CN109829879A - The detection method and device of vascular bundle - Google Patents

The detection method and device of vascular bundle Download PDF

Info

Publication number
CN109829879A
CN109829879A CN201811474473.XA CN201811474473A CN109829879A CN 109829879 A CN109829879 A CN 109829879A CN 201811474473 A CN201811474473 A CN 201811474473A CN 109829879 A CN109829879 A CN 109829879A
Authority
CN
China
Prior art keywords
vascular bundle
picture
high definition
scanned picture
training
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
Application number
CN201811474473.XA
Other languages
Chinese (zh)
Other versions
CN109829879B (en
Inventor
王汉坤
余雁
易武坤
黎静
石俊利
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Letterman Electronics Technology Co Ltd
International Center for Bamboo and Rattan
Original Assignee
Beijing Letterman Electronics Technology Co Ltd
International Center for Bamboo and Rattan
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Letterman Electronics Technology Co Ltd, International Center for Bamboo and Rattan filed Critical Beijing Letterman Electronics Technology Co Ltd
Priority to CN201811474473.XA priority Critical patent/CN109829879B/en
Publication of CN109829879A publication Critical patent/CN109829879A/en
Application granted granted Critical
Publication of CN109829879B publication Critical patent/CN109829879B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Image Analysis (AREA)

Abstract

The embodiment of the present invention provides a kind of detection method and device of vascular bundle, wherein method includes: the high definition scanned picture for obtaining plant cross section, the vascular bundle in the high definition scanned picture is detected by YOLOv3 model trained in advance, the coordinate information of each vascular bundle is obtained and counts the quantity of vascular bundle;Wherein, the YOLOv3 model is formed according to the label training of each vascular bundle in sample high definition scanned picture and sample high definition scanned picture.The processing flow of material is not only greatly saved in the embodiment of the present invention, also achieves the integrally-built comprehensive analysis of plant haulm.

Description

The detection method and device of vascular bundle
Technical field
The present embodiments relate to plnat monitoring technical fields, more particularly, to the detection method and device of vascular bundle.
Background technique
The quantity and tissue proportion of vascular bundle in the vascular plants stalk such as bamboo, rattan, palm are for studying vascular plant stem Stalk institutional framework and mechanical performance are most important, but there is no related patents, paper or method at present to be counted automatically and group Knit proportion calculating.
When counting to vascular bundle, previous research, which relies on, manually to be counted.Counts amount is very huge, By taking moso bamboo as an example, more than a 7000 root vascular bundle of bamboo ring, artificial counting is difficult to calculate clear.
Summary of the invention
The embodiment of the present invention provides a kind of vascular bundle for overcoming the above problem or at least being partially solved the above problem Detection method and device.
First aspect, the embodiment of the present invention provide a kind of detection method of vascular bundle, comprising:
The high definition scanned picture for obtaining plant cross section, by YOLOv3 model trained in advance to the high definition scanning figure Vascular bundle in piece is detected, and is obtained the coordinate information of each vascular bundle and is counted the quantity of vascular bundle, wherein is described YOLOv3 model is formed according to the label training of each vascular bundle in sample high definition scanned picture and sample high definition scanned picture.
The second aspect, the embodiment of the present invention provide a kind of detection device of vascular bundle, comprising:
Picture obtains module, for obtaining the high definition scanned picture in plant cross section;
Vascular bundle information detecting module, by YOLOv3 model trained in advance to the dimension pipe in the high definition scanned picture Shu Jinhang detection, obtain the coordinate information of each vascular bundle and count the quantity of vascular bundle, wherein the YOLOv3 model according to The label training of each vascular bundle forms in sample high definition scanned picture and sample high definition scanned picture.
The third aspect, the embodiment of the present invention provides a kind of electronic equipment, including memory, processor and is stored in memory Computer program that is upper and can running on a processor, is realized when the processor executes described program as first aspect provides Method the step of.
Fourth aspect, the embodiment of the present invention provide a kind of non-transient computer readable storage medium, are stored thereon with calculating Machine program is realized as provided by first aspect when the computer program is executed by processor the step of method.
The detection method and device of vascular bundle provided in an embodiment of the present invention obtain the high definition scanning figure in plant cross section Piece detects the vascular bundle in the high definition scanned picture by YOLOv3 model trained in advance, obtains each dimension pipe The coordinate information of beam and the quantity for counting vascular bundle, are not only greatly saved the processing flow of material, also achieve plant haulm Integrally-built comprehensive analysis.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the flow diagram of the detection method of vascular bundle provided in an embodiment of the present invention;
Fig. 2 is the structural schematic diagram of the detection device of vascular bundle provided in an embodiment of the present invention;
Fig. 3 is the entity structure schematic diagram of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
In order to overcome the above problem of the prior art, the embodiment of the invention provides a kind of detection method of vascular bundle, Inventive concept are as follows: plant vasular Shu Jinhang is detected by machine learning algorithm and image processing techniques, is counted, is solved artificial Mode carries out the problem that time-consuming, effort, precision are low in plant vasular beam quantity statistics and tissue proportion calculating process.
Fig. 1 is the flow diagram of the detection method of vascular bundle provided in an embodiment of the present invention, as shown in Figure 1, comprising:
S101, the high definition scanned picture for obtaining plant cross section.
The embodiment of the present invention is directed to target plant --- the vascular plant such as bamboo, rattan, palm, after intercepting detection position, It is polished after gas is dry using fine sandpaper, after obtaining the burnishing surface of high quality, is scanned using high definition scanner, obtain high definition Scanned picture.
Specifically, for vascular plants such as bamboo, rattan, palms, test position is intercepted, sample should be noted that selection knot when intercepting Structure is complete, the sample without manufacturing deficiency, and it is dry to be placed in gas under room temperature environment, after with sander carry out surface polishing treatment, sand paper Mesh number should be greater than 200 mesh, with guarantee polish precision, guarantee fibrous sheath and parenchyma cell boundary it is clear and legible, sample processing It is placed on well under Stereo microscope and checks, satisfactory sample surfaces carry out high definition scanning, and scanning of the embodiment of the present invention uses Equipment be high definition scanner, be scanned under grayscale mode, scanning resolution should be greater than 4800dpi.
It should be noted that the embodiment of the present invention selects fine sand paper polishing, fibrous sheath and thin not only can be accurately told The line of demarcation of parietal cell, and the macropores such as big conduit, bast in vascular bundle in sample can be filled, it avoids and leads Interference of the pore to fibrous sheath areal calculation.High definition scanner can quickly obtain the complete cross section picture of plant haulm, Not only convenient and efficient, gained information is also more comprehensively accurate.
S102, the vascular bundle in the high definition scanned picture is detected by YOLOv3 model trained in advance, is obtained It obtains the coordinate information of each vascular bundle and counts the quantity of vascular bundle.
It should be appreciated that YOLO model is the Ross Girshick needle after RCNN, fast-RCNN and faster-RCNN Another frame that DL target detection speed issue is proposed.The core concept of YOLO is exactly using whole figure as network Input, the directly position in output layer recurrence bounding box (bounding box) and its affiliated classification.The embodiment of the present invention YOLOv3 model is formed according to the label training of each vascular bundle in sample high definition scanned picture and sample high definition scanned picture, In, input of the sample high definition scanned picture as network, the label of each vascular bundle is as network in sample high definition scanned picture Output.YOLO model has tri- kinds of versions of v1, v2 and v3 at present, and the embodiment of the present invention uses v3 version, and the version is in Pascal 608x608 speed image is handled on Titan X reaches 20FPS, mAP@0.5 reaches 57.9% on COCO test-dev, with The result of RetinaNet is close, and fast 4 times of speed.
It should be noted that the detection method of the vascular bundle of the embodiment of the present invention passes through at machine learning algorithm and image Reason technology is detected and is counted to plant vasular Shu Jinhang, is solved manual type and is carried out plant vasular beam quantity statistics and tissue ratio The low problem of time-consuming, effort, precision during amount calculates.The embodiment of the present invention use for the first time machine learning algorithm to vascular bundle into Row identification and calculating, can not only accurately identify vascular bundle, obtain the quantity of vascular bundle, can also obtain every dimension pipe The accurate coordinate of beam provides important evidence for the distribution characteristics of vascular bundle in research stalk and the research and development of biomimetic material.
To vascular bundle area and tissue proportion calculate when, generally use the pictures such as Image-ProPlus processing Software carries out binary conversion treatment to picture obtained by Stereo microscope, carries out automatic frame favored area using software afterwards or manually delineate Region is calculated.In terms of areal calculation, because conventional software relies solely on artificial progress binary conversion treatment, boundary determines artificial Influence factor is too many, and difference is larger when different people is delineated, binaryzation and areal calculation inaccuracy, it is difficult to reach requirement, and Stereo microscope visual field very little is only capable of obtaining the small local picture of stalk, and the plant structure information of reflection is not comprehensive.
It is described to obtain often as a kind of alternative embodiment on the basis of the various embodiments described above in order to overcome the above problem The coordinate information of a vascular bundle and the quantity for counting vascular bundle, later further include:
Vascular bundle is grabbed from the high definition scanned picture according to the coordinate information of each vascular bundle and is carried out from binaryzation Vascular bundle after binary conversion treatment is restored to according to corresponding coordinate information the sky with the high definition scanned picture with size by reason In white picture, binaryzation picture is obtained.
The area that fibrous sheath is calculated according to the binaryzation picture, according in gliding smoothing integration method combination binaryzation picture The coordinate information of vascular bundle calculates the area in plant cross section, according to the face according to fibrous sheath area and plant cross section Long-pending ratio calculation tissue proportion.
It should be noted that in embodiments of the present invention, each vascular bundle is marked by a unique rectangle frame, Therefore, the coordinate of vascular bundle can also be indicated by corresponding rectangle frame, specifically, can be with the upper left corner of rectangle frame Coordinate of the coordinate of element as vascular bundle, in addition it is also possible to using the coordinate of the element at the multiple angles of rectangle frame as vascular bundle Coordinate.2) such as in the picture of a 1000*1000, the left upper apex coordinate of vascular bundle is (100,200), and bottom right vertex is sat (300,400) are designated as, then the pixel value of the vascular bundle can easily be grabbed by following Python code: wgs=img [200: 400,100:300,:]。
Since each vascular bundle all has unique coordinate, the embodiment of the present invention, which is realized, to be cleaned according to coordinate from height It traces designs and grabs the purpose of vascular bundle in piece, and in the prior art due to not accurately identifying the coordinate of vascular bundle, it can not be real The effect that now any vascular bundle is grabbed.
It is understood that vascular bundle color in scanned picture is deeper, and cell liquid, parenchyma cell are in scanned picture Color is shallower, therefore by capableing of the boundary of automatic identification fibrous sheath and parenchyma cell to vascular bundle progress binary conversion treatment, complete At the automatic calculating of fibrous sheath area, data are accurate and reliable.It is crosscutting that plant can be easily obtained by gliding smoothing integration method The area in face.
On the basis of the various embodiments described above, as a kind of alternative embodiment, the training method of the YOLOv3 model are as follows:
It is several subgraphs by sample high definition scanned picture cutting, vascular bundle complete in each subgraph is carried out with rectangle frame It marks and coordinate is generated according to rectangle frame, the training set as YOLOv3 model.
Specifically, high definition picture cutting obtained by scanner is several subgraph by the embodiment of the present invention, is passed through The coordinate of vascular bundle complete in sub-pictures is labeled (confines) respectively by LabelImg software, and obtaining one includes institute There is the XML file of the coordinate of vascular bundle.Later by subgraph and its comprehensive data as YOLOv3 model of corresponding XML file Collection.Total sample number amount should be greater than 10000.As a kind of optional embodiment, the embodiment of the present invention uses Keras version YOLO V3 Open Source Code is trained model.
KMeans cluster is carried out as sample characteristics to the height and width of each rectangle frame in the training set, obtains 9 clusters Center, each cluster centre are used to characterize the depth-width ratio of a rectangle frame.
It should be noted that in order to improve the precision of model, according to the requirement of YOLO v3, first with all marks of data set The height of frame is infused, it is wide as sample characteristics progress KMeans cluster (k=9), 9 cluster centres, i.e. 9 pairs of height/width values are obtained, and match It sets in config.json file, the reference of the size of alternative frame is generated as YOLOV3 model.
Data enhancing processing is carried out to the sample characteristics in training set in batches, the data enhancing processing includes random scaling And flip horizontal.
It should be noted that in order to increase the variability of sample and randomness, to improve model generalization ability and prevent Over-fitting, the embodiment of the present invention carry out data enhancing processing to sample characteristics, wherein random scaling i.e. to the height of rectangle frame and For width into row stochastic scaling, flip horizontal is then to rotate rectangle frame.
By data enhancing, treated that sample characteristics are input to pre-sets the YOLOv3 model of initiation parameter and carry out Training retains the weight of YOLOv3 model after the completion of each epoch training, and and epoch iteration stable in penalty values is not Training is terminated when declining again;
It should be noted that biggish batch_size is conducive to find the faster gradient side of LOSS decline in iterative process To the convergence of acceleration model.Suitable batch_size should be based on device configuration, and Ying Jinhang repetition test determines, otherwise will Memory is caused to overflow and can not train.For acceleration model training, and LOSS value as small as possible is obtained, is learned using dynamic adjustment The algorithm of habit rate.When LOSS value is when 2 epoch training can not still improve, learning rate will be reduced by one automatically The order of magnitude.
The embodiment of the present invention in the training process, complete the parameter of preservation model, and exports mould by each epoch training The key parameters such as penalty values, learning rate, the time-consuming of type, finally when penalty values are more stable, multiple epoch iteration is not continued to down When drop, then training can be terminated.
After model training terminates, verifying collection is tested with trained model, and calculate the mAP on verifying collection (mean accuracy), recall50 (recall rate when iou threshold value is 50%), recall75 (recalling when iou threshold value is 75% Rate) etc. indexs carry out the performance of evaluation model.
On the basis of the various embodiments described above, by YOLOv3 model trained in advance in the high definition scanned picture Vascular bundle is detected, and is obtained the coordinate information of each vascular bundle and is counted the quantity of vascular bundle, specifically:
By the high definition scanned picture according to offset cutting method cutting be 4 sets of subgraphs, the size and training of every set subgraph The cutting subgraph of collection is in the same size;
The test parameter of YOLOv3 model trained in advance is set, 4 sets of subgraphs are separately input into the YOLOv3 later In model, the coordinate information of each alternative frame in 4 sets of subgraphs is obtained.
Specifically, test parameter includes the size (high, wide) of network inputs, which must be 32 multiple, smaller, in advance Degree of testing the speed is faster, but precision may be impacted.Obj_thresh is the threshold value for examining alternative frame to whether there is object, nms_ Thresh, that is, non-maxima suppression threshold value, for defining duplicate alternative frame, to select the alternative frame of highest scoring.Wherein, Alternative frame is the prediction block that YOLOV3 model generates, for predict object position and affiliated classification.
Actual test sample is detected by trained model, model can detect well in addition to cutting line All vascular bundles except upper several incomplete vascular bundles.
In order to improve the accuracy of testing result, on the basis of the various embodiments described above, output token each completely manage by dimension The subgraph of the rectangle frame (i.e. alternative frame) of beam, later further include:
The alternative frame of 4 sets of subgraphs is filtered according to preset area threshold and depth-width ratio threshold value, after being filtered Alternative frame;For example, giving up for the alternative frame less than 4000 squares, since the shape of vascular bundle is closer square Shape, therefore the alternative frame that depth-width ratio is greater than 4:1 can be deleted.
The filtered alternative frame is merged, duplicate alternative frame is removed, records remaining standby in all subgraphs The coordinate information and quantity for selecting frame, respectively as the coordinate and quantity of vascular bundle.
It is predicted it should be noted that original high definition scanned picture is split for multiple subgraphs, the dimension on cutting line Due to imperfect, possibility is unpredictable or is only capable of one incomplete vascular bundle of prediction for tube bank.This problem is solved, can be adopted With offset 4 sets of subgraphs of cutting, then the prediction result of more set subgraphs is merged to the algorithm of duplicate removal, further boosting algorithm Overall precision.
On the basis of the various embodiments described above, the area of fibrous sheath is calculated according to the binaryzation picture, specifically:
It counts for characterizing the number of the pixel of fibrous sheath in the binaryzation picture, and represented multiplied by each pixel Real area, obtain the area of the fibrous sheath.
Since the calculating of fibrous sheath area needs to rely on binary picture, traditional Binarization methods are largely all based on The algorithm of threshold value (fixed threshold or adaptive threshold), it is easy to by picture integrated environment, background, coloration, color difference, noise Deng influence, precision is not high.And the embodiment of the present invention grabs each vascular bundle according to the coordinate information of vascular bundle from original image one by one It takes out, then all pixels value flattening of the vascular bundle is regard as feature, using Kmean clustering algorithm, by all pixels point Gather for 2 classes (black and white), then calculates separately the pixel value mean value of every one kind, the big conduct white of mean value, the small conduct of mean value Then black is then reduced into binary picture, so that the interference of exclusion integrated environment as much as possible, is clearly distinguished as far as possible The binary picture of all vascular bundles is reverted to the blank of an original image size according to coordinate information by vascular bundle and background In picture, whole binary picture is generated, based on whole binary picture, by counting the number of black pixel point, and is multiplied With the real area that each pixel represents, the area of fibrous sheath is calculated.
It is described according to vascular bundle in gliding smoothing integration method combination binaryzation picture on the basis of the various embodiments described above Coordinate information calculates the area in plant cross section, specifically:
Sliding window is set in the binary picture on piece, calculates in each sliding window the extreme lower position of fibrous sheath and most High position;
Horizontally and vertically scalable formation rectangle is carried out to the extreme lower position and extreme higher position according to preset threshold, as institute State a part in sliding window implants cross section;Wherein, the preset threshold according to the artificial results of measuring of all kinds of plants into Row determines;
By the rectangle binaryzation fill to in the blank picture of the same size of original image, until it is crosscutting to obtain entire plant The binary picture of EDS maps;
By calculating the pixel number in the characterization plant cross section in binary picture, and represented multiplied by each pixel Real area obtains the area in the plant cross section.
Specifically, it since vascular bundle is almost distributed in entire plant, can be distributed by the coordinate of all vascular bundles, Using gliding smoothing integration method approximate calculation sample cross-sectional area.Algorithm is as follows:
A) width=[width] px (such as 10~50 pixels) is taken, the slender type rectangle of height=picture altitude, from It is left-to-right, using [step] px as step-length, the binary picture for the vascular bundle entirety that smooth scan obtains.
B) mobile [step] px of each sliding sash, is intended to calculate minimum (min_height) and highest of fibrous sheath in the region Position (max_height), and horizontal and vertical expend_w_thresh, expend_h_thresh px (threshold value are extended respectively Can compare, be adjusted according to the artificial results of measuring of every class plant), an elongated rectangle is obtained, as the sliding sash Interior, example cross section part, binaryzation is filled into a blank picture with the same size of original image;
C) step b is repeated until sliding sash is moved to image right end, to obtain the two-value of entire example cross section distribution Change figure, is based on this figure, it, can be close by calculating the number of black pixel point, and the real area represented multiplied by each pixel Like the area for calculating example cross section.
Fig. 2 is the structural schematic diagram of the detection device of vascular bundle provided in an embodiment of the present invention, as shown in Fig. 2, the dimension pipe The detection device of beam includes: that picture obtains module 201 and vascular bundle information detecting module 202, in which:
Picture obtains the high definition scanned picture that module 201 is used to obtain plant cross section.
Specifically, the embodiment of the present invention is directed to target plant --- the vascular plant such as bamboo, rattan, palm, interception detection Behind position, is polished after gas is dry using fine sandpaper, after obtaining the burnishing surface of high quality, is scanned using high definition scanner, Obtain high definition scanned picture.
Vascular bundle information detecting module 202, by YOLOv3 model trained in advance in the high definition scanned picture Vascular bundle is detected, and is obtained the coordinate information of each vascular bundle and is counted the quantity of vascular bundle.
It should be appreciated that YOLO model is the Ross Girshick needle after RCNN, fast-RCNN and faster-RCNN Another frame that DL target detection speed issue is proposed.The core concept of YOLO is exactly using whole figure as network Input, the directly position in output layer recurrence bounding box (bounding box) and its affiliated classification.The embodiment of the present invention YOLOv3 model is formed according to the label training of each vascular bundle in sample high definition scanned picture and sample high definition scanned picture, In, input of the sample high definition scanned picture as network, the label of each vascular bundle is as network in sample high definition scanned picture Output.YOLO model has tri- kinds of versions of v1, v2 and v3 at present, and the embodiment of the present invention uses v3 version, and the version is in Pascal 608x608 speed image is handled on Titan X reaches 20FPS, mAP@0.5 reaches 57.9% on COCO test-dev, with The result of RetinaNet is close, and fast 4 times of speed.
It should be noted that the detection method of the vascular bundle of the embodiment of the present invention passes through at machine learning algorithm and image Reason technology is detected and is counted to plant vasular Shu Jinhang, is solved manual type and is carried out plant vasular beam quantity statistics and tissue ratio The low problem of time-consuming, effort, precision during amount calculates.The embodiment of the present invention use for the first time machine learning algorithm to vascular bundle into Row identification and calculating, can not only accurately identify vascular bundle, obtain the quantity of vascular bundle, can also obtain every dimension pipe The accurate coordinate of beam provides important evidence for branch's feature of vascular bundle in research stalk and the research and development of biomimetic material.
The detection device of vascular bundle provided in an embodiment of the present invention, the detection method for specifically executing above-mentioned each vascular bundle are implemented Example process, please specifically be detailed in the content of the detection method embodiment of above-mentioned each vascular bundle, details are not described herein.The embodiment of the present invention The processing flow of material is not only greatly saved in the detection device of the vascular bundle of offer, and it is integrally-built to also achieve plant haulm Analysis comprehensively;The work such as detection, counting, the area measurement of vascular bundle can be completed in a few minutes, not only greatly improve Efficiency has been saved the time, and the precision of resolution and calculating is also substantially increased.
Fig. 3 is the entity structure schematic diagram of electronic equipment provided in an embodiment of the present invention, as shown in figure 3, the electronic equipment It may include: processor (processor) 310,320, memory communication interface (Communications Interface) (memory) 330 and communication bus 340, wherein processor 310, communication interface 320, memory 330 pass through communication bus 340 Complete mutual communication.Processor 310 can call the meter that is stored on memory 330 and can run on processor 310 Calculation machine program, to execute the detection method of the various embodiments described above offer, for example, obtain the high definition scanning figure in plant cross section Piece detects the vascular bundle in the high definition scanned picture by YOLOv3 model trained in advance, obtains each dimension pipe The coordinate information of beam and the quantity for counting vascular bundle;Wherein, the YOLOv3 model is according to sample high definition scanned picture and sample The label training of each vascular bundle forms in this high definition scanned picture.
In addition, the logical order in above-mentioned memory 330 can be realized by way of SFU software functional unit and conduct Independent product when selling or using, can store in a computer readable storage medium.Based on this understanding, originally The technical solution of the inventive embodiments substantially part of the part that contributes to existing technology or the technical solution in other words It can be embodied in the form of software products, which is stored in a storage medium, including several fingers It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes the present invention respectively The all or part of the steps of a embodiment the method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic or disk Etc. the various media that can store program code.
The embodiment of the present invention also provides a kind of non-transient computer readable storage medium, is stored thereon with computer program, The computer program is implemented to carry out the detection method of the various embodiments described above offer when being executed by processor, for example, obtain The high definition scanned picture in plant cross section, by YOLOv3 model trained in advance to the vascular bundle in the high definition scanned picture It is detected, obtain the coordinate information of each vascular bundle and counts the quantity of vascular bundle;Wherein, the YOLOv3 model is according to sample The label training of each vascular bundle forms in this high definition scanned picture and sample high definition scanned picture.
The apparatus embodiments described above are merely exemplary, wherein described, unit can as illustrated by the separation member It is physically separated with being or may not be, component shown as a unit may or may not be physics list Member, it can it is in one place, or may be distributed over multiple network units.It can be selected according to the actual needs In some or all of the modules achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creativeness Labour in the case where, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation Method described in certain parts of example or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features; And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and Range.

Claims (10)

1. a kind of detection method of vascular bundle characterized by comprising
The high definition scanned picture for obtaining plant cross section, by YOLOv3 model trained in advance in the high definition scanned picture Vascular bundle detected, obtain the coordinate information of each vascular bundle and count the quantity of vascular bundle;
Wherein, the YOLOv3 model is according to the mark of each vascular bundle in sample high definition scanned picture and sample high definition scanned picture Note training forms.
2. the detection method of vascular bundle according to claim 1, which is characterized in that the coordinate for obtaining each vascular bundle Information and the quantity for counting vascular bundle, later further include:
Vascular bundle is grabbed from the high definition scanned picture according to the coordinate information of each vascular bundle and carries out binary conversion treatment, it will Vascular bundle after binary conversion treatment is restored to the blank sheet with the high definition scanned picture with size according to corresponding coordinate information In piece, binaryzation picture is obtained;
The area that fibrous sheath is calculated according to the binaryzation picture is managed according to tieing up in gliding smoothing integration method combination binaryzation picture The coordinate information of beam calculates the area in plant cross section, according to described according to fibrous sheath area and the area in plant cross section Ratio calculation tissue proportion.
3. detection method according to claim 1, which is characterized in that the training method of the YOLOv3 model are as follows:
It is several subgraphs by sample high definition scanned picture cutting, vascular bundle complete in each subgraph is marked with rectangle frame And coordinate is generated according to rectangle frame, the training set as YOLOv3 model;
KMeans cluster is carried out as sample characteristics to the height and width of each rectangle frame in the training set, obtains 9 cluster centres, Each cluster centre is used to characterize the depth-width ratio of a rectangle frame;
Data enhancing processing is carried out to the sample characteristics in training set in batches, the data enhancing processing includes random scaling and water Flat overturning;
By data enhancing, treated that sample characteristics are input to pre-sets the YOLOv3 model of initiation parameter and be trained, Retain the weight of YOLOv3 model after the completion of each epoch training, and and epoch iteration stable in penalty values no longer declines When terminate training;
Wherein, the parameter include crowd size batch_size, learning rate learning_rate, frequency of training nb_epochs, The detection threshold value ignore_thresh and GPU quantity GPUs for participating in training.
4. detection method according to claim 1, which is characterized in that it is described by YOLOv3 model trained in advance to institute The vascular bundle stated in high definition scanned picture is detected, and the coordinate information of each vascular bundle is obtained, specifically:
It according to offset cutting method cutting is 4 sets of subgraphs by the high definition scanned picture, the size of every set subgraph and training set Cutting subgraph is in the same size;
The test parameter of YOLOv3 model trained in advance is set, 4 sets of subgraphs are separately input into the YOLOv3 model later In, obtain the coordinate information of each alternative frame in 4 sets of subgraphs.
5. detection method according to claim 4, which is characterized in that detected to 4 sets of subgraphs difference alternative Frame coordinate information, later further include:
The alternative frame of 4 sets of subgraphs is filtered according to preset area threshold and depth-width ratio threshold value, is obtained filtered standby Select frame;
The filtered alternative frame is merged, duplicate alternative frame is removed, records remaining alternative frame in all subgraphs Coordinate and quantity, respectively as the coordinate and quantity of vascular bundle.
6. detection method according to claim 2, which is characterized in that described to calculate fibrous sheath according to the binaryzation picture Area, specifically:
It counts in the binaryzation picture for characterizing the number of the pixel of fibrous sheath, and the reality represented multiplied by each pixel Interphase product, obtains the area of the fibrous sheath.
7. detection method according to claim 2, which is characterized in that described according to gliding smoothing integration method combination binaryzation The coordinate information of vascular bundle in picture calculates the area in plant cross section, specifically:
Sliding window is set in the binary picture on piece, calculates the extreme lower position and highest order of fibrous sheath in each sliding window It sets;
Horizontally and vertically scalable formation rectangle is carried out to the extreme lower position and extreme higher position according to preset threshold, as the cunning The a part in dynamic window implants cross section;Wherein, the preset threshold carries out true according to the artificial results of measuring of all kinds of plants It is fixed;
The rectangle binaryzation is filled to the blank picture of the same size of original image, is divided until obtaining entire plant cross section The binary picture of cloth;
By calculating the pixel number of the characterization vascular bundle in binary picture, and the practical face represented multiplied by each pixel Product, obtains the area in the plant cross section.
8. a kind of detection device of vascular bundle characterized by comprising
Picture obtains module, for obtaining the high definition scanned picture in plant cross section;
Vascular bundle information detecting module, by YOLOv3 model trained in advance to the vascular bundle in the high definition scanned picture into Row detection, obtains the coordinate information of each vascular bundle and counts the quantity of vascular bundle;
Wherein, the YOLOv3 model is according to the mark of each vascular bundle in sample high definition scanned picture and sample high definition scanned picture Note training forms.
9. a kind of electronic equipment characterized by comprising
At least one processor;And
At least one processor being connect with the processor communication, in which:
The memory is stored with the program instruction that can be executed by the processor, and the processor calls described program to instruct energy Enough execute detection method as claimed in any of claims 1 to 7 in one of claims.
10. a kind of non-transient computer readable storage medium, which is characterized in that the non-transient computer readable storage medium is deposited Computer instruction is stored up, the computer instruction makes the computer execute detection as claimed in any of claims 1 to 7 in one of claims Method.
CN201811474473.XA 2018-12-04 2018-12-04 Method and device for detecting vascular bundle Active CN109829879B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811474473.XA CN109829879B (en) 2018-12-04 2018-12-04 Method and device for detecting vascular bundle

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811474473.XA CN109829879B (en) 2018-12-04 2018-12-04 Method and device for detecting vascular bundle

Publications (2)

Publication Number Publication Date
CN109829879A true CN109829879A (en) 2019-05-31
CN109829879B CN109829879B (en) 2021-01-15

Family

ID=66859835

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811474473.XA Active CN109829879B (en) 2018-12-04 2018-12-04 Method and device for detecting vascular bundle

Country Status (1)

Country Link
CN (1) CN109829879B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111414880A (en) * 2020-03-26 2020-07-14 电子科技大学 Method for detecting target of active component in microscopic image based on improved RetinaNet
CN114757918A (en) * 2022-04-15 2022-07-15 国际竹藤中心 Biomass material detection method and device, electronic equipment and storage medium
CN115965582A (en) * 2022-11-22 2023-04-14 哈尔滨岛田大鹏工业股份有限公司 Ultrahigh-resolution-based engine cylinder body and cylinder cover surface defect detection method

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101238987A (en) * 2007-09-06 2008-08-13 深圳先进技术研究院 Processing method of CT cerebral hemorrhage image
US20090170712A1 (en) * 2007-12-28 2009-07-02 Pioneer Hi-Bred International, Inc. Using Oligonucleotide Microarrays to Analyze Genomic Differences for the Prediction of Heterosis
US20100111396A1 (en) * 2008-11-06 2010-05-06 Los Alamos National Security Object and spatial level quantitative image analysis
CN104182763A (en) * 2014-08-12 2014-12-03 中国计量学院 Plant type identification system based on flower characteristics
CN107886497A (en) * 2017-10-05 2018-04-06 北京林业大学 A kind of wheat flag leaf automatic nondestructive detection and measuring method based on image
CN108527382A (en) * 2018-04-09 2018-09-14 上海方立数码科技有限公司 A kind of crusing robot
CN108596174A (en) * 2018-04-28 2018-09-28 中南大学 A kind of lesion localization method of skin disease image
CN108647652A (en) * 2018-05-14 2018-10-12 北京工业大学 A kind of cotton development stage automatic identifying method based on image classification and target detection

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101238987A (en) * 2007-09-06 2008-08-13 深圳先进技术研究院 Processing method of CT cerebral hemorrhage image
US20090170712A1 (en) * 2007-12-28 2009-07-02 Pioneer Hi-Bred International, Inc. Using Oligonucleotide Microarrays to Analyze Genomic Differences for the Prediction of Heterosis
US20100111396A1 (en) * 2008-11-06 2010-05-06 Los Alamos National Security Object and spatial level quantitative image analysis
CN104182763A (en) * 2014-08-12 2014-12-03 中国计量学院 Plant type identification system based on flower characteristics
CN107886497A (en) * 2017-10-05 2018-04-06 北京林业大学 A kind of wheat flag leaf automatic nondestructive detection and measuring method based on image
CN108527382A (en) * 2018-04-09 2018-09-14 上海方立数码科技有限公司 A kind of crusing robot
CN108596174A (en) * 2018-04-28 2018-09-28 中南大学 A kind of lesion localization method of skin disease image
CN108647652A (en) * 2018-05-14 2018-10-12 北京工业大学 A kind of cotton development stage automatic identifying method based on image classification and target detection

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
JIANJUN DU 等: "Micron-scale phenotyping quantification and three-dimensional microstructure reconstruction of vascular bundles within maize stalks based on micro-CT scanning", 《FUNCTIONAL PLANT BIOLOGY》 *
赵世峰 等: "基于OpenCV的复杂环境下图像二值化方法", 《电子测量技术》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111414880A (en) * 2020-03-26 2020-07-14 电子科技大学 Method for detecting target of active component in microscopic image based on improved RetinaNet
CN111414880B (en) * 2020-03-26 2022-10-14 电子科技大学 Method for detecting target of active component in microscopic image based on improved RetinaNet
CN114757918A (en) * 2022-04-15 2022-07-15 国际竹藤中心 Biomass material detection method and device, electronic equipment and storage medium
CN115965582A (en) * 2022-11-22 2023-04-14 哈尔滨岛田大鹏工业股份有限公司 Ultrahigh-resolution-based engine cylinder body and cylinder cover surface defect detection method
CN115965582B (en) * 2022-11-22 2024-03-08 哈尔滨岛田大鹏工业股份有限公司 Ultrahigh-resolution-based method for detecting surface defects of cylinder body and cylinder cover of engine

Also Published As

Publication number Publication date
CN109829879B (en) 2021-01-15

Similar Documents

Publication Publication Date Title
CN109636772A (en) The defect inspection method on the irregular shape intermetallic composite coating surface based on deep learning
Clausi et al. Design-based texture feature fusion using Gabor filters and co-occurrence probabilities
CN104346617B (en) A kind of cell detection method based on sliding window and depth structure extraction feature
CN113592845A (en) Defect detection method and device for battery coating and storage medium
CN109829879A (en) The detection method and device of vascular bundle
CN101984916B (en) Blood vessel diameter measuring method based on digital image processing technology
CN107564002A (en) Plastic tube detection method of surface flaw, system and computer-readable recording medium
CN108426994A (en) Digital holographic microscopy data are analyzed for hematology application
US11501446B2 (en) Segmenting 3D intracellular structures in microscopy images using an iterative deep learning workflow that incorporates human contributions
CN109509171A (en) A kind of Fabric Defects Inspection detection method based on GMM and image pyramid
CN110400322A (en) Fruit point cloud segmentation method based on color and three-dimensional geometric information
EP3785223A1 (en) Systems and methods for segmentation and analysis of 3d images
CN109584153A (en) Modify the methods, devices and systems of eye
CN112215217B (en) Digital image recognition method and device for simulating doctor to read film
CN109410138A (en) Modify jowled methods, devices and systems
CN107590512A (en) The adaptive approach and system of parameter in a kind of template matches
CN103268492B (en) A kind of corn grain type identification method
CN114299490B (en) Tumor microenvironment heterogeneity evaluation method
CN115170548A (en) Leather defect automatic detection method and device based on unsupervised learning
US11710308B1 (en) Seed germination detection method and apparatus
CN109583496A (en) A kind of network model and method for the classification of display panel large area defect
EP3475915A1 (en) Visual cardiomyocyte analysis
CN105528791B (en) A kind of quality evaluation device and its evaluation method towards touch screen hand-drawing image
CN112837271B (en) Melon germplasm resource character extraction method and system
CN113935906B (en) Strong reflection stripe noise removing method for Fourier domain optical coherence tomography

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant