CN103116747B - Automatically the method and system of corn stem leaf image is identified - Google Patents
Automatically the method and system of corn stem leaf image is identified Download PDFInfo
- Publication number
- CN103116747B CN103116747B CN201310076981.3A CN201310076981A CN103116747B CN 103116747 B CN103116747 B CN 103116747B CN 201310076981 A CN201310076981 A CN 201310076981A CN 103116747 B CN103116747 B CN 103116747B
- Authority
- CN
- China
- Prior art keywords
- image
- point
- gray
- chain sequence
- chain
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Abstract
The invention discloses a kind of corn stem recognition methods, comprise the following steps: (1) gathers RGB color image; (2) described RGB color image is converted into gray level image; (3) described gray level image is carried out pretreatment; (4) by binary conversion treatment, pretreated described gray level image is converted into binary image; (5) described binary image is carried out straight-line detection; (6) calculate the shape facility that described binary image is detected the zones of different of the line segmentation obtained, and the RGB color image collected is identified by the shape facility according to described zones of different, obtain final stem stalk target. The present invention can identify the stem stalk of Semen Maydis in real time, exactly, and automatic fertilizer spreaders can be made to avoid when applying fertilizer operation because the obstruction of stem stalk causes the generation of the problem such as device damage or injury milpa.
Description
Technical field
The present invention relates to technical field of image processing, particularly relate to a kind of corn stem recognition methods.
Background technology
In corn fertilizing operation process, fertilizer apparatus on automatic fertilizer spreaders can collide with " barriers " such as the stem stalks of Semen Maydis, cause device damage or injury crop, it is thus desirable to these " barriers " are carried out in real time, identify accurately and location so that automatic fertilizer spreaders changes mobile route before or stops mobile being likely to collide with " barrier ".
This identification and location technology can be realized by machine vision, adopt camera acquisition image, utilize the equipment such as computer, dsp chip that image is processed. Prior art adopts the aberration R-B image to corn stem carry out hough convert straight-line detection and then realize the identification of stem stalk, although the line detection method based on Hough transform is a kind of algorithm reached its maturity, but amount of calculation is relatively larger, it is unfavorable for application in real time and conversion process lost end points and the length information of line segment.
Summary of the invention
The technical problem to be solved in the present invention is: provide a kind of corn stem recognition methods, it can identify the stem stalk of Semen Maydis in real time, exactly, and automatic fertilizer spreaders can be made to avoid when applying fertilizer operation because the obstruction of stem stalk causes the generation of the problem such as device damage or injury milpa.
For solving the problems referred to above, the invention provides a kind of Semen Maydis recognition methods, comprise the following steps:
(1) RGB color image is gathered;
(2) described RGB color image is converted into gray level image;
(3) described gray level image is carried out pretreatment;
(4) by binary conversion treatment, pretreated described gray level image is converted into binary image;
(5) described binary image is carried out straight-line detection;
(6) calculate the shape facility that described binary image is detected the zones of different of the line segmentation obtained, and the RGB color image collected is identified by the shape facility according to described zones of different, obtain final stem stalk target.
Described pretreatment is that Lifting Wavelet processes, and it can extract gray level image low frequency profile information, suppress high-frequency noise.
When the pretreatment in step (3) is Lifting Wavelet process, step (4) farther includes:
Gray level image after pixel after Lifting Wavelet processes is reduced adopts differential operator to carry out binary conversion treatment, converts gray images into binary image.
Step (5) farther includes:
Utilizing zero-crossing examination to obtain the boundary point of binary image, and then boundary point is converted into by edge following algorithm the symbolic information of chain representation, the chain sequence obtained based on frontier tracing carries out straight-line detection.
The present invention is by after carrying out gray processing, Lifting Wavelet, binary conversion treatment to the RGB color image of corn stem, pass through boundary points detection, utilize the straight-line detection based on chain code and then the stem stalk of Semen Maydis can be identified in real time, exactly, making automatic fertilizer spreaders avoid when applying fertilizer operation because the obstruction of stem stalk causes the generation of the problem such as device damage or injury milpa.
Accompanying drawing explanation
Fig. 1 is the flow chart of corn stem recognition methods described in embodiment of the present invention;
Fig. 2 is the schematic diagram of 8 directional chain-code.
Detailed description of the invention
Below in conjunction with drawings and Examples, the specific embodiment of the present invention is described in further detail. Following example are used for illustrating the present invention, but are not limited to the scope of the present invention.
As it is shown in figure 1, a kind of corn stem recognition methods of the present invention, comprise the following steps:
(1) RGB color image is gathered;
(2) described RGB color image is converted into gray level image;
In this step, the RGB color image collected is carried out gray processing process, coloured image is converted into gray level image. If in coloured image, the color of certain point original is RGB (R, G, B), it is possible to by following several method, be converted into gray scale:
1) floating-point arithmetic: Gray=R*0.3+G*0.59+B*0.11;
2) integer method: Gray=(R*30+G*59+B*11)/100;
3) displacement method: Gray=(R*76+G*151+B*28) > > 8;
4) mean value method: Gray=(R+G+B)/3;
5) green: Gray=G is only taken;
After trying to achieve Gray by any of the above-described kind of method, R, G, B unification Gray in original RGB (R, G, B) is replaced, form new color RGB (Gray, Gray, Gray), replace original RGB (R, G, B) with it and just obtain gray level image.
In image processing process, for utilizing the color characteristic of coloured image to be split from image background by destination object, often coloured image is carried out dimension-reduction treatment, describe the color of image merely with a certain characteristic component, coloured image is converted into gray level image. Meanwhile, by the gray processing of coloured image, also make image data amount significantly reduce, substantially increase image processing speed.In the present invention, owing to the color information of image is unimportant to be fetched culm morphology information, gray level image remains the shape information of stem stalk in original color image preferably.
(3) described gray level image is carried out pretreatment;
Above-mentioned pretreatment is that Lifting Wavelet processes, and the gray level image after converting is carried out Lifting Wavelet process, and described Lifting Wavelet processes can extract gray level image low frequency profile information, suppress high-frequency noise;
Lifting Wavelet is that one wavelet transformation more fast and effectively realizes method, is referred to as Second Generation Wavelet Transformation. Lifting Wavelet does not rely on Fourier transformation, inherits the feature of the multiresolution of first generation small echo, and the coefficient after wavelet transformation is integer, it is not necessary to extra internal memory, and this bit manipulation available carries out computing, it is possible to realize the wavelet transformation of arbitrary image size. By means of factorization wavelet transformation, all wavelet transformations can both realize by Lifting Wavelet pattern.
In this step, Lifting Wavelet processes and includes division, prediction and update, it is considered to signal sj={sj,l|0��l��2j, the low frequency signal that it obtains after one-level wavelet transformation is sj-1With high-frequency signal dj-1, then the process that implements building low-resolution image is:
I. division: will input signal sjBeing divided into odd, even two subsets, the set of the element composition being positioned at even subscript position is designated as evenj-1={sj,2l|0��l��2j-1-1}, the set of the element composition being positioned at strange subscript position is designated as oddj-1={sj,2l+1|0��l��2j-1-1};
II. prediction: on the basis based on initial data dependency, removes prediction or interpolation odd numbered sequences with the predictive value of even order;
III. update: being intended to of renewal looks for a better subset sj-1So that it is keep a certain scalar characteristic Q (x) (as root-mean-square value is constant) of artwork, i.e. Q (sj-1)=Q(sj);
Wherein, for Lifting Wavelet, there is division operator Split, predictive operator P and update operator U so that
(evenj-1, oddj-1)=Split(sj),
dj-1=oddj-1-P(evenj-1),
sj-1=evenj-1+U(dj-1)��
One width pixel is the image of M �� N, after a Lifting Wavelet, takes its low frequency component and builds low-resolution image, and its pixel isAdditionally, due to noise is distributed in the HFS of image mostly, therefore restrained effectively high-frequency noise based on the low-resolution image constructed by Lifting Wavelet.
(4) by binary conversion treatment, pretreated described gray level image is converted into binary image;
When the pretreatment in step (3) is Lifting Wavelet process, this step farther includes: the gray level image after the pixel after Lifting Wavelet processes is reduced adopts differential operator to carry out binary conversion treatment, converts gray images into binary image;
(5) described binary image is carried out straight-line detection;
This step farther includes:
Utilizing zero-crossing examination to obtain the boundary point of binary image, and then boundary point is converted into by edge following algorithm the symbolic information of chain representation, the chain sequence obtained based on frontier tracing carries out straight-line detection;
In this step, edge following algorithm mainly includes scanning process and tracking process, hypothetical boundary is present between pixel, by limit, unit is constituted, it is divided into horizontal and vertical two kinds, if currently investigating point coordinates is (x, y), judge whether that horizontal sides unit is with (x+1, y) for reference point, judge whether that vertical edges unit is with (x, y+1) for reference point, when investigating point and being identical with the value of reference point, border is not had to pass through between the two point, the value of limit unit is zero, it it is all impact point or background dot is divided into positive zero-sum negative zero two kinds according to them, border is had to pass through when the value investigating point and reference point is different between 2, the value non-zero of limit unit, it is impact point or background dot takes plus or minus respectively according to investigating point,
Tracking direction provides as follows: impact point is all the time on the left side of direction of advance, external boundary for a connected region, tracking direction is counterclockwise, inner boundary (border that inner void is formed) is then for clockwise, when quadraturing on closing chain code basis, external boundary chain code area defined area is just, and the defined region of inner boundary chain code and amass as negative;
The value of limit unit also show corresponding tracking direction simultaneously, and certain limit unit direction is just, for horizontal sides unit, to the right, for vertical edges unit, tracking direction is upwards for tracking direction;
Whether labelling in tracking process is beaten on the impact point when front unit, to distinguish this boundary point tracked mistake;
When there is non-zero limit unit between current investigation point and reference point, according to all the other situations of 2 in these 22 �� 2 windows constituted, determining next step tracking direction, i.e. follow-up limit unit, when return to origin, tracking process terminates.
Wherein, scanning process adopts the mode of network scanning to find the boundary point not yet followed the tracks of, and chain code adopts 8 directional chain-code. Above-mentioned frontier tracing core algorithm drastically increases efficiency of algorithm.
Image is done network scanning, and the region only intersected with grid lines its is possible to tracked, and those regions entirely fallen within grid are then left in the basket. Therefore, network scanning is adopted can also to be greatly enhanced processing speed.
The chain sequence obtained based on frontier tracing carries out in straight-line detection process, adopts following form to represent the chain sequence of object boundary:
I:Len, X, Y, d1,d2,��,dn
Wherein I is the numbering of current chain code string, and Len is the total length of chain code in current chain code string, X, and Y is the image coordinate that current chain code strings initial point, d1,d2,��,dnFor the direction code of pixel each on current chain code string, then the step of straight-line detection is:
IV. minimum length of straigh line checking: order judges each bar chain sequence, if the length of chain sequence is less than default minimum length of straigh line threshold value LT, then gives up this chain sequence;
V. from chain sequence, straightway is extracted according to straightway degree of approximation criterion:
Order judges each bar chain sequence;
1. from the off, select to meet the subchain sequence of minimum length of straigh line constraint successively, the straightway degree of approximation S of this section is calculated according to formula (1), if S >=ST, ST is default minimum straightway similarity threshold, then this section meets line constraint, turns 2., otherwise give up this section, continue to judge next cross-talk chain sequence;
Wherein, | | ps-pe| | represent end points psWith peBetween ideal line distance, Len (ps,pe) represent that subchain sequence is from end points psTo end points pePhysical length through pixel, S then represents the straightway degree of approximation, additionally, when calculating the physical length of chain sequence, if direction code is 0,2,4,6, physical length between two pixels that then chain code connects is 1, if direction code is 1,3,5,7, then the physical length between two pixels is that 2(is referring to Fig. 2);
If 2. subchain the last period sequence also meets line constraint, then consider that can two adjacent strip chain sequence be merged into a longer straightway, by the last period chain sequence starting point to this section chain sequence terminal between this section of chain code regard a new chain sequence as, calculate its straightway degree of approximation S '. If S ' >=ST, then merge, otherwise this cross-talk chain sequence is marked as a new straightway;
If 3. oneself arrives chain sequence terminal, then terminate, otherwise turn and 1. continue to judge next cross-talk chain sequence.
Wherein, described default minimum length of straigh line threshold value LT is 23 pixels, and default minimum straightway similarity threshold ST is 0.91.
Object boundary is first carried out chain code following by the present invention, then carries out line segment extraction in the chain code set of strings obtained, and its advantage is that amount of calculation is little, and can obtain the information such as the position of straightway, length, direction simultaneously. The line detection algorithm that the present invention adopts, from practical application, has broken away from the constraint of Freeman ideal line criterion, only with minimum length of straigh line and two constrained parameters of minimum line similar degree, it is possible to realize straight-line detection rapidly and accurately.This algorithm detection speed is fast, practical, enhances the real-time of corn stem identification process.
(6) calculate the shape facility that described binary image is detected the zones of different of the line segmentation obtained, and the RGB color image collected is identified by the shape facility according to described zones of different, obtain final stem stalk target.
In this step, described shape facility has multiple parameter, and the present embodiment only calculates form parameter and eccentricity. Straight line region is identified as stem stalk, utilizes form parameter to by other region of line segmentationAnd eccentricityIt is identified, meet the step that region recognition is stem stalk that F is more than 2.3 and e is more than 5, wherein, the area in Area and Per respectively region and girth, the border long axis length in c and a respectively region and minor axis length, major axis refers to 2 lines that on zone boundary, distance is farthest, and in line vertical with major axis on zone boundary, the longest line segment is called short axle.
Embodiment of above is merely to illustrate the present invention; and it is not limitation of the present invention; those of ordinary skill about 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 fall within scope of the invention, and the scope of patent protection of the present invention should be defined by the claims.
Claims (1)
1. the method for an automatic identification corn stem, it is characterised in that comprise the following steps:
(1) RGB color image of corn stem is gathered;
(2) described RGB color image is converted into gray level image;
If in coloured image, the color of certain point original is RGB (R, G, B), it is possible to be converted into gray scale by the following method:
1) floating-point arithmetic: Gray=R*0.3+G*0.59+B*0.11;
2) integer method: Gray=(R*30+G*59+B*11)/100;
3) displacement method: Gray=(R*76+G*151+B*28) > > 8;
4) mean value method: Gray=(R+G+B)/3;
5) green: Gray=G is only taken;
After trying to achieve Gray by any of the above-described kind of method, R, G, B unification Gray in original RGB (R, G, B) is replaced, form new color RGB (Gray, Gray, Gray), replace original RGB (R, G, B) with it and just obtain gray level image;
(3) described gray level image carrying out pretreatment, described pretreatment is that Lifting Wavelet processes, and it can extract gray level image low frequency profile information, suppress high-frequency noise;
Described Lifting Wavelet processes and includes division, prediction and update, it is considered to signal sj={ sj,l|0��l��2j, the low frequency signal that it obtains after one-level wavelet transformation is sj-1With high-frequency signal dj-1, then the process that implements building low-resolution image is:
I. division: will input signal sjBeing divided into odd, even two subsets, the set of the element composition being positioned at even subscript position is designated as evenj-1={ sj,2l|0��l��2j-1-1}, the set of the element composition being positioned at strange subscript position is designated as oddj-1={ sj,2l+1|0��l��2j-1-1};
II. prediction: on the basis based on initial data dependency, removes prediction or interpolation odd numbered sequences with the predictive value of even order;
III. update: being intended to of renewal looks for a better subset sj-1So that it is keep a certain scalar characteristic Q (x) of artwork, i.e. Q (sj-1)=Q (sj), a certain scalar characteristic Q (x) of wherein said maintenance artwork is constant for root-mean-square value;
Wherein, for Lifting Wavelet, there is division operator Split, predictive operator P and update operator U so that
(evenj-1, oddj-1)=Split (sj),
dj-1=oddj-1-P(evenj-1),
sj-1=evenj-1+U(dj-1);
(4) by binary conversion treatment, pretreated described gray level image is converted into binary image, gray level image after pixel after Lifting Wavelet processes is reduced adopts differential operator to carry out binary conversion treatment, and pretreated gray level image is converted into binary image;
(5) described binary image is carried out straight-line detection, zero-crossing examination is utilized to obtain the boundary point of binary image, and then by edge following algorithm, boundary point is converted into the symbolic information of chain representation, the chain sequence obtained based on frontier tracing carries out straight-line detection;
Described edge following algorithm mainly includes scanning process and tracking process, hypothetical boundary is present between pixel, by limit, unit is constituted, it is divided into horizontal and vertical two kinds, if currently investigating point coordinates is (x, y), judge whether that horizontal sides unit is with (x+1, y) for reference point, judge whether that vertical edges unit is with (x, y+1) for reference point, when investigating point and being identical with the value of reference point, border is not had to pass through between the two point, the value of limit unit is zero, it it is all impact point or background dot is divided into positive zero-sum negative zero two kinds according to them, border is had to pass through when the value investigating point and reference point is different between 2, the value non-zero of limit unit, it is impact point or background dot takes plus or minus respectively according to investigating point,
Tracking direction provides as follows: impact point is all the time on the left side of direction of advance, external boundary for a connected region, tracking direction is counterclockwise, inner boundary is then clockwise, when quadraturing on closing chain code basis, external boundary chain code area defined area is just, and the defined region of inner boundary chain code and amass as negative;
The value of limit unit also show corresponding tracking direction simultaneously, and certain limit unit direction is just, for horizontal sides unit, to the right, for vertical edges unit, tracking direction is upwards for tracking direction;
Whether labelling in tracking process is beaten on the impact point when front unit, to distinguish this boundary point tracked mistake;
When there is non-zero limit unit between current investigation point and reference point, according to all the other situations of 2 in these 22 �� 2 windows constituted, determining next step tracking direction, i.e. follow-up limit unit, when return to origin, tracking process terminates;
Described scanning process adopts the mode of network scanning to find the boundary point not yet followed the tracks of, and described chain code is 8 directional chain-code, and described inner boundary is the border that inner void is formed;
The described chain sequence obtained based on frontier tracing carries out in straight-line detection process, adopts following form to represent the chain sequence of object boundary:
I:Len, X, Y, d1,d2,��,dn
Wherein I is the numbering of current chain code string, and Len is the total length of chain code in current chain code string, X, and Y is the image coordinate that current chain code strings initial point, d1,d2,��,dnFor the direction code of pixel each on current chain code string, then the step of straight-line detection is:
IV. minimum length of straigh line checking: order judges each bar chain sequence, if the length of chain sequence is less than default minimum length of straigh line threshold value LT, then gives up this chain sequence;
V. from chain sequence, straightway is extracted according to straightway degree of approximation criterion:
Order judges each bar chain sequence;
1. from the off, select to meet the subchain sequence of minimum length of straigh line constraint successively, the straightway degree of approximation S of this section is calculated according to formula (1), if S >=ST, ST is default minimum straightway similarity threshold, then this section meets line constraint, turns 2., otherwise give up this section, continue to judge next cross-talk chain sequence;
Wherein, | | ps-pe| | represent end points psWith peBetween ideal line distance, Len (ps,pe) represent that subchain sequence is from end points psTo end points peThrough the physical length of pixel, S then represents the straightway degree of approximation, additionally, when calculating the physical length of chain sequence, if direction code is 0,2,4,6, then the physical length between two pixels that chain code connects is 1, if direction code is 1,3,5,7, then the physical length between two pixels is 2;
If 2. subchain the last period sequence also meets line constraint, then consider that can two adjacent strip chain sequence be merged into a longer straightway, by the last period chain sequence starting point to this section chain sequence terminal between this section of chain code regard a new chain sequence as, calculate its straightway degree of approximation S ', if S ' >=ST, then merge, otherwise this cross-talk chain sequence is marked as a new straightway;
If 3. oneself arrives chain sequence terminal, then terminate, otherwise turn and 1. continue to judge next cross-talk chain sequence;
Wherein said default minimum length of straigh line threshold value LT is 23 pixels, and default minimum straightway similarity threshold ST is 0.91;
(6) shape facility that described binary image is detected the zones of different of the line segmentation obtained is calculated, and the RGB color image collected is identified by the shape facility according to described zones of different, obtain final stem stalk target, detailed process is: straight line region is identified as stem stalk, utilizes form parameter to by other region of line segmentationAnd eccentricityBeing identified, meeting the region recognition that F is more than 2.3 and e is more than 5 is stem stalk, wherein, and the area in Area and Per respectively region and girth, the border long axis length in c and a respectively region and minor axis length.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310076981.3A CN103116747B (en) | 2013-03-11 | 2013-03-11 | Automatically the method and system of corn stem leaf image is identified |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310076981.3A CN103116747B (en) | 2013-03-11 | 2013-03-11 | Automatically the method and system of corn stem leaf image is identified |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103116747A CN103116747A (en) | 2013-05-22 |
| CN103116747B true CN103116747B (en) | 2016-06-08 |
Family
ID=48415119
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310076981.3A Expired - Fee Related CN103116747B (en) | 2013-03-11 | 2013-03-11 | Automatically the method and system of corn stem leaf image is identified |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103116747B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103714342B (en) * | 2013-12-20 | 2016-10-05 | 华北电力大学（保定） | Insulator chain automatic positioning method of taking photo by plane based on bianry image shape facility |
| CN107734120A (en) * | 2016-08-10 | 2018-02-23 | 北京东舟技术股份有限公司 | The fluency method for testing performance of terminal device |
| CN107941802B (en) * | 2017-10-31 | 2020-05-08 | 华中农业大学 | Potted rice leaf rolling degree measuring method based on machine vision |
| CN109886094A (en) * | 2019-01-08 | 2019-06-14 | 中国农业大学 | A kind of crop growth of cereal crop seedlings seedling gesture capturing analysis method and device |
| CN113807129A (en) * | 2020-06-12 | 2021-12-17 | 广州极飞科技股份有限公司 | Crop area identification method and device, computer equipment and storage medium |
| CN115147423B (en) * | 2022-09-06 | 2022-11-22 | 江苏欧罗曼家纺有限公司 | Cotton top detection method based on improved chain code method |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101645172A (en) * | 2009-09-09 | 2010-02-10 | 北京理工大学 | Rapid detection method for straight line in digital image |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8144986B2 (en) * | 2008-09-05 | 2012-03-27 | The Neat Company, Inc. | Method and apparatus for binarization threshold calculation |
-
2013
- 2013-03-11 CN CN201310076981.3A patent/CN103116747B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101645172A (en) * | 2009-09-09 | 2010-02-10 | 北京理工大学 | Rapid detection method for straight line in digital image |
Non-Patent Citations (5)
| Title |
|---|
| 一种基于小波提升变换的多尺度边缘提取算法;葛雯等;《东北大学学报（自然科学版）》;20070430;第28卷(第4期);第473-475页 * |
| 一种快速实用的直线检测算法;孙涵等;《计算机应用研究》;20060228(第2期);第256-257页 * |
| 基于图像处理的植物叶片参数测量系统;牛珂等;《农村经济与科技》;20110630;第22卷(第6期);第205-206页 * |
| 基于病症图像的玉米病害智能诊断研究;赖军臣;《中国博士学位论文全文数据库 信息科技辑》;20120115(第1期);第I138-58页 * |
| 对一种快速边缘跟踪算法的讨论;史册;《小型微型计算机系统》;20000630;第21卷(第6期);第641-645页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103116747A (en) | 2013-05-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103116747B (en) | Automatically the method and system of corn stem leaf image is identified | |
| CN107403197B (en) | Crack identification method based on deep learning | |
| CN102254319B (en) | Method for carrying out change detection on multi-level segmented remote sensing image | |
| CN103400151B (en) | The optical remote sensing image of integration and GIS autoregistration and Clean water withdraw method | |
| CN109145872B (en) | CFAR and Fast-RCNN fusion-based SAR image ship target detection method | |
| Shao et al. | Application of a fast linear feature detector to road extraction from remotely sensed imagery | |
| CN103020605B (en) | Bridge identification method based on decision-making layer fusion | |
| CN103455991B (en) | A kind of multi-focus image fusing method | |
| CN105719259A (en) | Pavement crack image detection method | |
| CN108921076B (en) | Pavement crack disease self-adaptive constant false alarm detection method based on image | |
| CN109766887A (en) | A kind of multi-target detection method based on cascade hourglass neural network | |
| CN103049763A (en) | Context-constraint-based target identification method | |
| CN104268515A (en) | Sperm morphology anomaly detection method | |
| CN103903268A (en) | Road surface crack detection method based on multiple lower-layer marked features | |
| CN101645172A (en) | Rapid detection method for straight line in digital image | |
| CN103745468A (en) | Significant object detecting method based on graph structure and boundary apriority | |
| CN104952070A (en) | Near-rectangle guide based remote-sensing cornfield image segmentation method | |
| CN103077507B (en) | Beta algorithm-based multiscale SAR (Synthetic Aperture Radar) image denoising method | |
| CN110363160A (en) | A kind of Multi-lane Lines recognition methods and device | |
| CN116452667A (en) | Target identification and positioning method based on image processing | |
| CN117392157B (en) | Edge-aware protective cultivation straw coverage rate detection method | |
| CN101853503A (en) | Spectral line inflexion multi-scale optimizing segmentation method and application thereof | |
| CN117314940B (en) | Laser cutting part contour rapid segmentation method based on artificial intelligence | |
| CN105354798B (en) | SAR image denoising method based on geometry priori and dispersion similarity measure | |
| CN104966296A (en) | Moving window N-Smoothlets image edge detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20160608 Termination date: 20170311 |