CN101922914B - Crown information extraction method and system based on high spatial resolution remote sense image - Google Patents
Crown information extraction method and system based on high spatial resolution remote sense image Download PDFInfo
- Publication number
- CN101922914B CN101922914B CN2010102645506A CN201010264550A CN101922914B CN 101922914 B CN101922914 B CN 101922914B CN 2010102645506 A CN2010102645506 A CN 2010102645506A CN 201010264550 A CN201010264550 A CN 201010264550A CN 101922914 B CN101922914 B CN 101922914B
- Authority
- CN
- China
- Prior art keywords
- tree crown
- image
- forest
- crown
- threshold value
- 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.)
- Expired - Fee Related
Links
Images
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a crown information extraction method and system based on the high spatial resolution remote sense image. The method comprises the following steps: obtaining a forest land remote sense image; preprocessing the remote sense image, obtaining the preprocessed remote sense image; adding a forest form map on to the preprocessed remote sense image, using the subcompartment boundary of the forest form map as the clipping region to clip the image information corresponding to each subcompartment in the forest form map, extracting single forest stand image information; utilizing the maximally stable extremal region method to divide the crown and background areas of the single forest stand image information; and extracting the crown image of single tree in the multi-tree crown area, and calculating the number of trees and the crown factor of each tree in the crown distribution map. By using the method and system of the invention, the connected crowns are easy to separate under the condition that the crown image contains fewer pixels, the operation efficiency is high; and the efficiency and degree of automation of the forest resource investigation can be efficiently increased, the forest resource information can be accurately obtained, and manpower and material resources are saved.
Description
Technical field
The present invention relates to the technical field of tree crown information extraction, relate in particular to a kind of crown information extraction method and system based on high spatial resolution remote sense image.
Background technology
The actual standing forest individual plant hat of many at present employings width of cloth method for quantitative measuring obtains the parameters such as profile of individual plant tree crown.This method is the center with the trunk, measures the radius of 8 directions of trunk or 4 directions respectively, obtains the profile of individual plant tree crown then with circular or oval-shaped approximation method; And and then the size of pro form bill strain tree crown; This method workload is big, and expense is high, and the estimation frequency is low; Be difficult to obtain continuous observation data, can't decision support promptly and accurately be provided for forest inventory control.
Based on high-resolution remote sensing image, the crown identification method mainly contains method (local maximum is called for short LM) (Blazquez, 1989 based on local maximum; Dralle et al, 1996), based on method (Contour-based is called for short CB) (Pinz, et al, 1993 of profile; Gougeon, 1995), template matching method (template-matching is called for short TM) (Pollock, 1996; Tarp-Johansen, 2002), 3D modelling (3D-model) (Sheng et al, 2001; Gong et al, 2002), region growing method (region growing is called for short RG) (Brandtberg and Walter, 1998; Culvenor, 2002; Erikson, 2003) and watershed segmentation method (Watershed Segmentation is called for short WS) (Gonzalez and Woods, 2002, Wang et al., 2004, Najman et al., 2005).
The cardinal principle of LM method is that the peak value of hypothesis canopy reflectance is positioned at or is in close proximity to the treetop, thereby finds local maximum through image filtering, can finally detect the position, treetop.Though this method has advantage simply fast, when image is changed by background interference and illumination, be easy to generate a plurality of local maximums.The performance of its crown identification can descend greatly.
The main strategy of CB method is to detect the tree crown border with edge-detection algorithm, also promptly mainly is the separator bar of attempting finding tree crown and its background.Under than the fine scale chi, can confirm all branches of a tree crown more accurately, branch has occupied most Strength Changes; Under thick engineer's scale, adjacent tree crown can interconnect.Like this, cluster is set to become and is changed the place that takes place, so finds in the reality that suitable ratio is very difficult with the adaptation of individual plant tree crown borderline phase strictly.
The TM method comprises that mainly model generates and two processes of template matches.Different sizes and the shape that outward appearance showed of tree have been considered, to the different mathematical model of various parameters structures of different seeds.At first consider geometric configuration and the radiation feature of tree, in case obtained these knowledge, the best match position of just most likely setting through a moving window removal search.
The 3D method is not overripened at present, and the resurfacing that the images match that existing a kind of concrete application is based on model goes to obtain a tree crown is accomplished the tree crown extraction, and the tree crown surface model is considered shape of tree-crown, illumination and sensor model or the like.OO this tree crown method for distilling, formality is loaded down with trivial details, wants height parameter at every turn, and stability is bad as a result in extraction.
The RG method is cut apart according to the spectral characteristic of image, utilizes the tree crown central point that produces in the single ebon hat detection process as seed points, is starting point with the spectral value of seed points; Set a threshold value, whether the spectral value of the interior each point of neighborhood is close with seed points around judging, if the two spectrum intervals is less than threshold value; Then the two belongs to same tree crown zone; It is merged in the zone at seed points place, otherwise, this tree crown zone then do not belonged to.
The WS method is based on mathematical morphology; The spectral value of image is thought of as the landform of fluctuating, earlier raw video is generated gradient image, then gradient image is carried out the processing of edge thinning; The local maximum of final search gradient image, these local maximums are the tree crown border.Watershed algorithm is to faint edge sensitive, but watershed algorithm has the characteristics of over-segmentation, need carry out pre-service or aftertreatment to image.Tree crown extraction effect on aerial image is pretty good, but for satellite image, is not enough height because the spatial resolution of image is compared with the tree crown size, and therefore for satellite image, the tree crown extraction effect is relatively poor.
Above method is all putting forward under concrete image and landscape conditions, based on above-mentioned all be that the closing of crop degree is lower under the more simple relatively standing forest condition, just suitable under the general prerequisite uniformly of tree crown.For the higher standing forest in blocks of canopy density, existing research method is difficult to accurately estimate the individual plant tree crown.Therefore, existing method is inappropriate for the quantitative estimation of carrying out forest zone standing forest hat width of cloth size.
Summary of the invention
The object of the present invention is to provide a kind of crown information extraction method and system based on high spatial resolution remote sense image.
On the one hand, the invention provides a kind of crown information extraction method based on high spatial resolution remote sense image, said method comprises the steps: data acquisition step, obtains the forest land remote sensing image; Pre-treatment step to said remote sensing image pre-service, is obtained pretreated remote sensing image data; Single forest stand image information extraction step, on said pretreated remote sensing image, the stack forest form map is the clipping region with forest form map bottom class border, determines each bottom class's corresponding image information in the forest form map, extracts single forest stand image information; Segmentation procedure is utilized the method in maximum stable extremal district, cuts apart the tree crown and the background area of said single forest stand image information; Single ebon hat extraction step is preced with under the regional situation for many ebons in said tree crown zone, extracts single ebon hat, and the crown outline of sketching out, generates single ebon hat distribution plan; Tree crown factor calculation step is calculated the tree crown factor of said tree crown distribution plan number of trees and every one tree, and the said tree crown factor comprises the hat width of cloth, area, center point coordinate; Show the tree crown factor information according to said center point coordinate distribution, show the space distribution of forest land tree crown in proportion.
Above-mentioned crown information extraction method, preferred said data acquisition step also comprises: the reference mark data of obtaining 1: 10000 digital terrain figure, digitizing forest form map and field operation DGPS; And said pre-treatment step is: with said 1: 10000 digital terrain figure, in conjunction with the reference mark data of said field operation DGPS, the high spatial resolution remote sense image that obtains is carried out orthorectify.
Above-mentioned crown information extraction method, in the preferred said segmentation procedure, said maximum stable extremal district obtains according to following method: attempt segmentation procedure, with all probable values of said single forest stand image as the threshold value of attempting; The pixel of supposing to be lower than said trial threshold value is the tree crown zone, and the pixel that is higher than said trial threshold value is the background area; All said trial threshold values are changed to maximal value gradually from minimum value, successively said single forest stand image is cut apart; Statistic procedure, the trial threshold value of merging appears in record localized target zone, and statistics is attempted under the threshold value area in each localized target zone at this; Segmentation threshold is confirmed step, and when reaching a certain said trial threshold value, the area in said localized target zone can obviously increase, and thinks that then this threshold value is the localized target zone and the segmentation threshold of background; Segmentation procedure according to said segmentation threshold, is cut apart said single forest stand image, obtains a maximum stable extremal region, and this maximum stable extremal region reaches and is said tree crown, and all the other zones are said background area.
Above-mentioned crown information extraction method; In the preferred said single ebon hat extraction step, use successively shrinkage method to extract single ebon and be preced with, comprise the steps: amplification procedure with the condition growth method; Carry out maximum stable extremal distinguish cut after, with the most contiguous method for resampling image is amplified 4 times; Collapse step is used shrinkage method mark list ebon hat center successively; Growth step extracts single ebon hat with the condition growth method; According to the minimum dimension of standing forest list ebon hat, set a threshold value; False tree crown determining step be false tree crown if the area of the said single ebon hat that extracts, is then confirmed this list ebon hat that extracts less than this threshold value, and deletion should vacation tree crown zone.
On the other hand, the present invention also provides a kind of tree crown information extracting system based on high spatial resolution remote sense image, and said system comprises: data acquisition module is used to obtain the forest land remote sensing image; Pre-processing module is used for pretreated remote sensing image data is obtained in said remote sensing image pre-service; The single forest stand image information extraction modules is used at said pretreated remote sensing image, and the stack forest form map is the clipping region with forest form map bottom class border, determines each bottom class's corresponding image information in the forest form map, extracts single forest stand image information; Cut apart module, be used to utilize the method in maximum stable extremal district, cut apart the tree crown and the background area of said single forest stand image information, extract the tree crown zone; Single ebon hat extraction module is used under the situation of said tree crown zone for many ebon hats zone, extracting single ebon hat, and the crown outline of sketching out, generates single ebon hat distribution plan; Tree crown factor calculation module is used to calculate the tree crown factor of said tree crown distribution plan number of trees and every one tree, and the said tree crown factor comprises the hat width of cloth, area, center point coordinate; Show the tree crown factor information according to said center point coordinate distribution, show the space distribution of forest land tree crown in proportion.
Above-mentioned tree crown information extracting system, preferred said data acquisition module is further used for obtaining the reference mark data of 1: 10000 digital terrain figure, digitizing forest form map and field operation DGPS; And said pre-processing module is further used for: with said 1: 10000 digital terrain figure, in conjunction with the reference mark data of said field operation DGPS, the high spatial resolution remote sense image that obtains is carried out orthorectify.
Above-mentioned tree crown information extracting system, also comprises preferred said cutting apart in the module: attempt cutting apart module, be used for all probable values with said single forest stand image as the threshold value of attempting; The pixel of supposing to be lower than said trial threshold value is the tree crown zone, and the pixel that is higher than said trial threshold value is the background area; All said trial threshold values are changed to maximal value gradually from minimum value, successively said single forest stand image is cut apart; Statistical module is used to write down the trial threshold value that merging appears in the localized target zone, and statistics is attempted under threshold value the area in each localized target zone at this; The segmentation threshold determination module is used for when reaching a certain said trial threshold value, and the area in said localized target zone can obviously increase, and confirms that then this threshold value is the localized target zone and the segmentation threshold of background;
Cut apart module, be used for according to said segmentation threshold said single forest stand image being cut apart, obtain a maximum stable extremal region, this maximum stable extremal region reaches and is said tree crown, and all the other zones are said background area.
Above-mentioned tree crown information extracting system, preferred said single ebon hat extraction module comprises: amplification module, be used for carry out maximum stable extremal distinguish cut after, with the most contiguous method for resampling single forest stand image is amplified 4 times; Shrink module, be used for shrinkage method mark list ebon hat center successively; Pop-in upgrades is used for extracting single ebon hat with the condition growth method; False tree crown judge module is used for the minimum dimension according to standing forest list ebon hat, sets a threshold value; If the area of the said single ebon hat that extracts, is then confirmed this list ebon hat that extracts less than this threshold value and is false tree crown, and deletion should vacation tree crown zone.
In terms of existing technologies, the present invention comprises at tree crown and well separates the connection tree crown under the less situation of pixel, make isolated single ebon hat produce very little distortion, and operation efficiency is high.And then, can effectively improve the efficient and the automaticity of forest inventory investigation, obtain forest reserves information timely and accurately, use manpower and material resources sparingly.
Description of drawings
Fig. 1 is the flow chart of steps that the present invention is based on the crown information extraction method embodiment of high spatial resolution remote sense image;
Fig. 2 is for successively contraction and condition growth method extract single ebon hat process flow diagram;
Fig. 3 is the hat of shrinkage method mark list ebon successively process flow diagram;
Fig. 4 is single ebon hat process flow diagram for the condition growth method extracts;
Fig. 5 is the structured flowchart that the present invention is based on the tree crown information extracting system embodiment of high spatial resolution remote sense image.
Embodiment
For make above-mentioned purpose of the present invention, feature and advantage can be more obviously understandable, below in conjunction with accompanying drawing and embodiment the present invention done further detailed explanation.
With reference to Fig. 1, Fig. 1 is the flow chart of steps that the present invention is based on the crown information extraction method embodiment of high spatial resolution remote sense image, comprises the steps:
1) data acquisition step
(A) image data obtains: the high spatial resolution remote sense image in forest land can be aviation digitized video or high-resolution satellite image.
(B) 1: 10000 digital terrain figure
(C) digitizing forest form map
(D) field operation DGPS High Accuracy Control point data.
2) pre-service of remote sensing image
With 1: 10000 digital terrain figure,, the high spatial resolution remote sense image that obtains is carried out orthorectify in conjunction with field operation high-precision GPS reference mark data.
3) single forest stand image information extraction:
Just penetrating on the remote sensing image above-mentioned, the gloomy phasor that superposes is the clipping region with forest form map bottom class border, determines each bottom class's corresponding image information in the forest form map.
4) utilize maximum stable extremal district (MSER) method to cut apart tree crown and background area
Cut apart the tree crown zone with the MSER method, a MSER (maximum stable extremal district) selects suitable threshold and obtains connected component topography, and the stationarity of these connected components is detected the final plateau region of acquisition.Tree crown can be thought some zones that spectral value is more approaching.
Arthmetic statement: with all probable values of piece image, as threshold value.Suppose that the pixel that is lower than threshold value is the tree crown zone, the pixel that is higher than threshold value is the background area.Threshold value is changed to maximal value gradually from minimum value, and on some threshold values, merging can appear in the target area.Add up the area in each localized target zone, when reaching a certain threshold value, the area in localized target zone can obviously increase.Think that then this threshold value is the segmentation threshold of localized target zone and background.Tree crown zone after cutting apart is a maximum stable extremal region.
The MSER arithmetic result is similar to the algorithm that a kind of suboptimization threshold value is selected.General Threshold Segmentation Algorithm adopts same threshold value that image is cut apart in the entire image scope, when image range bigger, when textural characteristics enriches, the segmentation effect that single threshold value can not play.The MSER algorithm adopts independently threshold value to each localized target zone.Regional area at image is grown, until obtaining extreme value stabilized zone (tree crown zone just).Algorithm essence is exactly in the zones of different scope, generates adaptive segmentation threshold, has obtained segmentation effect preferably.The cutting apart and compare based on the local maximum method of this object-oriented zone, the uncertainty of having avoided Local Extremum to choose has improved the stability that target is chosen.To the identification of tree crown target area, be not subject to noise, when illumination changes, also have more stable effect.
5) single ebon hat extracts automatically
Arboreal growth in the forest land interconnects between tree crown to certain time limit usually, and promptly canopy density increase.In forest resourceies investigation, what a lot of occasions were concerned about is the characteristic of single trees, like its tree crown size shape or characteristic parameter etc.Therefore, need find out the connecting line between tree crown, separate the tree crown that links to each other, reach the purpose that single wood extracts, this also is statistics standing forest trees number and the basis of calculating the tree crown factor.The method for distilling of single ebon hat mainly is from many ebon hats zone, to isolate single ebon hat zone.Many ebon hats here are meant the tree crown zone that two or more trees are connected with each other.
It is relatively low to the present invention is directed to image resolution, and the less situation of pixel that single ebon hat that will extract in the image comprises provides a kind of method that combines with the condition growth method of successively shrinking.Successively shrinkage method more after a little while, can make full use of less shape information at pixel; And, can well separate the connection tree crown through shrinking and growth.This method comprises under the less situation of pixel at tree crown, make shape of tree-crown in contraction and growth course, be difficult for distortion, and operation efficiency is high.
Below, explain and successively shrink and condition growth method single ebon hat of extraction from many ebon hats.
Carrying out extracting the tree crown zone after MSER cuts apart, the tree crown zone is adopted successively to shrink with the condition growth method can extract single ebon hat.This algorithm is regarded interconnective tree crown as the some zones that are interconnected, and earlier connected region is made marks, and for morphologic algorithm provides marking image, through shrinking and the growth computing, detects the single tree crown that image comprises, and extracts profile information.
With reference to Fig. 2, this method is made detailed description.
A) image resampling amplifies
Carrying out with the most contiguous method for resampling image being amplified 4 times after MSER cuts apart.The shape that so both can keep original image can guarantee that again enough big zone carries out morphologic operation.According to experiment, when amplifying more than 4 times, 4 times of no significant differences of result and amplification.
B) successively shrinkage method mark list ebon is preced with the center
When successively shrinkage method is handled, the connected region that exists in the image is dwindled and can not make its disappearance.Result is that each regions contract becomes a point or do not have interior zonule of putting, and this point or zonule are called nucleus.The promptly corresponding tree crown of each nucleus.Like this, through just can know the position and sum of single ebon hat in the image to these nucleus countings.
When carrying out shrink process, do not make unborn connected region disappearance in the image, making regions contract is the nucleus that isolates.The key that realizes is not make the zone to disappear and when finishes contraction process.
The input picture of shrinkage operation and output image should separate.Operation steps is following:
Step 2 subtracts 1 with the pixel value of all non-zero pixels in the distance map.
Step 3 detects all connected regions in the input picture.According to the connected region in the input picture, whether also there is non-zero pixels in output image relevant position inspection, if there has not been non-zero pixels in certain zone in output image, then on output image, recovering should the zone, makes it be unlikely to disappear.
Step 4 all need be recovered as if All Ranges in the output image, then termination procedure.
Fig. 3 shows above-mentioned 4 steps.
C) the condition growth method extracts single ebon hat
Condition growth method: make edges of regions outwards expand a pixel and be called a secondary growth.The condition growth is not unconfined growth, and it can only be grown in the regional extent of original image, and the nucleus that simultaneously mark is come out can not link together in growth course.These two steps of process gather, and the zone that linked together originally is just by separated from each other.
With reference to Fig. 4, Fig. 4 is single ebon hat process flow diagram for the condition growth method extracts, and operation steps is following:
Step 2 is found out the pixel that this secondary growth increases newly through the distance of input, output image after each growth process.
Step 3 detects newly-increased pixel, if it then removes this pixel for interior point in the growth output image.
Step 4 is checked the pixel that each is newly-increased, if this pixel is then removed in the background area of this pixel in original image.
Step 5 detects the linking number that increases pixel newly in the growth input picture, if its linking number is also removed this pixel greater than 1.
Step 6 is counted the newly-increased pixel of not deletion in testing process.This number is 0 and stops long-living process.
D) false tree crown zone removal
According to the minimum dimension of standing forest list ebon hat, set a threshold value; If the area in said tree crown zone less than this threshold value, confirms that then this tree crown zone is false tree crown, and deletion should vacation tree crown zone.
6) tree crown factor calculation
In order to realize the rapid extraction of the tree crown factor, on the basis of tree crown distribution plan, developed the automatic extractive technique of the tree crown factor.
(1) design and function: algorithm design adopts the OOP technology, and grid map is handled.To the wide extraction of single ebon crown gear, generate the tree crown distribution plan according to above-mentioned.This algorithm is exactly the tree crown factor of every one tree during calculating is published picture: the hat width of cloth, and area, center point coordinate, and, be shown as the space distribution of the tree that lives in proportion according to the distributed demonstration tree crown of center point coordinate factor information.
(2) algorithm is realized: image is the collection of pixels with the storage of behavior unit.Regard the tree crown distribution plan as a forest object Forest, it is the set of tree object: Forest={Tree_1, Tree_2 ... Tree_n} for the ease of the generation and the management of tree object, has introduced the line segment object.Tree is made up of some line segment objects liking: Tree{Segment1, Segment2 ... Segmentn}, line segment object (Segment) attribute has: in the image row number, start position, final position.It is a line segment aggregate L{Segment_1 that each trade of image file is done, Segment_2 ..., Segment_m}, entire image file regard a row set F{Line_1 as, Line_2 ..., Line_n}.Therefore, the set Tree{Segment_1 that every one tree is made up of neighbouring n1 line segment segment, Segment_2 ..., Segment_n1}.Wherein, it is wide that the longest line segment is the things of trees, and line segment number n1 is that the north and south of trees is wide, and the length overall of all line segments is the area of tree, and the length and width of tree crown are extrapolated center point coordinate.Can be depicted as the spatial distribution map of the tree that lives in view of the above.
On the other hand, the present invention also provides a kind of tree crown information extracting system based on high spatial resolution remote sense image, with reference to Fig. 5, comprising:
Data acquisition module 50 is used to obtain the forest land remote sensing image; Pre-processing module 51 is used for pretreated remote sensing image data is obtained in said remote sensing image pre-service; Single forest stand image information extraction modules 52 is used at said pretreated remote sensing image, and the gloomy phasor that superposes is the clipping region with forest form map bottom class border, determines each bottom class's corresponding image information in the forest form map, extracts single forest stand image information; Cut apart module 53, be used to utilize the method in maximum stable extremal district, cut apart the tree crown and the background area of said single forest stand image information, and extract the tree crown zone; Single ebon hat extraction module 54 is used under the situation of said tree crown zone for many ebon hats zone, extracting single ebon hat, and the crown outline of sketching out, generates single ebon hat distribution plan; Tree crown factor calculation module 55 is used to calculate the tree crown factor of said tree crown distribution plan number of trees and every one tree, and the said tree crown factor comprises the hat width of cloth, area, center point coordinate; Show the tree crown factor information according to said center point coordinate distribution, show the space distribution of forest land tree crown in proportion.
More than to having done detailed explanation based on the crown information extraction method of high spatial resolution remote sense image; Tree crown information extracting system based on high spatial resolution remote sense image is identical with said method embodiment principle; Repeat no more at this, relevant part is mutually with reference to getting final product.
To sum up, have following characteristics in the present invention:
(1) mathematical morphology MSER algorithm is used for tree crown image front and back background segment,, has improved the stability of target area information extraction owing to be based on cutting apart of zone; Cutting techniques based on many threshold values; For the textural characteristics more complicated, the image that noise is bigger also has good effect.
(2) for the relatively low image of resolution, adopt the most contiguous image after amplifying 4 times, carry out single ebon hat mark through shrinkage method successively, carry out tree crown through the condition growth then and cut apart, the tree crown that links to each other is had good segmentation effect.
(3) automaticity is high, can count the area of tree crown automatically, girth, and the hat width of cloth can be widely used in forestry investigation, applications such as forest resource monitoring, forestry remote sensing.
More than a kind of crown information extraction method and system based on high spatial resolution remote sense image provided by the present invention described in detail; Used specific embodiment among this paper principle of the present invention and embodiment are set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, part all can change on embodiment and range of application.In sum, this description should not be construed as limitation of the present invention.
Claims (4)
1. the crown information extraction method based on high spatial resolution remote sense image is characterized in that, said method comprises the steps:
Data acquisition step is obtained the forest land remote sensing image;
Pre-treatment step to said remote sensing image pre-service, is obtained pretreated remote sensing image data;
Single forest stand image information extraction step, on said pretreated remote sensing image, the stack forest form map is the clipping region with forest form map bottom class border, determines each bottom class's corresponding image information in the forest form map, extracts single forest stand image information;
Segmentation procedure is utilized the method in maximum stable extremal district, cuts apart the tree crown and the background area of said single forest stand image information;
Single ebon hat extraction step is under the situation in many ebon hats zone in said tree crown zone, extracts single ebon hat, and the crown outline of sketching out, generates single ebon hat distribution plan;
Tree crown factor calculation step is calculated the tree crown factor of said tree crown distribution plan number of trees and every one tree, and the said tree crown factor comprises the hat width of cloth, area, center point coordinate; Show the tree crown factor information according to said center point coordinate distribution, show the space distribution of forest land tree crown in proportion; Wherein
Said data acquisition step also comprises: the reference mark data of obtaining 1: 10000 digital terrain figure, digitizing forest form map and field operation DGPS;
And said pre-treatment step is:
With said 1: 10000 digital terrain figure,, the high spatial resolution remote sense image that obtains is carried out orthorectify in conjunction with the reference mark data of said field operation DGPS; And
In the said segmentation procedure, said tree crown and background area obtain according to following method:
Cut apart pre-treatment step, with all probable values of said single forest stand image as the threshold value of attempting; The pixel of supposing to be lower than said trial threshold value is the tree crown zone, and the pixel that is higher than said trial threshold value is the background area; All said trial threshold values are changed to maximal value gradually from minimum value, successively said single forest stand image is cut apart;
Statistic procedure, the trial threshold value of merging appears in record localized target zone, and statistics is attempted under the threshold value area in each localized target zone at this;
Segmentation threshold is confirmed step, and when reaching a certain said trial threshold value, the area in said localized target zone can obviously increase, and thinks that then this threshold value is the localized target zone and the segmentation threshold of background;
Tree crown and background area obtaining step according to said segmentation threshold, are cut apart said single forest stand image, obtain a maximum stable extremal region, and this maximum stable extremal region is said tree crown, and all the other zones are said background area.
2. crown information extraction method according to claim 1 is characterized in that, in the said single ebon hat extraction step, uses successively shrinkage method to extract single ebon with the condition growth method and is preced with, and comprises the steps:
Amplification procedure, carry out maximum stable extremal distinguish cut after, with the most contiguous method for resampling image is amplified 4 times;
Collapse step is used shrinkage method mark list ebon hat center successively;
Growth step extracts single ebon hat with the condition growth method;
False tree crown determining step according to the minimum dimension of standing forest list ebon hat, is set a threshold value; If the area of the said single ebon hat that extracts, is then confirmed this list ebon hat that extracts less than this threshold value and is false tree crown, and deletion should vacation tree crown zone.
3. tree crown information extracting system based on high spatial resolution remote sense image is characterized in that said system comprises:
Data acquisition module is used to obtain the forest land remote sensing image;
Pre-processing module is used for pretreated remote sensing image data is obtained in said remote sensing image pre-service;
The single forest stand image information extraction modules is used at said pretreated remote sensing image, and the stack forest form map is the clipping region with forest form map bottom class border, determines each bottom class's corresponding image information in the forest form map, extracts single forest stand image information;
Cut apart module, be used to utilize the method in maximum stable extremal district, cut apart the tree crown and the background area of said single forest stand image information;
Single ebon hat extraction module is used under the situation of said tree crown zone for many ebon hats zone, extracting single ebon hat, and the crown outline of sketching out, generates single ebon hat distribution plan;
Tree crown factor calculation module is used to calculate the tree crown factor of said tree crown distribution plan number of trees and every one tree, and the said tree crown factor comprises the hat width of cloth, area, center point coordinate; Show the tree crown factor information according to said center point coordinate distribution, show the space distribution of forest land tree crown in proportion, wherein
Said data acquisition module is further used for obtaining the reference mark data of 1: 10000 digital terrain figure, digitizing forest form map and field operation DGPS; And said pre-processing module is further used for:
With said 1: 10000 digital terrain figure,, the high spatial resolution remote sense image that obtains is carried out orthorectify in conjunction with the reference mark data of said field operation DGPS;
Said cutting apart in the module, also comprise:
Cut apart pre-processing module, be used for all probable values of said single forest stand image as the threshold value of attempting; The pixel of supposing to be lower than said trial threshold value is the tree crown zone, and the pixel that is higher than said trial threshold value is the background area; All said trial threshold values are changed to maximal value gradually from minimum value, successively said single forest stand image is cut apart;
Statistical module is used to write down the trial threshold value that merging appears in the localized target zone, and statistics is attempted under threshold value the area in each localized target zone at this;
The segmentation threshold determination module is used for when reaching a certain said trial threshold value, and the area in said localized target zone can obviously increase, and confirms that then this threshold value is the localized target zone and the segmentation threshold of background;
Tree crown and background area acquisition module are used for according to said segmentation threshold said single forest stand image being cut apart, and obtain a maximum stable extremal region, and this maximum stable extremal region is said tree crown, and all the other zones are said background area.
4. tree crown information extracting system according to claim 3 is characterized in that, said single ebon hat extraction module comprises:
Amplification module, be used for carry out maximum stable extremal distinguish cut after, with the most contiguous method for resampling single forest stand image is amplified 4 times;
Shrink module, be used for shrinkage method mark list ebon hat center successively;
Pop-in upgrades is used for extracting single ebon hat with the condition growth method;
False tree crown judge module is used for the minimum dimension according to standing forest list ebon hat, sets a threshold value; If the area of the said single ebon hat that extracts, is then confirmed this list ebon hat that extracts less than this threshold value and is false tree crown, and deletion should vacation tree crown zone.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2010102645506A CN101922914B (en) | 2010-08-27 | 2010-08-27 | Crown information extraction method and system based on high spatial resolution remote sense image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2010102645506A CN101922914B (en) | 2010-08-27 | 2010-08-27 | Crown information extraction method and system based on high spatial resolution remote sense image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101922914A CN101922914A (en) | 2010-12-22 |
CN101922914B true CN101922914B (en) | 2012-02-22 |
Family
ID=43337988
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2010102645506A Expired - Fee Related CN101922914B (en) | 2010-08-27 | 2010-08-27 | Crown information extraction method and system based on high spatial resolution remote sense image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101922914B (en) |
Families Citing this family (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102096818B (en) * | 2011-01-31 | 2012-12-12 | 中国林业科学研究院资源信息研究所 | Object-oriented automatic extracting method and system for outline and parameter of remote sensing image crown |
CN102393180B (en) * | 2011-10-19 | 2013-09-04 | 中国林业科学研究院资源信息研究所 | Method for automatically extracting forest stand upper layer tree parameters from LiDAR point cloud data |
CN102419818B (en) * | 2011-10-28 | 2014-12-10 | 中国林业科学研究院资源信息研究所 | LiDAR (Light Detecting and Ranging) data single-tree extraction method with combination of morphological canopy control and watershed |
CN103528570A (en) * | 2013-10-23 | 2014-01-22 | 环境保护部卫星环境应用中心 | Spatial distribution acquiring method and system for spartina alterniflora in northern sea area of Guangxi |
CN104361592A (en) * | 2014-11-14 | 2015-02-18 | 中国林业科学研究院资源信息研究所 | Method and device for estimating forest above-ground biomass |
CN105241423B (en) * | 2015-09-18 | 2017-03-08 | 北京林业大学 | A kind of evaluation method based on unmanned plane photogram to high canopy density Stand Volume |
CN105403876B (en) * | 2015-12-24 | 2018-01-30 | 中国林业科学研究院资源信息研究所 | The measuring method and device of forest canopy density |
CN105868717B (en) * | 2016-03-29 | 2019-05-28 | 南京信息工程大学 | A kind of high-resolution remote sensing image list wood crown information extraction method based on classification |
CN109325451B (en) * | 2017-12-31 | 2022-05-06 | 苏州中科天启遥感科技有限公司 | High-spatial-resolution cultivated land block full-automatic extraction method based on deep learning |
CN109166148B (en) * | 2018-08-08 | 2021-08-27 | 长春工程学院 | Method for extracting crown diameter in remote sensing image of regional gray and directional component operator |
CN109765932A (en) * | 2019-01-31 | 2019-05-17 | 交通运输部天津水运工程科学研究所 | A kind of desert shrubbery cover degree unmanned plane investigation method |
CN111310614B (en) * | 2020-01-22 | 2023-07-25 | 航天宏图信息技术股份有限公司 | Remote sensing image extraction method and device |
CN111598915B (en) * | 2020-05-19 | 2023-06-30 | 北京数字绿土科技股份有限公司 | Point cloud single wood segmentation method, device, equipment and computer readable medium |
CN112069947A (en) * | 2020-08-25 | 2020-12-11 | 航天信德智图(北京)科技有限公司 | Fruit forest grain number statistical system based on density analysis |
CN112580504B (en) * | 2020-12-17 | 2023-01-17 | 中国科学院空天信息创新研究院 | Tree species classification counting method and device based on high-resolution satellite remote sensing image |
CN112861336B (en) * | 2021-02-01 | 2023-07-25 | 中国林业科学研究院资源信息研究所 | Virtual simulation method for determining forest breaking area by angle gauge |
CN117274844B (en) * | 2023-11-16 | 2024-02-06 | 山东科技大学 | Rapid extraction method for field peanut seedling emergence condition by using unmanned aerial vehicle remote sensing image |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1996044A (en) * | 2006-12-26 | 2007-07-11 | 中国林业科学研究院资源信息研究所 | Canopy spatial statistics quantitative estimation method based on remote sensing image with fine spatial resolution |
CN101403795A (en) * | 2008-11-18 | 2009-04-08 | 北京交通大学 | Remote sensing survey method and system for estimating tree coverage percentage of city |
GB2458278A (en) * | 2008-03-11 | 2009-09-16 | Geoffrey Cross | A method of recognising signs in images taken from video data |
CN101672915A (en) * | 2009-09-23 | 2010-03-17 | 中国林业科学研究院资源信息研究所 | High spatial resolution remote sensing image crown outline delineation system and method |
-
2010
- 2010-08-27 CN CN2010102645506A patent/CN101922914B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1996044A (en) * | 2006-12-26 | 2007-07-11 | 中国林业科学研究院资源信息研究所 | Canopy spatial statistics quantitative estimation method based on remote sensing image with fine spatial resolution |
GB2458278A (en) * | 2008-03-11 | 2009-09-16 | Geoffrey Cross | A method of recognising signs in images taken from video data |
CN101403795A (en) * | 2008-11-18 | 2009-04-08 | 北京交通大学 | Remote sensing survey method and system for estimating tree coverage percentage of city |
CN101672915A (en) * | 2009-09-23 | 2010-03-17 | 中国林业科学研究院资源信息研究所 | High spatial resolution remote sensing image crown outline delineation system and method |
Non-Patent Citations (3)
Title |
---|
覃先林,李增元,易浩若.高空间分辨率卫星遥感影像树冠信息提取方法研究.《遥感技术与应用》.2005,第20卷(第2期),228-232. * |
黄建文,陈永富,鞠洪波.基于面向对象技术的退耕还林树冠遥感信息提取研究.《林业科学》.2006,第42卷68-71. * |
黄建文,鞠洪波,赵峰,陈巧,马红.利用遥感进行退耕还林成活率及长势监测方法的研究.《遥感学报》.2007,第11卷(第6期),899-905. * |
Also Published As
Publication number | Publication date |
---|---|
CN101922914A (en) | 2010-12-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101922914B (en) | Crown information extraction method and system based on high spatial resolution remote sense image | |
CN102096818A (en) | Object-oriented automatic extracting method and system for outline and parameter of remote sensing image crown | |
Jing et al. | Automated delineation of individual tree crowns from LiDAR data by multi-scale analysis and segmentation | |
Duncanson et al. | An efficient, multi-layered crown delineation algorithm for mapping individual tree structure across multiple ecosystems | |
Bao et al. | Field-based architectural traits characterisation of maize plant using time-of-flight 3D imaging | |
Ke et al. | A review of methods for automatic individual tree-crown detection and delineation from passive remote sensing | |
Véga et al. | PTrees: A point-based approach to forest tree extraction from lidar data | |
CN101672915B (en) | High spatial resolution remote sensing image crown outline delineation system and method | |
Xu et al. | Automatic extraction of street trees’ nonphotosynthetic components from MLS data | |
CN103646246A (en) | Decision tree model based multispectral remote sensing image river information extraction method | |
CN103295018A (en) | Method for precisely recognizing fruits covered by branches and leaves | |
CN104331686A (en) | Manual assisting type recognition system of coverage rate of straw on soil surface | |
CN1996044B (en) | Canopy spatial statistics quantitative estimation method based on remote sensing image with fine spatial resolution | |
Chehata et al. | Object-based forest change detection using high resolution satellite images | |
Rashidi et al. | Extraction of ground points from LiDAR data based on slope and progressive window thresholding (SPWT) | |
CN101788685B (en) | Remote sensing earthquake damage information extracting and digging method based on pixels | |
Yu et al. | Regional sampling of forest canopy covers using UAV visible stereoscopic imagery for assessment of satellite-based products in Northeast China | |
CN105894501A (en) | Single-tree detection and crown describing method for high-resolution remote sensing image | |
CN116805413A (en) | Automatic calculation method for oil tea seedling stage phenotype characteristics based on three-dimensional point cloud | |
Li et al. | Street tree information extraction and dynamics analysis from mobile LiDAR point cloud | |
CN112381029B (en) | Method for extracting airborne LiDAR data building based on Euclidean distance | |
CN114241333A (en) | Method for accurately identifying new landslide region based on multi-source time sequence remote sensing image | |
CN116977593A (en) | Single wood segmentation method based on super-voxel concave-convex segmentation and color region growth | |
CN111768101B (en) | Remote sensing cultivated land change detection method and system taking account of physical characteristics | |
Jalbuena et al. | Object-based image analysis for mangroves extraction using LIDAR datasets and orthophoto |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120222 Termination date: 20180827 |