data, strengthens characteristics of image;

Forest zone positioning unit, carries out Iamge Segmentation by the Fuzzy C-Means Algorithm improved to forest zone image, determines the position in forest zone in image;

Plague of insects positioning unit, by the Based On Method of Labeling Watershed Algorithm based on hybrid template to forest zone Image Segmentation Using, determines the position in plague of insects region in image;

Plague of insects stage unit, on the basis determining plague of insects regional location, the plague of insects degree of combined ground enquiry data to described plague of insects region carries out classification.Wherein, described plague of insects stage unit also comprises: plague of insects rating database, records the plague of insects level data of the regional obtained by manual research mode; Prewarning unit, for sending warning message when plague of insects grade exceedes threshold value.

In sum, the forest pest and disease monitoring method for early warning based on unmanned plane image analysis technology that the present invention proposes and system thereof, UAV flight's video camera can be passed through, the picture of taking photo by plane of specifying forest zone is recorded in the equipment shootings such as GPS registering instrument, the data such as image, transmit it in monitoring and warning platform, by platform, image procossing is carried out to obtained image, comprise image registration, splicing and enhancing etc., extract characteristics of image, image is classified, thus the location realized pest disaster region, forest zone and the function such as disease and pest classification and early warning, meet timely to forest zone disease and pest situation, comprehensively, the demand of efficient monitoring and early warning.

Last it is noted that above embodiment is only in order to illustrate technical scheme of the present invention, be not intended to limit; Although with reference to previous embodiment to invention has been detailed description, those of ordinary skill in the art is to be understood that: it still can be modified to the technical scheme described in previous embodiment, or carries out equivalent replacement to wherein portion of techniques feature; And these amendments or replacement, do not make the essence of appropriate technical solution depart from the spirit and scope of embodiment of the present invention technical scheme.

Claims (10)

1., based on a forest pest and disease monitoring method for early warning for unmanned plane graphical analysis, it is characterized in that, comprise the following steps:

S1: by the video camera shooting forest zone image that unmanned plane carries;

S2: carry out pre-service in early stage to the image that described image acquisition units collects, to improve view data, strengthens characteristics of image;

S3: by the Fuzzy C-Means Algorithm improved, Iamge Segmentation is carried out to forest zone image, determine the position in forest zone in image;

S4: by the Based On Method of Labeling Watershed Algorithm based on hybrid template to forest zone Image Segmentation Using, determine the position in plague of insects region in image;

S5: on the basis determining plague of insects regional location, the plague of insects degree of combined ground enquiry data to described plague of insects region carries out classification.

2. forest pest and disease monitoring method for early warning according to claim 1, is characterized in that, the mode coming and going scanning by unmanned plane course line in step S1 takes image, the data overlap degree 70% between image.

3. forest pest and disease monitoring method for early warning according to claim 1, is characterized in that, the Fuzzy C-Means Algorithm improved described in step S3 comprises the following steps:

A1: clustering initialization

For n vector x _ithe limited data set X={x of composition ₁, x ₂, x ₃x _n, wherein n is natural number, given initial cluster center set V={v ₀, v ₁v _n-1, primary iteration number of times k=0, clusters number is c (1<c<n), Weighting exponent m (m>0), maximum iteration time T, end condition threshold epsilon;

A2: the subordinate function matrix U asking for X ^(k)={ u _ij ^(k), wherein i, j are the locus of image slices vegetarian refreshments, u _ijfor being subordinate to angle value;

For arbitrary natural number i and r, work as d _ir ^(k)during > 0, wherein d _irfor Euclidean distance is estimated;

Carrying out stretch processing to being subordinate to angle value, obtaining the membership function stretched:

$y=\left\{\begin{array}{cc}0& x\&le;0.2\\ \frac{3x-1}{2}& 0.2<x<1\\ 1& x>1\end{array}\right.$

Wherein x is for being subordinate to angle value u _ik;

A3: the cluster centre set V asking for renewal ^(k+1)

$\&ForAll;j,{v_j}^{(k+1)}=\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}{\left(y^{\left(k\right)}\right)}^m{x}_i/\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}{\left(y^{\left(k\right)}\right)}^m$

A4: judge cluster termination condition,

If || V ^(k)-V ^(k+1)|| < ε or k>T, then stop, otherwise make k=k+1, turns to step a2.

4. forest pest and disease monitoring method for early warning according to claim 1, is characterized in that,

The described Based On Method of Labeling Watershed Algorithm based on hybrid template is on the basis of Based On Method of Labeling Watershed Algorithm, utilizes field character matrix plate operator and improves template operator composition hybrid template, carried out the gradient information of pixel in computed image by the average solving both; Wherein

Described field character matrix plate operator is:

$M_1=\left[\begin{array}{ccccc}\frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}\\ 0& -p& -p& -p& 0\\ \frac{(p^2-p)}{2}& -p& 8& -p& \frac{(p^2-p)}{2}\\ 0& -p& -p& -p& 0\\ \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}\end{array}\right]$

Described improvement template operator is:

$M_2=\left[\begin{array}{ccccc}-1& 0& -2& 0& 1\\ 0& -3& -4& 3& 0\\ -2& -4& 0& 4& 2\\ 0& -3& 4& 3& 0\\ -1& 0& 2& 0& 1\end{array}\right]$

Calculate pixel Grad:

$G=\frac{1}{2}{M}_{1}A+\frac{1}{2}{M}_{2}A$

In above formula, p is differential order, and A is the pixel matrix of original image.

5. forest pest and disease monitoring method for early warning according to claim 4, it is characterized in that, step S4 comprises:

Calculate the Grad of each pixel based on hybrid template and sort;

Carry out image to spread unchecked, marker image vegetarian refreshments, obtain preliminary classification result;

Color space conversion, to LUV, upgrades pixel mark, obtains secondary classification image and combined region color average;

Pixel constant for label information is labeled as watershed divide, as Iamge Segmentation border.

6., based on a forest pest and disease monitoring early warning system for unmanned plane graphical analysis, it is characterized in that, comprising:

Image acquisition units, for the video camera shooting forest zone image by unmanned plane carries;

Image pre-processing unit, carries out pre-service in early stage for the image collected described image acquisition units, to improve view data, strengthens characteristics of image;

Forest zone positioning unit, carries out Iamge Segmentation by the Fuzzy C-Means Algorithm improved to forest zone image, determines the position in forest zone in image;

Plague of insects positioning unit, by the Based On Method of Labeling Watershed Algorithm based on hybrid template to forest zone Image Segmentation Using, determines the position in plague of insects region in image;

Plague of insects stage unit, on the basis determining plague of insects regional location, the plague of insects degree of combined ground enquiry data to described plague of insects region carries out classification.

7. forest pest and disease monitoring early warning system according to claim 6, is characterized in that, the Fuzzy C-Means Algorithm of described improvement comprises the following steps:

A1: clustering initialization

For n vector x _ithe limited data set X={x of composition ₁, x ₂, x ₃x _n, wherein n is natural number, given initial cluster center set V={v ₀, v ₁v _n-1, primary iteration number of times k=0, clusters number is c (1<c<n), Weighting exponent m (m>0), maximum iteration time T, end condition threshold epsilon;

A2: the subordinated-degree matrix U asking for X ^(K)={ u _ij ^(K), wherein i, j are natural number, u _ijfor being subordinate to angle value;

For arbitrary natural number i and r, work as d _ir ^(k)during > 0, wherein d _irfor Euclidean distance is estimated;

Carrying out stretch processing to being subordinate to angle value, obtaining the membership function stretched:

$y=\left\{\begin{array}{cc}0& x\&le;0.2\\ \frac{3x-1}{2}& 0.2<x<1\\ 1& x>1\end{array}\right.$

Wherein x is for being subordinate to angle value u _ik;

A3: the cluster centre set V asking for renewal ^(k+1)

$\&ForAll;j,{v_j}^{(k+1)}=\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}{\left(y^{\left(k\right)}\right)}^m{x}_i/\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}{\left(y^{\left(k\right)}\right)}^m$

A4: judge cluster termination condition

If || V ^(k)-V ^(k+1)|| < ε or k>T, then stop, otherwise make k=k+1, turns to step a2.

8. forest pest and disease monitoring early warning system according to claim 6, it is characterized in that, the described Based On Method of Labeling Watershed Algorithm based on hybrid template is on the basis of Based On Method of Labeling Watershed Algorithm, utilize field character matrix plate operator and improve template operator composition hybrid template, carried out the gradient information of pixel in computed image by the average solving both; Wherein

Described field character matrix plate operator is:

$M_1=\left[\begin{array}{ccccc}\frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}\\ 0& -p& -p& -p& 0\\ \frac{(p^2-p)}{2}& -p& 8& -p& \frac{(p^2-p)}{2}\\ 0& -p& -p& -p& 0\\ \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}& 0& \frac{(p^2-p)}{2}\end{array}\right]$

Described improvement template operator is:

$M_2=\left[\begin{array}{ccccc}-1& 0& -2& 0& 1\\ 0& -3& -4& 3& 0\\ -2& -4& 0& 4& 2\\ 0& -3& 4& 3& 0\\ -1& 0& 2& 0& 1\end{array}\right]$

Calculate pixel Grad:

$G=\frac{1}{2}{M}_{1}A+\frac{1}{2}{M}_{2}A$

In above formula, p is differential order, and A is the pixel matrix of original image.

9. forest pest and disease monitoring early warning system according to claim 8, is characterized in that, the described Based On Method of Labeling Watershed Algorithm based on hybrid template comprises the steps:

Calculate the Grad of each pixel based on hybrid template and sort;

Carry out image to spread unchecked, marker image vegetarian refreshments, obtain preliminary classification result;

Color space conversion, to LUV, upgrades pixel mark, obtains secondary classification image and combined region color average;

Pixel constant for label information is labeled as watershed divide, as Iamge Segmentation border.

10. forest pest and disease monitoring early warning system according to claim 1, is characterized in that, described plague of insects stage unit also comprises:

Plague of insects rating database, records the plague of insects level data of the regional obtained by manual research mode;

Prewarning unit, for sending warning message when plague of insects grade exceedes threshold value.