CN104318546A - Multi-scale analysis-based greenhouse field plant leaf margin extraction method and system - Google Patents
Multi-scale analysis-based greenhouse field plant leaf margin extraction method and system Download PDFInfo
- Publication number
- CN104318546A CN104318546A CN201410513634.7A CN201410513634A CN104318546A CN 104318546 A CN104318546 A CN 104318546A CN 201410513634 A CN201410513634 A CN 201410513634A CN 104318546 A CN104318546 A CN 104318546A
- Authority
- CN
- China
- Prior art keywords
- image
- edge
- pseudo
- obtains
- blade
- 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.)
- Granted
Links
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the digital image processing technical field and relates to a multi-scale analysis-based greenhouse field plant leaf margin extraction method and system. According to the method, differences of information of images in different scale spaces are utilized, and different segmentation methods are selected, and comprehensive analysis is performed, and as a result, ideal segmentation results can be obtained; after an experimental image is obtained, proper smoothing filtering is performed on the image, and classified processing is performed on different types of pseudo edges in the image; and based on different situations of Canny edge detection in a comprehensive scale space and OTSU threshold segmentation in different scales and different characteristics of various kinds of pseudo edges, with methods such as morphological processing and logical operation methods adopted, internal and external pseudo edges are removed through utilizing bitwise operation, and therefore, edge detection accuracy can be improved, and leaf recognition error rate can be reduced.
Description
Technical field
The present invention relates to digital image processing techniques field, particularly relate to a kind of greenhouse Field Plants blade edge extracting method based on multiscale analysis and system thereof.
Background technology
Blade is the manufacture organ of fruit tree nutrition material, is the basis of yield composition.Need in agricultural production to obtain a large amount of leaf growth information.For passing through Image Acquisition leaf growth information, realizing orchard and producing monitoring automatically, needing to carry out Image enhancing and dividing algorithm research to field fruit tree leaf image.Iamge Segmentation is the basis of graphical analysis and image understanding, can provide important information for further image procossing.Leaf image segmentation can provide important evidence for plant characteristics, as blade area calculate, blade disease and pest detects and blade three-dimensional reconstruction etc., thus can the upgrowth situation of Real-Time Monitoring plant, shift to an earlier date pre-preventing disease and pest etc.Contribute to scientic planting, improve crop yield etc.
Iamge Segmentation is according to concrete mission requirements, interested part (target) is separated from other information (background) of image, being the important component part of image procossing, is also the basis of graphical analysis and understanding, especially a computer vision technique.Target Segmentation is the committed step of graphical analysis and image understanding, simultaneously accurately complete plant leaf blade marginal information is also the primary prerequisite obtained based on the crop biomass of digital image processing techniques, such as upgrowth situation detection can be carried out, measuring chlorophyll content, disease and pest early warning, and the blade three-dimensional configuration of based target feature rebuilds recovery, and then realize harmless accurate three-dimensional plant leaf area calculating, and the visualized management etc. in digital orchard.No matter be the acquisition of the growth such as blade area girth, the stem diameter key element based on image, or nutrition condition, degree of ripeness detect, and the basis of all analyses is the accurate extraction of required target in image.Other gain of parameters be only based upon on target extraction method are accurately only valuable, and the analysis done based on these parameters and the conclusion obtained are only significant.Analyze correct texture, the shape information that can be provided for disease and pest and analyze accurately, the corresponding image points mated can be provided, the motion orientation that picking robot is correct can be provided in three-dimensional reconstruction.Only have and completely accurately Iamge Segmentation, target to be separated, the feature extraction of based target and parameter measurement just can weed out a large amount of image redundancy information, converted image half-tone information is to more abstract series of features data set, outstanding and complete expression important information, the convenient analysis to pictures subsequent and understanding.
Rim detection segmentation is then normally carried out on the basis of gradient image.Utilize the extreme value of image first order derivative or the zero point of second derivative as the basic foundation judging marginal point.The advantage of rim detection is that edge local is accurate, fast operation.Rim detection is generally divided into three steps: first utilize some edge detection operators to detect marginal point possible in image; Secondly, to there being certain thickness edge to carry out complicated edge thinning, obtain accurate, that thickness is pixel edge; Finally, edge closure technology is utilized to obtain the edge closed.But leaf image has following characteristics usually: background is complicated, blade is more, easy overlap, blade surface is smooth not, the heterochromia of vein and blade is comparatively large, and some local graded of leaf edges is not obvious, and blade is difficult to separate with petiole connecting place, branch overlapping, the on-the-spot leaf image of some fruit tree shows as the region of several different brightness due to illumination difference, inner gradient change is too obvious.These features cause split as a result, continuity and the closure at edge can not be ensured, there is many broken edge, pseudo-edges in a large number in high detail areas again, be difficult to determine which edge is real object edge.
Summary of the invention
(1) technical matters that will solve
The technical problem to be solved in the present invention is: propose a kind of greenhouse Field Plants blade edge extracting method based on multiscale analysis and system thereof, implements different disposal to dissimilar pseudo-edge, improves rim detection degree of accuracy, reduces leaf recognition error rate.
(2) technical scheme
For achieving the above object, the invention provides a kind of greenhouse Field Plants blade edge extracting method based on multiscale analysis, the method comprises:
S1. capture greenhouse Field Plants image, intercept the target subimage comprising single whole blade, described target subimage is converted to gray level image, use wiener filtering and noise reduction;
S2. adopt four layers of ' db5 ' wavelet decomposition to the image obtained after S1 denoising, and reconstruct ground floor image and obtain image f1, reconstruct third layer image obtains image f2;
S3. the image f1 obtained step S2 uses canny rim detection to obtain low yardstick edge image f3, use OTSU Threshold segmentation to obtain the OTSU Threshold segmentation image f4 of low yardstick, use OTSU Threshold segmentation to obtain the Threshold segmentation image f5 of high yardstick to the image f2 that step S2 obtains;
S4. the process of first kind pseudo-edge: carry out outside pseudo-edge process to the low yardstick threshold binary image f4 that step S3 obtains, obtains the exterior sheathing area image f6 for wiping outside pseudo-edge;
S5. the edge image f3 logical and obtained with step S3 again after the image f6 obtained step S4 uses logical not operation, obtains the image f7 of the outside pseudo-edge of erasing;
S6. in-profile process being carried out to the high yardstick threshold binary image f5 that step S3 obtains, obtaining the inner formword image f18 for wiping inner pseudo-edge;
S7. the inner formword image f18 obtained step S6 uses logical not operation, then with the image f7 logic and operation that step S5 obtains, obtain final edge image f19;
Preferably, carry out outside pseudo-edge process described in S4 to image f4, the exterior sheathing area image f6 obtained for wiping outside pseudo-edge comprises:
S4.1. be the circle of N1 pixel with radius be structural elements, by this structural elements to image f4 example etching operation, wherein N1 is positive integer;
S4.2. in the image that obtains of wadding warp S4.1 due to hole that etching operation causes;
S4.3. the connected region that in the image obtained through S4.2, area is maximum is chosen;
S4.4. be the circle of N2 pixel with radius be structural elements, carry out morphological erosion operation by this structural elements to the largest connected region that S4.3 chooses, wherein N2 is positive integer;
S4.5. the connected region that in the image obtained through S4.4, area is maximum is chosen;
S4.6. be the circle of N3 pixel with radius be structural elements, by this structural elements, morphological dilation is carried out to the largest connected region that S4.5 chooses, obtain the exterior sheathing image f6 for wiping outside pseudo-edge, wherein N3 is positive integer, and N3 is greater than N2 and N1 sum;
Preferably, N1 equals 2, N2 and equals 4, N4 and equal 8.
Preferably, describedly carry out in-profile process to image f5, the inner formword image f18 obtained for wiping inner pseudo-edge comprises:
S6.1. the process of Equations of The Second Kind pseudo-edge is carried out to the high yardstick threshold binary image f5 that step S3 obtains, obtain the image f11 wiping Equations of The Second Kind pseudo-edge;
S6.2. the 3rd class pseudo-edge process is carried out to image f5, obtain the image f16 wiping the 3rd class pseudo-edge;
S6.3. the image f16 process that image f11 and S6.2 obtained S6.1 obtains, obtains the image f18 for wiping inner pseudo-edge;
Preferably, carry out the process of Equations of The Second Kind pseudo-edge described in S6.1 to image f5, the image f11 obtaining wiping Equations of The Second Kind pseudo-edge comprises:
S6.1.1. construct the circular configuration unit that radius is N4, by this structural elements to image f5 example etching operation, to disconnect adhesion, wherein N4 is positive integer;
S6.1.2. the hole of image that obtains of wadding warp S6.1.1, chooses the connected region that area is maximum;
S6.1.3. construct the circular configuration unit that radius is N5, carry out morphological dilation by this structural elements to the maximum area connected domain chosen through S6.1.2, obtain image f8, wherein N5 is positive integer;
S6.1.4. the image f8 obtained with step S6.1.3 carries out second time OTSU Threshold segmentation for mask to the image f2 that step S2 obtains, and obtains image f9;
S6.1.5. to the image f7 that threshold binary image f9 and step S5 obtains, with image f7 logic XOR Map as f9, thus with the region be separated in the edge conjunction image f9 in image f7;
S6.1.6. the inner void of image that obtains of wadding warp S6.1.5, then chooses the connected region that area is maximum, using this connected region as the inner formword highlight area image f11 in order to wipe Equations of The Second Kind pseudo-edge.
Preferably, N4 equals 6, N5 and equals 3.
Preferably, describedly carry out the 3rd class pseudo-edge process to the 5th image f5, the image f16 obtaining wiping the 3rd class pseudo-edge comprises:
S6.2.1. after the image f8 obtained in S6.1.3 being removed the high light parts of images f9 obtained in S6.1.4, remaining region is designated as image f10;
S6.2.2. for the edge image f7 that image f10 and step S5 obtains, four arms of tectonic level direction and vertical direction respectively for the linear stauros constitutive element of N6 pixel is namely centered by that pixel that will expand, the linear crosswise structural elements of horizontal and vertical list pixel extension; By this structural elements, morphological dilation is carried out to image f7, thus the canny edge in small, broken bits disconnected under connecting low yardstick, wherein N6 is positive integer;
S6.2.3. morphologic thinning operation is carried out to the image obtained through S6.2.2, obtain image f12;
S6.2.4. the image f10 obtained by step S6.2.1 carries out logical not operation, then carries out logic and operation with image f12, and to obtain the part edge subimage of image f12 in image f10, the image of the internal edge remembered is image f13;
S6.2.5. the edge of the edge of image f13 and image f10 is carried out logic and operation, thus remain the edge of refinement and be positioned at the texture of image f10 inside, realize closing image f13, obtain image f14;
S6.2.6. construct radius be N7 circular configuration unit, by this structural elements to image f10 example etching operation, then with image f14 step-by-step logic XOR, to connect the pieces of debris in f10, split outside adhesion, the image remembered is f15 simultaneously, and wherein N7 is positive integer;
S6.2.7. blank map is as the hole in f15, chooses the inner vein template that the maximum connected region of area is erasing the 3rd class edge as image f16, f16.
Preferably, N6 equals 4, N7 and equals 4.
Preferably, described to image f11 and image f16 process, the image f18 obtained for wiping inner pseudo-edge comprises:
S6.3.1. the image f16 that the image f11 obtained in step S6.1 and step S6.2 obtains is carried out logic and operation, obtain image f17;
S6.3.2. the inner void in the image f17 obtained in S6.3.1 is filled;
S6.3.3. construct radius be the circle of N8 pixel as structural elements, carry out morphological erosion operation by this structural elements to the image obtained through S6.3.2, obtain the inner formword image f18 for wiping inner pseudo-edge, wherein N8 is positive integer.
Preferably, N8 equals 2.
The present invention also provides a kind of Field Plants field, greenhouse blade edge extraction system based on multiscale analysis, and this system comprises:
Pretreatment module: obtain and filtering for blade subgraph, comprising: first image capture device obtains the original image of plant leaf blade; Utilize the window close with leaf blade size, with particular step size, traversal cuts blade original image, and acquisition comprises the subregion of individual blade as target subimage; After obtaining blade subgraph, be converted into gray-scale map, wiener filtering process is carried out to gray level image, remove the noise of image;
Exterior contour processing module, processes first kind pseudo-edge, carries out db5 wavelet decomposition to former figure, the first yardstick is reconstructed and canny segmentation and first time OTSU segmentation; Obtain edge image and area image, Morphological scale-space is carried out to the area image obtained, obtains the exterior sheathing image removing outside adhesion, exterior sheathing image and edge image logical operation, obtain the canny edge image wiping outside adhesion;
The foreground portion subarea processing module of inner high light and secondary OTSU, Equations of The Second Kind pseudo-edge is processed, highlight area and part canny edge image are carried out logic XOR, namely by the part that part original canny edge splicing disconnects, filling out hole to the image after logical operation and getting the maximum connected region of area is Equations of The Second Kind target internal highlight bar; Rear and the canny edge logical and to the operation of this highlight bar logic NOT, obtains the image of erasing blade interior highlight bar pseudo-edge;
Inner vein pseudo-edge processing module, processes the 3rd class pseudo-edge, to refinement after canny edge swell, adopts the logic XOR canny closed edge of refinement to be connected by the different block in tertiary target region, and cuts outside adhesion; Hole is filled out to the image obtained, then etching operation, get namely the largest connected region of area removes pseudo-edge image-region as tertiary target region;
Blade integral edge acquisition module, logic and operation is carried out to the interior zone of obtained inside highlight area and removal the 3rd class pseudo-edge, after the morphological image obtained is filled, carry out morphological erosion operation again, the image obtained is inner formword region, by the edge image logic and operation obtained with the outside pseudo-edge of removal again after this area image logic NOT, the image obtained is final blade integral edge image.
The present invention also provides a kind of Field Plants field, greenhouse blade edge extraction system based on multiscale analysis, and it is characterized in that, this system comprises:
Canny link block: for canny edge conjunction fragment region, to obtain the inner formword for removing the inner pseudo-edge interference in canny edge, comprise: with the original canny edge line of part with logic XOR is carried out to the foreground area of secondary OTSU, the inside fragment of the foreground area of connecting secondary OTSU, the hole of image inside after filling logical operation, to obtain the largest connected region of prospect; By constructing linear cross structure unit, canny edge is expanded, then the refinement edge, inside that unlimited Refinement operation obtains the background area of secondary OTSU is carried out to the image after expanding; By the closed edge line of the outer peripheral logical and of the background area of this edge and secondary OTSU, logic XOR is carried out to this background area, is namely connected the image fragment in this region by the closed edge line after canny refinement, and split outside adhesion; Then fill the hole of image inside after computing, remove outside adhesion and the background connected region being connected inner fragment to obtain;
Dimensional analysis module: using the instrument of multiscale analysis as research characteristics of image, first at different scale spatially, the characteristics of image corresponding with different dividing methods carries out Research on differences, the difference of refining multiple dimensioned epigraph feature is expressed; By utilizing morphological operation and logical operation, by the image digitwise operation on single for difference yardstick, parse the leaf image complete edge removing outside and inner pseudo-edge.
Preferably, also comprise: classification processing module, for pseudo-edge comparatively meticulous under low yardstick and region thereof are divided into three major types: first kind pseudo-edge and region: the pseudo-edge of real blade outside, the pseudo-edge of the background parts namely after first time OTSU process and region; Equations of The Second Kind pseudo-edge and region: the inner and pseudo-edge of the GTG homogeneous area that brightness is slightly high of real blade, the pseudo-edge of the prospect namely after secondary OTSU process and region: the prospect part luma that the OTSU process of its brightness ratio obtains is slightly high are a part for the prospect that an OTSU process obtains; 3rd class edge and region: the texture pseudo-edge that real blade interior intensity differs greatly, namely this region is the background that secondary OTSU process obtains, and it is also a part for the prospect of an inter-class variance, and visual performance is that interior intensity differs greatly; Then for adopting different disposal routes for the nature and characteristic of the different pseudo-edge of this three class, the true edge not comprising above three class pseudo-edges is finally obtained.
(3) beneficial effect
Method of the present invention comes in every shape for From Strawberry Leaves, background is complicated and mutually overlapping, and the situation that the reflection of greenhouse black thin film causes the blade local greenhouse Field Plants blade segmentation result that causes of uneven illumination undesirable, and due to blade self clean mark, graded is larger, to the present situation that the detection of true edge is also disturbed to some extent, propose a kind of greenhouse Field Plants blade edge extracting method based on multiscale analysis, utilize the difference of different scale spatial image information, select different dividing methods, comprehensively analyze, obtain desirable segmentation result.After acquisition experimental image, suitable smothing filtering is carried out to image, for the different pseudo-edges existed in image, carry out the inside and outside pseudo-edge of process removing in a variety of ways, improve rim detection degree of accuracy, reduce leaf recognition error rate.In addition, a kind of new canny edge closure edge fit mode that the present invention also proposes, is processed canny edge by structure cross curve meta structure, the edge obtained more true and accurate.Using the instrument of multiscale analysis as research characteristics of image in this patent, edge extracting precision is high, and fast operation is applied widely.
Accompanying drawing explanation
Fig. 1 is based on the greenhouse Field Plants blade edge extracting method process flow diagram of multiscale analysis;
The outside pseudo-edge processing flow chart of Fig. 2;
Fig. 3 Equations of The Second Kind pseudo-edge processing flow chart;
Fig. 4 the 3rd class pseudo-edge processing flow chart;
Embodiment
The greenhouse Field Plants blade edge extraction algorithm based on multiscale analysis that the present invention proposes and system thereof, be described in detail as follows in conjunction with the accompanying drawings and embodiments.
Method of the present invention and system thereof pass through experiment sieving, different for greenhouse Field Plants leaf morphology, background is complicated and mutually overlapping, and the situation that the reflection of greenhouse black thin film causes the blade local greenhouse Field Plants blade segmentation result that causes of uneven illumination undesirable, and due to blade self clean mark, graded is larger, to the present situation that the detection of true edge is also disturbed to some extent, a kind of greenhouse Field Plants blade edge extracting method based on multiscale analysis proposed by the invention and system thereof, utilize the difference of different scale spatial image information, select different dividing methods, comprehensively analyze, can be complete, accurately, high success rate ground extracts the plant image scene blade edge obtained by real-time on-line system.It is to comprise the subregion of individual blade as target image, using locked cut zone subgraph near subregion as handling object; Comprehensive utilization graph and image processing operator, carries out filtering to region subgraph and strengthens process; Propose different partitioning algorithms for dissimilar pseudo-edge, utilize Morphological scale-space, adopt OTSU threshold value and canny operator to split the method combined, carry out the logical operation process of the mutual supplement with each other's advantages of extracting image border.
Greenhouse Field Plants is described for strawberry in the following embodiments, it should be appreciated by those skilled in the art that the greenhouse Field Plants blade edge extracting method of multiscale analysis of the present invention and system thereof go for other greenhouse Field Plants equally.
As shown in Figure 1, according to the present invention, a kind of greenhouse Field Plants blade edge extracting method based on multiscale analysis, comprises step:
S1. capture the strawberry image under greenhouse natural light condition, cut-away view picture, obtain the target subimage comprising single whole blade.Reading images, after being converted to gray level image, uses wiener filtering and noise reduction;
Intercepting the method comprising the target subimage of single whole blade is: first image capture device obtains the original image of plant leaf blade; Utilize the window close with leaf blade size, with particular step size, traversal cuts blade original image, and acquisition comprises the subregion of individual blade as target subimage.
S2. four layers of ' db5 ' wavelet decomposition are adopted to the image after S1 denoising, and reconstruct ground floor and third layer image, be designated as f1 and f2 respectively;
S3. use canny rim detection and OTSU Threshold segmentation to the f1 that S2 obtains, use OTSU Threshold segmentation to f2, the image of scoring respectively after cutting is f3, f4 and f5.Wherein f3 is the canny edge image of low yardstick, and f4 is the OTSU Threshold segmentation image of low yardstick, and f5 is the OTSU Threshold segmentation image of high yardstick;
It is emphasised that pseudo-edge comparatively meticulous under low yardstick and region thereof are divided into three major types by the present invention: 1. first kind edge and region: the pseudo-edge of real blade outside and region (first time inter-class variance the pseudo-edge of background (background) part and region); 2. Equations of The Second Kind edge and region: the inner and pseudo-edge of the GTG homogeneous area that brightness is slightly high of real blade (pseudo-edge of the foreground (prospect) of second time inter-class variance and region: between its brightness ratio first time class, foreground part luma is slightly high, the part of foreground for first time inter-class variance); 3. the 3rd class edge and region: the texture pseudo-edge (background of second time inter-class variance is a part of the foreground of inter-class variance for the first time, shows as interior intensity and differs greatly) that real blade interior intensity differs greatly.Nature and characteristic for the different pseudo-edge of this three class adopts different disposal routes, finally obtains the true edge not comprising above three class pseudo-edges.
S4. the process of first kind pseudo-edge is carried out in this step: the low yardstick threshold binary image f4 obtained S3 carries out outside pseudo-edge process (refer to Fig. 2 and below to its corresponding explanation), obtains the first kind exterior sheathing region f6 for wiping outside pseudo-edge;
S5. the exterior sheathing region f6 that edge image f3 and S4 obtained S3 obtains, obtains the edge image f7 removing outside pseudo-edge by logical operation.To f6 use logical not operation, then with edge image f3 logical and, obtain erasing outside pseudo-edge edge image f7, be expressed as: f7=f3 & (~ f6);
S6. to the high yardstick threshold binary image f5 application in-profile handling procedure that S3 obtains, the inner formword region f18 for wiping inner pseudo-edge is obtained;
S7. the edge image f7 that inner formword region f18 and S5 obtained S6 obtains carries out logical operation and obtains final edge image f19: namely internally template area f18 uses logical not operation, again with the f7 logic and operation wiping outside pseudo-edge, obtain final edge image f19, be expressed as: f19=(~ f18) & f7.
As shown in Figure 2, S4 comprises further:
S4.1. be the circle of 2 pixels with radius be structural elements, by this structural elements to f4 example etching operation, disconnect more weak adhesion and contact, here, the radius of structural elements also can be other sizes;
S4.2. due to hole that etching operation causes in the image after wadding warp S4.1 process;
S4.3. the connected region that area is maximum is chosen;
S4.4. be the circle of 4 pixels with radius be structural elements, by this structural elements, morphological erosion operation carried out to largest connected region, disconnect stronger adhesion and contact; Here, the radius of structural elements also can be other sizes.
S4.5. the connected region that the area in the image obtained after selecting step 4.4 process is maximum;
S4.6. be the circle of 8 with radius be structural elements, by this structural elements, morphological dilation is carried out to largest connected region, using the region of these 2 pixels larger than true edge as the exterior sheathing f6 being used for wiping outside pseudo-edge, here, the radius of this structural elements also can be other sizes
Wherein, S6 specifically comprises:
S6.1. to the process of Equations of The Second Kind pseudo-edge: carry out in-profile process to the high yardstick threshold binary image f5 that step S3 obtains, the inner formword highlight area image f11 wiping Equations of The Second Kind pseudo-edge is obtained;
S6.2. the 3rd class pseudo-edge process is carried out to the high yardstick threshold binary image f5 that step S3 obtains, obtain the inner vein template f16 wiping the 3rd class pseudo-edge;
S6.3. by f11 and f16 process, the inner formword region f18 for wiping inner pseudo-edge is obtained;
As shown in Figure 3, step S6.1 specifically comprises:
S6.1.1. for the high yardstick OTSU Threshold segmentation image f5 that S3 obtains, structure radius is the circular configuration unit of 6, by this structural elements to f5 example etching operation, disconnects adhesion;
S6.1.2. fill the hole of image after disconnecting adhesion, get the connected region that area is maximum;
S6.1.3. construct the circular configuration unit that radius is 3, carry out morphological dilation to maximum area connected domain, after note process, image is f8, and here, the radius of this structural elements also can be other sizes;
S6.1.4. be that mask carries out secondary OTSU Threshold segmentation to the reconstructed image f2 that S2 obtains with f8, be designated as f9;
The edge image f7 of the threshold binary image f9 obtained after S6.1.5. splitting for secondary OTSU and the outside pseudo-edge of removing, with f7 logic XOR f9, with the region be separated in the edge conjunction f9 in f7, can cut outside adhesion simultaneously, be expressed as: f7^f9; Realize wiring with canny edge herein: the fragment connecting f9 inside with the edge (part at original edge is taken out with f6: f7=f6 & f3) of f7, cut outside adhesion simultaneously;
S6.1.6. fill the inner void produced in S6.1.5, and get the maximum connected region of area, using the inner formword highlight area image f11 of this region as erasing Equations of The Second Kind pseudo-edge;
As shown in Figure 4, the process of the 3rd class pseudo-edge is comprised further:
S6.2.1. f8 in S6.1.4 is removed the remaining region of highlight regions f9 and be designated as f10;
S6.2.2. for f10 and the edge image f7 removing outside pseudo-edge, the linear stauros constitutive element of vertical each 4 pixels of tectonic level namely centered by that pixel that will expand, the linear crosswise structural elements that horizontal and vertical list pixel extends.Morphological dilation is carried out to image f7, the canny edge in small, broken bits disconnected under connecting low yardstick;
S6.2.3. carry out morphologic thinning operation to interconnective expansion edge, note image is f12;
S6.2.4. obtain the part edge subimage of refinement edge image f12 in f10, remember that the image of this internal edge is f13, be expressed as: f13=f12 & (~ f10);
S6.2.5. the edge of f13 is superposed with the edge of f10, thus remain the edge of refinement and be positioned at the texture of f10 inside, realize closing f13, remember that the image now obtained is f14, be expressed as: f14=f13 & (edge line of f10);
S6.2.6., after f10 being corroded four pixels, with image f14 logic XOR, to connect the pieces of debris in f10, split outside adhesion, the image now remembered is f15 simultaneously; Be expressed as: f15=f10^f14;
S6.2.7. blank map is as the hole in f15, and gets the maximum connected region of area, and note image is f16, using the inner vein template of f16 as erasing the 3rd class edge;
Wherein, inner formword image f18 obtaining step comprises further:
S6.3.1. the inner vein template f16 region that blade interior height light image f11 and S6.2 obtained in S6.1 obtains is added, obtains image f17, be expressed as: f17=f16 & f11;
S6.3.2. blank map is as the inner void produced in f17;
S6.3.3. structure radius is the circle of 2 is structural elements, carrying out morphological erosion operation by this structural elements to filling out the image behind hole, obtaining 2 pixels less of true edge, for wiping the inner formword image f18 of inner pseudo-edge, here, the radius of this structural elements also can be other sizes.
Morphological erosion and expansion refer to: using certain structure oneself defined as structural elements, utilize the size of this structural elements for benchmark corrosion or dilation pixel value.Corrosion is exactly that the pixel value of the pixel value of each pixel of original image and each pixel of structural elements is subtracted each other according to certain rule, and expansion is exactly that pixel value is increased.Morphologic thinning is exactly the tiny branch removing the edge image extracted.Morphologic thinning is exactly the tiny branch removing the edge image extracted.
The invention allows for a kind of greenhouse Field Plants blade edge extraction system based on multiscale analysis, this system comprises:
Pretreatment module: blade subgraph obtains and filtering, comprising: first image capture device obtains the original image of plant leaf blade; Utilize the window close with leaf blade size, with particular step size, traversal cuts blade original image, and acquisition comprises the subregion of individual blade as target subimage.After obtaining blade subgraph, be converted into gray-scale map, wiener filtering process is carried out to gray level image, remove the noise of image.
Exterior contour processing module, first kind pseudo-edge is processed, db5 wavelet decomposition is carried out to former gray level image, first yardstick is reconstructed and canny segmentation and first time OTSU segmentation, obtain edge image and area image, Morphological scale-space is carried out to the area image obtained, obtains the exterior sheathing image removing outside adhesion, exterior sheathing image and edge image logical operation, obtain the canny edge image wiping outside adhesion.
Inner highlight area processing module, processes Equations of The Second Kind pseudo-edge, highlight area and canny edge image is carried out logic XOR, namely by the part that part original canny edge splicing disconnects.Filling out hole to the image after logical operation and getting the maximum connected region of area is Equations of The Second Kind target internal highlight bar, the rear and canny edge logical and to the operation of this highlight bar logic NOT, obtains the image of erasing blade interior highlight bar pseudo-edge.
Inner vein pseudo-edge processing module, processes the 3rd class pseudo-edge, to refinement after canny edge swell, adopts XOR computing refinement canny closed edge to be connected by the different block in tertiary target region, and cuts outside adhesion.Hole is filled out to the image obtained, then etching operation, get namely the largest connected region of area removes pseudo-edge image-region as tertiary target region.
Blade integral edge acquisition module: logic and operation is carried out to the interior zone of obtained inside highlight area and removal the 3rd class pseudo-edge, after the morphological image obtained is filled, carry out morphological erosion operation again, the image obtained is inner formword region, by the edge image logic and operation obtained with the outside pseudo-edge of removal again after this area image logic NOT, the image obtained is final blade integral edge image.
The invention allows for a kind of hothouse plants field blade edge extraction system based on multiscale analysis, it is characterized in that, this system comprises:
(1) Canny link block: a kind of method proposing utilization canny edge conjunction fragment region newly, to obtain the inner formword that may be used for removing the inner pseudo-edge interference in canny edge.
With the original canny edge line of part with logic XOR is carried out to the foreground region of secondary OTSU, the inside fragment in the foreground region of connecting secondary OTSU, the hole of image inside after filling logical operation, to obtain the largest connected region of foreground.
By constructing linear cross structure unit, canny edge is expanded, then the refinement edge, inside that unlimited Refinement operation obtains the background region of secondary OTSU is carried out to the image after expanding.By the closed edge line of the outer peripheral logical and in the background region of this edge and secondary OTSU.Use the closed edge line of this refinement to carry out logic XOR to this background region, then fill the hole of image inside after computing, remove outside adhesion and the background connected region being connected inner fragment to obtain.
(2) dimensional analysis module: using the instrument of multiscale analysis as research characteristics of image, first at different scale spatially, the characteristics of image corresponding with different dividing methods carries out Research on differences, the difference of refining multiple dimensioned epigraph feature is expressed.By utilizing morphological operation and logical operation, by the image digitwise operation on single for difference yardstick, parse the leaf image complete edge removing outside and inner pseudo-edge.
This system can also comprise:
Classification processing module, for being divided into three major types by pseudo-edge comparatively meticulous under low yardstick and region thereof: 1. first kind pseudo-edge and region: the pseudo-edge (pseudo-edge of the background part after first time inter-class variance OTSU process and region) of real blade outside; 2. Equations of The Second Kind pseudo-edge and region: the inner and pseudo-edge of the GTG homogeneous area that brightness is slightly high of real blade (pseudo-edge of the foreground after secondary inter-class variance OTSU process and region: the foreground part luma that its brightness ratio inter-class variance obtains is slightly high, the part for the foreground that inter-class variance obtains); 3. the 3rd class edge and region: (this region is the background that secondary inter-class variance obtains to the texture pseudo-edge that real blade interior intensity differs greatly, it is also a part of the foreground of an inter-class variance, and visual performance is that interior intensity differs greatly).Nature and characteristic for the different pseudo-edge of this three class adopts different disposal routes, finally obtains the true edge not comprising above three class pseudo-edges.
Greenhouse Field Plants blade edge extracting method based on multiscale analysis proposed by the invention and system thereof, have the following advantages:
After acquisition experimental image, suitable smothing filtering is carried out to image.For the dissimilar pseudo-edge existed in image, classification process is carried out to the pseudo-edge in image.And use the method for dimensional analysis, on different single yardsticks, the inside and outside pseudo-edge of process removing is carried out in step-by-step in a variety of ways, improves rim detection degree of accuracy, reduces leaf recognition error rate.In addition, a kind of method of utilization canny edge conjunction fragment region newly that also proposes of the present invention.And by structure cross curve meta structure, canny edge is processed, the closed edge in order to join domain fragment obtained.Make obtained blade whole object edge more true and accurate.Using the instrument of multiscale analysis as research characteristics of image in this patent, edge extracting precision is high, and fast operation is applied widely.
Of the present invention the proposed greenhouse Field Plants blade edge extracting method based on multiscale analysis, can be realized by various hardware, software and firmware, such as can carry out image procossing to complete method of the present invention in MATLAB, but the present invention is not limited thereto.
Above embodiment is only for illustration of the present invention; and be not limitation of the present invention; the those of ordinary skill of relevant technical field; without departing from the spirit and scope of the present invention; can also make a variety of changes and modification; therefore all equivalent technical schemes also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.
Claims (13)
1., based on a greenhouse Field Plants blade edge extracting method for multiscale analysis, it is characterized in that, the method comprising the steps of:
S1. capture greenhouse Field Plants image, intercept the target subimage comprising single whole blade, described target subimage is converted to gray level image, use wiener filtering and noise reduction;
S2. adopt four layers of ' db5 ' wavelet decomposition to the image obtained after S1 denoising, and reconstruct ground floor image and obtain image f1, reconstruct third layer image obtains image f2;
S3. the image f1 obtained step S2 uses canny rim detection to obtain low yardstick edge image f3, use OTSU Threshold segmentation to obtain the OTSU Threshold segmentation image f4 of low yardstick, use OTSU Threshold segmentation to obtain the Threshold segmentation image f5 of high yardstick to the image f2 that step S2 obtains;
S4. the process of first kind pseudo-edge: carry out outside pseudo-edge process to the low yardstick threshold binary image f4 that step S3 obtains, obtains the exterior sheathing area image f6 for wiping outside pseudo-edge;
S5. the edge image f3 logical and obtained with step S3 again after the image f6 obtained step S4 uses logical not operation, obtains the image f7 of the outside pseudo-edge of erasing;
S6. in-profile process being carried out to the high yardstick threshold binary image f5 that step S3 obtains, obtaining the inner formword image f18 for wiping inner pseudo-edge;
S7. the inner formword image f18 obtained step S6 uses logical not operation, then with the image f7 logic and operation that step S5 obtains, obtain final edge image f19.
2., as claimed in claim 1 based on the greenhouse Field Plants blade edge extracting method of multiscale analysis, it is characterized in that, carry out outside pseudo-edge process described in S4 to image f4, the exterior sheathing area image f6 obtained for wiping outside pseudo-edge comprises:
S4.1. be the circle of N1 pixel with radius be structural elements, by this structural elements to image f4 example etching operation, wherein N1 is positive integer;
S4.2. in the image that obtains of wadding warp S4.1 due to hole that etching operation causes;
S4.3. the connected region that in the image obtained through S4.2, area is maximum is chosen;
S4.4. be the circle of N2 pixel with radius be structural elements, carry out morphological erosion operation by this structural elements to the largest connected region that S4.3 chooses, wherein N2 is positive integer;
S4.5. the connected region that in the image obtained through S4.4, area is maximum is chosen;
S4.6. be the circle of N3 pixel with radius be structural elements, by this structural elements, morphological dilation is carried out to the largest connected region that S4.5 chooses, obtain the exterior sheathing image f6 for wiping outside pseudo-edge, wherein N3 is positive integer, and N3 is greater than N2 and N1 sum.
3. as claimed in claim 2 based on the greenhouse Field Plants blade edge extracting method of multiscale analysis, it is characterized in that, N1 equals 2, N2 and equals 4, N4 and equal 8.
4. as described in claim 1 based on the greenhouse Field Plants blade edge extracting method of multiscale analysis, it is characterized in that, describedly carry out in-profile process to image f5, the inner formword image f18 obtained for wiping inner pseudo-edge comprises:
S6.1. the process of Equations of The Second Kind pseudo-edge is carried out to the high yardstick threshold binary image f5 that step S3 obtains, obtain the image f11 wiping Equations of The Second Kind pseudo-edge;
S6.2. the 3rd class pseudo-edge process is carried out to image f5, obtain the image f16 wiping the 3rd class pseudo-edge;
S6.3. the image f16 process that image f11 and S6.2 obtained S6.1 obtains, obtains the image f18 for wiping inner pseudo-edge.
5. as claimed in claim 4 based on the greenhouse Field Plants blade edge extracting method of multiscale analysis, it is characterized in that, carry out the process of Equations of The Second Kind pseudo-edge described in S6.1 to image f5, the image f11 obtaining wiping Equations of The Second Kind pseudo-edge comprises:
S6.1.1. construct the circular configuration unit that radius is N4, by this structural elements to image f5 example etching operation, to disconnect adhesion, wherein N4 is positive integer;
S6.1.2. the hole of image that obtains of wadding warp S6.1.1, chooses the connected region that area is maximum;
S6.1.3. construct the circular configuration unit that radius is N5, carry out morphological dilation by this structural elements to the maximum area connected domain chosen through S6.1.2, obtain image f8, wherein N5 is positive integer;
S6.1.4. the image f8 obtained with step S6.1.3 carries out second time OTSU Threshold segmentation for mask to the image f2 that step S2 obtains, and obtains image f9;
S6.1.5. to the image f7 that threshold binary image f9 and step S5 obtains, with image f7 logic XOR Map as f9, thus with the region be separated in the edge conjunction image f9 in image f7;
S6.1.6. the inner void of image that obtains of wadding warp S6.1.5, then chooses the connected region that area is maximum, using this connected region as the inner formword highlight area image f11 in order to wipe Equations of The Second Kind pseudo-edge.
6., as claimed in claim 5 based on the greenhouse Field Plants blade edge extracting method of multiscale analysis, it is characterized in that, N4 equals 6, N5 and equals 3.
7. as claimed in claim 4 based on the greenhouse Field Plants blade edge extracting method of multiscale analysis, it is characterized in that, describedly carry out the 3rd class pseudo-edge process to the 5th image f5, the image f16 obtaining wiping the 3rd class pseudo-edge comprises:
S6.2.1. after the image f8 obtained in S6.1.3 being removed the high light parts of images f9 obtained in S6.1.4, remaining region is designated as image f10;
S6.2.2. for the edge image f7 that image f10 and step S5 obtains, four arms of tectonic level direction and vertical direction are respectively the linear stauros constitutive element of N6 pixel, namely construct centered by that pixel that will expand, linear crosswise structural elements that horizontal and vertical four direction extends N6 single pixel; By this structural elements, morphological dilation is carried out to image f7, thus the canny edge in small, broken bits disconnected under connecting low yardstick, wherein N6 is positive integer;
S6.2.3. morphologic thinning operation is carried out to the image obtained through S6.2.2, obtain image f12;
S6.2.4. the image f10 obtained by step S6.2.1 carries out logical not operation, then carries out logic and operation with image f12, and to obtain the part edge subimage of image f12 in image f10, the image of the internal edge remembered is image f13;
S6.2.5. the edge of the edge of image f13 and image f10 is carried out logic and operation, thus remain the edge of refinement and be positioned at the texture of image f10 inside, realize closing image f13, obtain image f14;
S6.2.6. construct radius be N7 circular configuration unit, by this structural elements to image f10 example etching operation, then with image f14 step-by-step logic XOR, to connect the pieces of debris in f10, split outside adhesion, the image remembered is f15 simultaneously, and wherein N7 is positive integer;
S6.2.7. blank map is as the hole in f15, chooses the inner vein template that the maximum connected region of area is erasing the 3rd class edge as image f16, f16.
8., as claimed in claim 7 based on the greenhouse Field Plants blade edge extracting method of multiscale analysis, it is characterized in that, N6 equals 4, N7 and equals 4.
9., as claimed in claim 4 based on the greenhouse Field Plants blade edge extracting method of multiscale analysis, it is characterized in that, described to image f11 and image f16 process, the image f18 obtained for wiping inner pseudo-edge comprises:
S6.3.1. the image f16 that the image f11 obtained in step S6.1 and step S6.2 obtains is carried out logic and operation, obtain image f17;
S6.3.2. the inner void in the image f17 obtained in S6.3.1 is filled;
S6.3.3. construct radius be the circle of N8 pixel as structural elements, carry out morphological erosion operation by this structural elements to the image obtained through S6.3.2, obtain the inner formword image f18 for wiping inner pseudo-edge, wherein N8 is positive integer.
10., as claimed in claim 9 based on the greenhouse Field Plants blade edge extracting method of multiscale analysis, it is characterized in that, N8 equals 2.
11. 1 kinds of Field Plants field, greenhouse blade edge extraction systems based on multiscale analysis, it is characterized in that, this system comprises:
Pretreatment module: obtain and filtering for blade subgraph, comprising: first image capture device obtains the original image of plant leaf blade; Utilize the window close with leaf blade size, with particular step size, traversal cuts blade original image, and acquisition comprises the subregion of individual blade as target subimage; After obtaining blade subgraph, be converted into gray-scale map, wiener filtering process is carried out to gray level image, remove the noise of image;
Exterior contour processing module, processes first kind pseudo-edge, carries out db5 wavelet decomposition to former figure, the first yardstick is reconstructed and canny segmentation and first time OTSU segmentation; Obtain edge image and area image, Morphological scale-space is carried out to the area image obtained, obtains the exterior sheathing image removing outside adhesion, exterior sheathing image and edge image logical operation, obtain the canny edge image wiping outside adhesion;
The foreground portion subarea processing module of inner high light and secondary OTSU, Equations of The Second Kind pseudo-edge is processed, highlight area and part canny edge image are carried out logic XOR, namely by the part that part original canny edge splicing disconnects, filling out hole to the image after logical operation and getting the maximum connected region of area is Equations of The Second Kind target internal highlight bar; Rear and the canny edge logical and to the operation of this highlight bar logic NOT, obtains the image of erasing blade interior highlight bar pseudo-edge;
Inner vein pseudo-edge processing module, processes the 3rd class pseudo-edge, to refinement after canny edge swell, adopts the logic XOR canny closed edge of refinement to be connected by the different block in tertiary target region, and cuts outside adhesion; Hole is filled out to the image obtained, then etching operation, get namely the largest connected region of area removes pseudo-edge image-region as tertiary target region;
Blade integral edge acquisition module, logic and operation is carried out to the interior zone of obtained inside highlight area and removal the 3rd class pseudo-edge, after the morphological image obtained is filled, carry out morphological erosion operation again, the image obtained is inner formword region, by the edge image logic and operation obtained with the outside pseudo-edge of removal again after this area image logic NOT, the image obtained is final blade integral edge image.
12. 1 kinds of Field Plants field, greenhouse blade edge extraction systems based on multiscale analysis, it is characterized in that, this system comprises:
Canny link block: for canny edge conjunction fragment region, to obtain the inner formword for removing the inner pseudo-edge interference in canny edge, comprise: with the original canny edge line of part with logic XOR is carried out to the foreground area of secondary OTSU, the inside fragment of the foreground area of connecting secondary OTSU, the hole of image inside after filling logical operation, to obtain the largest connected region of prospect; By constructing linear cross structure unit, canny edge is expanded, then the refinement edge, inside that unlimited Refinement operation obtains the background area of secondary OTSU is carried out to the image after expanding; By the closed edge line of the outer peripheral logical and of the background area of this edge and secondary OTSU, logic XOR is carried out to this background area, is namely connected the image fragment in this region by the closed edge line after canny refinement, and split outside adhesion; Then fill the hole of image inside after computing, remove outside adhesion and the background connected region being connected inner fragment to obtain;
Dimensional analysis module: using the instrument of multiscale analysis as research characteristics of image, first at different scale spatially, the characteristics of image corresponding with different dividing methods carries out Research on differences, the difference of refining multiple dimensioned epigraph feature is expressed; By utilizing morphological operation and logical operation, by the image digitwise operation on single for difference yardstick, parse the leaf image complete edge removing outside and inner pseudo-edge.
Field Plants field, the greenhouse blade edge extraction system based on multiscale analysis of 13. claims 12, it is characterized in that, also comprise: classification processing module, for pseudo-edge comparatively meticulous under low yardstick and region thereof are divided into three major types: first kind pseudo-edge and region: the pseudo-edge of real blade outside, the pseudo-edge of the background parts namely after first time OTSU process and region; Equations of The Second Kind pseudo-edge and region: the inner and pseudo-edge of the GTG homogeneous area that brightness is slightly high of real blade, the pseudo-edge of the prospect namely after secondary OTSU process and region: the prospect part luma that the OTSU process of its brightness ratio obtains is slightly high are a part for the prospect that an OTSU process obtains; 3rd class edge and region: the texture pseudo-edge that real blade interior intensity differs greatly, namely this region is the background that secondary OTSU process obtains, and it is also a part for the prospect of an inter-class variance, and visual performance is that interior intensity differs greatly; Then for adopting different disposal routes for the nature and characteristic of the different pseudo-edge of this three class, the true edge not comprising above three class pseudo-edges is finally obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410513634.7A CN104318546B (en) | 2014-09-29 | 2014-09-29 | Multi-scale analysis-based greenhouse field plant leaf margin extraction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410513634.7A CN104318546B (en) | 2014-09-29 | 2014-09-29 | Multi-scale analysis-based greenhouse field plant leaf margin extraction method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104318546A true CN104318546A (en) | 2015-01-28 |
| CN104318546B CN104318546B (en) | 2017-04-26 |
Family
ID=52373772
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410513634.7A Active CN104318546B (en) | 2014-09-29 | 2014-09-29 | Multi-scale analysis-based greenhouse field plant leaf margin extraction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104318546B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104809721A (en) * | 2015-04-09 | 2015-07-29 | 香港中文大学深圳研究院 | Segmentation method and device of cartoon |
| CN106296662A (en) * | 2016-07-28 | 2017-01-04 | 北京农业信息技术研究中心 | Maize leaf image partition method and device under field conditions |
| CN106468543A (en) * | 2015-08-21 | 2017-03-01 | 浙江托普云农科技股份有限公司 | A kind of method for measuring leaf area based on image procossing |
| CN106570876A (en) * | 2016-10-24 | 2017-04-19 | 南京理工大学 | Ghost imaging image edge extracting method |
| CN106815819A (en) * | 2017-01-24 | 2017-06-09 | 河南工业大学 | Many strategy grain worm visible detection methods |
| CN108564020A (en) * | 2018-04-08 | 2018-09-21 | 陕西科技大学 | Micro- gesture identification method based on panorama 3D rendering |
| CN109584240A (en) * | 2018-12-20 | 2019-04-05 | 成都理工大学 | Come down rear crack displacement image-recognizing method |
| CN110363784A (en) * | 2019-06-28 | 2019-10-22 | 青岛理工大学 | A kind of recognition methods being overlapped fruit |
| CN111311573A (en) * | 2020-02-12 | 2020-06-19 | 贵州理工学院 | Branch determination method and device and electronic equipment |
| CN111696125A (en) * | 2020-06-17 | 2020-09-22 | 广西科技大学 | Method for extracting edges of overlapped blades |
| CN113344959A (en) * | 2021-08-06 | 2021-09-03 | 山东捷瑞数字科技股份有限公司 | Residual material analysis processing method and device and material conveying system |
| CN113517877A (en) * | 2021-04-30 | 2021-10-19 | 华中科技大学 | Steel wire rope online detection signal noise reduction method and system based on generalized morphological filtering |
| CN117036359A (en) * | 2023-10-10 | 2023-11-10 | 成都中轨轨道设备有限公司 | Contact net geometric parameter measurement method based on binocular machine vision |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107704841A (en) * | 2017-10-25 | 2018-02-16 | 云南电网有限责任公司电力科学研究院 | A kind of image/video abnormal state detection identifies system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101286233A (en) * | 2008-05-19 | 2008-10-15 | 重庆邮电大学 | Fuzzy edge detection method based on object cloud |
| CN102226907A (en) * | 2011-05-24 | 2011-10-26 | 武汉嘉业恒科技有限公司 | License plate positioning method and apparatus based on multiple characteristics |
| US20140201126A1 (en) * | 2012-09-15 | 2014-07-17 | Lotfi A. Zadeh | Methods and Systems for Applications for Z-numbers |
-
2014
- 2014-09-29 CN CN201410513634.7A patent/CN104318546B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101286233A (en) * | 2008-05-19 | 2008-10-15 | 重庆邮电大学 | Fuzzy edge detection method based on object cloud |
| CN102226907A (en) * | 2011-05-24 | 2011-10-26 | 武汉嘉业恒科技有限公司 | License plate positioning method and apparatus based on multiple characteristics |
| US20140201126A1 (en) * | 2012-09-15 | 2014-07-17 | Lotfi A. Zadeh | Methods and Systems for Applications for Z-numbers |
Non-Patent Citations (3)
| Title |
|---|
| BAIHE LANG 等: "An Adaptive Edge Detection Method Based on Canny Operator", 《ADVANCED MATERIALS RESEARCH》 * |
| JIANLUN WANG 等: "An Adaptive Thresholding algorithm of field leaf image", 《COMPUTERS AND ELECTRONICS IN AGRICULTURE》 * |
| 董金勇 等: "田间枣树叶片复杂目标图像综合分割方法", 《农业机械学报》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104809721A (en) * | 2015-04-09 | 2015-07-29 | 香港中文大学深圳研究院 | Segmentation method and device of cartoon |
| CN104809721B (en) * | 2015-04-09 | 2017-11-28 | 香港中文大学深圳研究院 | A kind of caricature dividing method and device |
| CN106468543A (en) * | 2015-08-21 | 2017-03-01 | 浙江托普云农科技股份有限公司 | A kind of method for measuring leaf area based on image procossing |
| CN106296662A (en) * | 2016-07-28 | 2017-01-04 | 北京农业信息技术研究中心 | Maize leaf image partition method and device under field conditions |
| CN106296662B (en) * | 2016-07-28 | 2019-07-02 | 北京农业信息技术研究中心 | Maize leaf image partition method and device under field conditions |
| CN106570876B (en) * | 2016-10-24 | 2019-05-07 | 南京理工大学 | A method of extracting terrible image edge |
| CN106570876A (en) * | 2016-10-24 | 2017-04-19 | 南京理工大学 | Ghost imaging image edge extracting method |
| CN106815819B (en) * | 2017-01-24 | 2019-08-13 | 河南工业大学 | More strategy grain worm visible detection methods |
| CN106815819A (en) * | 2017-01-24 | 2017-06-09 | 河南工业大学 | Many strategy grain worm visible detection methods |
| CN108564020B (en) * | 2018-04-08 | 2021-07-13 | 陕西科技大学 | Micro-gesture recognition method based on panoramic 3D image |
| CN108564020A (en) * | 2018-04-08 | 2018-09-21 | 陕西科技大学 | Micro- gesture identification method based on panorama 3D rendering |
| CN109584240A (en) * | 2018-12-20 | 2019-04-05 | 成都理工大学 | Come down rear crack displacement image-recognizing method |
| CN109584240B (en) * | 2018-12-20 | 2022-05-03 | 成都理工大学 | Landslide trailing edge crack displacement image identification method |
| CN110363784A (en) * | 2019-06-28 | 2019-10-22 | 青岛理工大学 | A kind of recognition methods being overlapped fruit |
| CN111311573A (en) * | 2020-02-12 | 2020-06-19 | 贵州理工学院 | Branch determination method and device and electronic equipment |
| CN111311573B (en) * | 2020-02-12 | 2024-01-30 | 贵州理工学院 | Branch determination method and device and electronic equipment |
| CN111696125A (en) * | 2020-06-17 | 2020-09-22 | 广西科技大学 | Method for extracting edges of overlapped blades |
| CN111696125B (en) * | 2020-06-17 | 2022-05-24 | 广西科技大学 | Method for extracting edges of overlapped blades |
| CN113517877A (en) * | 2021-04-30 | 2021-10-19 | 华中科技大学 | Steel wire rope online detection signal noise reduction method and system based on generalized morphological filtering |
| CN113344959A (en) * | 2021-08-06 | 2021-09-03 | 山东捷瑞数字科技股份有限公司 | Residual material analysis processing method and device and material conveying system |
| CN113344959B (en) * | 2021-08-06 | 2021-11-09 | 山东捷瑞数字科技股份有限公司 | Residual material analysis processing method and device and material conveying system |
| CN117036359A (en) * | 2023-10-10 | 2023-11-10 | 成都中轨轨道设备有限公司 | Contact net geometric parameter measurement method based on binocular machine vision |
| CN117036359B (en) * | 2023-10-10 | 2023-12-08 | 成都中轨轨道设备有限公司 | Contact net geometric parameter measurement method based on binocular machine vision |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104318546B (en) | 2017-04-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104318546A (en) | Multi-scale analysis-based greenhouse field plant leaf margin extraction method and system | |
| CN109146948B (en) | Crop growth phenotype parameter quantification and yield correlation analysis method based on vision | |
| CN108665487B (en) | Transformer substation operation object and target positioning method based on infrared and visible light fusion | |
| Schiewe | Segmentation of high-resolution remotely sensed data-concepts, applications and problems | |
| Khoshelham et al. | Performance evaluation of automated approaches to building detection in multi-source aerial data | |
| CN103310218B (en) | A kind of overlap blocks fruit precise recognition method | |
| CA2858166C (en) | Method and system for characterising plant phenotype | |
| Canaz Sevgen et al. | An improved RANSAC algorithm for extracting roof planes from airborne lidar data | |
| Moriondo et al. | Use of digital images to disclose canopy architecture in olive tree | |
| CN103279762B (en) | Common growth form of fruit decision method under a kind of physical environment | |
| CN103295018A (en) | Method for precisely recognizing fruits covered by branches and leaves | |
| Matikainen et al. | Multispectral airborne laser scanning for automated map updating | |
| Grigillo et al. | Automated building extraction from IKONOS images in suburban areas | |
| CN115690081A (en) | Tree counting method, system, storage medium, computer equipment and terminal | |
| Meyer et al. | CherryPicker: Semantic skeletonization and topological reconstruction of cherry trees | |
| CN110309808A (en) | A kind of adaptive smog root node detection method under a wide range of scale space | |
| CN113160210A (en) | Drainage pipeline defect detection method and device based on depth camera | |
| Zhao et al. | Automatic sweet pepper detection based on point cloud images using subtractive clustering | |
| Carlinet et al. | Morphological object picking based on the color tree of shapes | |
| Menaka et al. | Change detection in deforestation using high resolution satellite image with Haar wavelet transforms | |
| Das et al. | SAR image segmentation for land cover change detection | |
| CN111768101B (en) | Remote sensing cultivated land change detection method and system taking account of physical characteristics | |
| Coca et al. | Normalized compression distance for SAR image change detection | |
| CN113723833A (en) | Method and system for evaluating afforestation actual performance quality, terminal equipment and storage medium | |
| Bouchahma et al. | Optical-flow-based approach for the detection of shoreline changes using remote sensing data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |