CN113255434A - Apple identification method fusing fruit features and deep convolutional neural network - Google Patents
Apple identification method fusing fruit features and deep convolutional neural network Download PDFInfo
- Publication number
- CN113255434A CN113255434A CN202110376545.2A CN202110376545A CN113255434A CN 113255434 A CN113255434 A CN 113255434A CN 202110376545 A CN202110376545 A CN 202110376545A CN 113255434 A CN113255434 A CN 113255434A
- Authority
- CN
- China
- Prior art keywords
- fruit
- apple
- image
- frame
- shape
- 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
- 235000013399 edible fruits Nutrition 0.000 title claims abstract description 85
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 19
- 238000010586 diagram Methods 0.000 claims description 15
- 238000001514 detection method Methods 0.000 claims description 9
- 238000009826 distribution Methods 0.000 claims description 9
- 238000001228 spectrum Methods 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 6
- 238000003709 image segmentation Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 230000010354 integration Effects 0.000 claims description 2
- 238000003064 k means clustering Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 5
- 241000220225 Malus Species 0.000 description 38
- 230000003595 spectral effect Effects 0.000 description 4
- 238000012271 agricultural production Methods 0.000 description 2
- 230000032683 aging Effects 0.000 description 1
- 235000021016 apples Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 235000012055 fruits and vegetables Nutrition 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002420 orchard Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention discloses an apple identification method fusing fruit characteristics and a deep convolutional neural network, which takes fast RCNN as a basic network and realizes the customization of an apple identification frame by fusing the fruit characteristics into an input layer, RPN and a position regression branch of the fast RCNN, thereby effectively improving the accuracy of apple identification and positioning. The algorithm fully considers the color, the shape and other characteristics of the apple, and has an important effect on improving the fruit identification accuracy under the conditions of adhesion and shielding.
Description
Technical Field
The invention relates to the field of computer vision and agricultural engineering, in particular to an apple identification method used in a natural environment, and specifically relates to an apple identification method fusing fruit characteristics and a deep convolutional neural network.
Background
A large amount of labor force is needed to carry out concentrated picking operation in the fruit mature period, so that the demand for labor force is large, and the demand time is concentrated. However, the aging of agricultural population in China is serious at present, and it is predicted that the population actually engaged in agricultural production will be reduced sharply and the labor cost of agricultural production will be improved remarkably in the next fifteen years. Therefore, the research and development of efficient picking robots for replacing manual fruit and vegetable picking operation is urgent.
The identification and positioning of apples is the focus of research on apple picking robots. The fruit identification method based on the deep convolutional neural network is a mainstream method for identifying the fruit at present, and the speed and the effect of identification exceed those of the traditional fruit identification method. However, most of the fruit identification methods based on the deep convolutional neural network do not comprehensively consider the characteristics of the fruits, and the accuracy of identification and positioning is difficult to be further improved. Due to the complex orchard environment, fruits are affected by various factors such as light, adhesion and shielding, and the like, the structure of the deep convolutional neural network needs to be improved and optimized by combining the characteristics of the fruits, so that the accuracy of identification and positioning is effectively improved.
Disclosure of Invention
Aiming at the technical problems, the technical scheme provides the apple identification method fusing fruit features and the deep convolutional neural network, and the problems can be effectively solved.
The invention is realized by the following technical scheme:
an apple identification method fusing fruit features and a deep convolutional neural network, which takes fast RCNN as a basic network and realizes the customization of an apple identification frame by fusing the fruit features into an input layer, RPN and a position regression branch of the fast RCNN, and comprises the following steps:
step 1: taking a fast RCNN target detection frame as a basic network frame, and fusing fruit characteristics into an input layer, an RPN and a position regression branch of the fast RCNN;
step 2: expressing the color, shape, texture, edge, spectrum and three-dimensional characteristics of the apple in a graphical mode, and integrating the color, shape, texture, edge, spectrum and three-dimensional characteristics of the apple and a fruit image into a multi-channel image as input of a target detection framework;
and step 3: determining a region where the fruit can exist according to the color characteristics of the fruit, and limiting the generation position of a preselected frame in the RPN to the region;
and 4, step 4: the shape of a preselection frame in the RPN is set to be circular, so that the preselection frame is adaptive to the shape of an apple, and partial background interference on the characteristics of the purified fruit is effectively shielded;
and 5: setting the size of a preselected frame according to the size distribution of the apple fruits in the image so as to enable the size of the preselected frame to be matched with the size of the apple fruits;
step 6: and (3) replacing the branch for performing rectangular box position regression in the fast RCNN framework by adopting a branch structure capable of performing oval shape prediction and position regression.
Further, the fruit features in step 2 are graphically expressed, wherein the color features are expressed by a color difference graph or a cluster graph, the shape features are expressed by a gradient graph, the texture features are expressed by a code graph, the edge features are expressed by a side graph, the spectral features are expressed by a multilayer spectral reflection graph, and the three-dimensional shape features are expressed by a point cloud graph or an intensity graph.
Further, the step 2 of integrating the fruit image into a multi-channel image as an input of the target detection framework specifically includes the following steps: and superposing the graphical characteristics of the fruits on the third dimension of the fruit image to form an image containing a plurality of channels, and modifying the number of channels of the fast RCNN input layer according to the number of the image channels.
Further, the step 3 of determining the possible existing area of the fruit according to the color characteristics of the fruit comprises the following specific operation steps:
firstly, generating a color difference graph Gray of the apple image by adopting an R-G operator, wherein the specific R-G operator is as follows:
then 1/10 with the best threshold value calculated by adopting a maximum inter-class variance method is used for image segmentation, and expansion is carried out on the connected region to obtain a region where the apple may exist; the maximum inter-class variance method classifies the pixels in the image into 2 classes by finding the optimal threshold Thr, and the specific formula is as follows:
whereinAndinter-class variance and intra-class variance, respectively, for the 2 classes of pixel values; mu.s1And mu2Respectively, the mean values of the 2 types of pixel values; p is a radical of1And p2The proportion of the number of the 2 types of pixels to the total number of the pixels is respectively;andrespectively, the variance of the class 2 pixel values.
Further, the step 4 of setting the shape of the preselected frame in the RPN to be a circle includes the following specific steps:
firstly, the shape of a preselected frame in the RPN is set to be square and used for participating in operation, and the specific operation mode is as follows: generating 3 square pre-selection frames with different sizes by taking each pixel of a possible fruit existing area as a center, and then mapping the pre-selection frames onto the feature map according to the size proportion of the input image and the feature map;
and then, when the features in the preselected frame are extracted, the shape of the preselected frame is modified into a circle by adopting a mask pattern, so that the preselected frame is adaptive to the shape of the apple.
Further, setting the size of the preselected frame in the step 5, firstly marking the size of the fruits in the image in a large scale, then analyzing and drawing a size distribution diagram of the sample fruits by adopting a clustering algorithm, and finally selecting a proper value in the diagram as the size of the preselected frame; the specific clustering algorithm is as follows: first in the size distribution range [ x ]1,x2]Selecting 3 seed points with the values of x1+ h, x1+3h and x1+5h respectively, wherein h is (x)2-x1) 6; and then clustering by using a k-means clustering algorithm and taking the seed point as a center to determine a proper preselected box size.
Further, the specific operation steps of the elliptic regression branch in step 6 are as follows:
firstly, pixel-level segmentation of a local area is realized on the basis of a full convolution network, and then the central coordinate, the major and minor axis radius, the inclination angle and the offset of the ellipse are predicted through an inserted parameter regression module.
(III) advantageous effects
Compared with the prior art, the apple identification method fusing fruit characteristics and the deep convolutional neural network has the following beneficial effects:
(1) according to the technical scheme, under the framework of a deep convolutional neural network, various characteristics of the fruits such as color, texture and shape are effectively fused, particularly, a circular preselection frame and an elliptical prediction branch structure are designed for the shape characteristics of the fruits, the identification and positioning accuracy of the fruits is improved, and the fruit adhesion and the fruit shielding are more remarkable.
(2) The technical scheme adds the characteristic input on the basis of fast RCNN, so that the characteristics of an input network are more diversified, the convolutional network is facilitated to learn the corresponding characteristics, and the identification accuracy is improved. Meanwhile, the RPN network is improved, and the shape, the size and the generation mode of the preselected frame are mainly modified, so that the preselected frame is more matched with the shape of the apple, and the target characteristics are purified. And moreover, a branch capable of performing elliptic regression and position prediction is provided, and an original rectangular positioning frame regression branch is replaced, so that the network can better identify and position adhesion and shielding fruits.
Drawings
Fig. 1 is an overall structure diagram of a fruit identification method according to the present invention.
FIG. 2 is a block diagram of the fast RCNN object detection framework of the present invention.
FIG. 3a is a color feature expression diagram of an apple image according to the present invention.
FIG. 3b is a diagram showing the shape characteristics of an apple image according to the present invention.
FIG. 3c is an edge feature representation of an apple image according to the present invention.
FIG. 3d is a three-dimensional shape feature representation of an apple image according to the present invention.
Fig. 4a is an original drawing of an apple of the present invention.
FIG. 4b is a color difference diagram of the apple represented by the R-G operator in the present invention.
Fig. 4c is an exemplary diagram of the area where an apple may exist in the present invention.
FIG. 5 is a diagram illustrating mask patterns of an apple of the present invention.
Fig. 6 is a graph showing the effect of apple identification in the present invention.
Detailed Description
The technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. The described embodiments are only some embodiments of the invention, not all embodiments. Various modifications and improvements of the technical solutions of the present invention may be made by those skilled in the art without departing from the design concept of the present invention, and all of them should fall into the protection scope of the present invention.
Example 1:
an apple identification method fusing fruit features and a deep convolutional neural network, which takes fast RCNN as a basic network and realizes the customization of an apple identification frame by fusing the fruit features into an input layer, RPN and a position regression branch of the fast RCNN, and comprises the following steps:
step 1: taking a Faster RCNN target detection framework as a basic network framework, as shown in FIG. 2; on the basis, the fruit features are fused into the input layer, RPN and position regression branch of fast RCNN, and the overall structure diagram of fast RCNN after the fruit features are fused is shown in FIG. 1.
Comparing fig. 1 and fig. 2, it can be seen that:
(1) the input of the characteristics is added on the basis of the fast RCNN, so that the characteristics of an input network are more diversified, the convolution network is facilitated to learn the corresponding characteristics, and the identification accuracy is improved.
(2) The RPN network is improved, and the shape and the size of the preselected frame and the generation mode are mainly modified, so that the preselected frame is more matched with the shape of the apple, and the target characteristics are purified.
(3) A branch capable of performing elliptic regression and position prediction is provided, and an original rectangular positioning frame regression branch is replaced, so that the network can better identify and position adhesion and shielding fruits.
Step 2: the color, shape, texture, edge, spectrum and stereo characteristics of the apple are expressed in a graphical mode, and are integrated with the fruit image into a multi-channel image to be used as the input of the target detection framework. The color feature is expressed by a color difference graph or a cluster graph, the shape feature is expressed by a gradient graph, the texture feature is expressed by a code graph, the edge feature is expressed by a side line graph, the spectral feature is expressed by a multilayer spectral reflection graph, and the three-dimensional shape feature is expressed by a point cloud graph or an intensity graph.
The specific integration is as follows: and superposing the graphical characteristics of the fruits on the third dimension of the fruit image to form an image containing a plurality of channels, and modifying the number of channels of the fast RCNN input layer according to the number of the image channels. (feel space, cannot be as good as the last patent; provide an example; let it be filled).
Fig. 3a is a three-channel color apple image, fig. 3b adopts an R-G color operator to perform graphical expression of apple color features, fig. 3c adopts a canny operator to perform graphical expression of edge features of a fruit image, and fig. 3d adopts a depth map to perform graphical expression of three-dimensional features of the fruit image.
And step 3: and determining the possible existing region of the fruit according to the color characteristics of the fruit, and limiting the generation position of the preselected frame in the RPN to the region. Firstly, generating a color difference graph Gray of the apple image by adopting an R-G operator, wherein the specific R-G operator is as follows:
then 1/10 with the best threshold value calculated by adopting a maximum inter-class variance method is used for image segmentation, and expansion is carried out on the connected region to obtain a region where the apple may exist; the maximum inter-class variance method classifies the pixels in the image into 2 classes by finding the optimal threshold Thr, and the specific formula is as follows:
whereinAndinter-class variance and intra-class variance, respectively, for the 2 classes of pixel values; mu.s1And mu2Are respectively asMean of class 2 pixel values; p is a radical of1And p2The proportion of the number of the 2 types of pixels to the total number of the pixels is respectively;andrespectively, the variance of the class 2 pixel values.
The region occupies 1/5 the total area of the image, and the generation of the pre-selection frame in the region can greatly reduce the calculation amount of the RPN network and reduce the risk of false identification; fig. 4 shows an example of a possible fruit region, fig. 4b is a color difference diagram of an R-G operator, and fig. 4c is a possible fruit region.
And 4, step 4: the shape of the preselection frame in the RPN is set to be circular, so that the preselection frame is adaptive to the shape of the apple, and partial background interference on the characteristics of the purified fruit is effectively shielded. Firstly, the shape of a preselected frame in the RPN is set to be square and used for participating in operation, and the specific operation mode is as follows: generating 3 square pre-selection frames with different sizes by taking each pixel of a possible fruit existing area as a center, and then mapping the pre-selection frames onto the feature map according to the size proportion of the input image and the feature map;
and then, when the features in the preselected frame are extracted, the shape of the preselected frame is modified into a circle by adopting a mask pattern, so that the preselected frame is adaptive to the shape of the apple. Fig. 5 shows a mask pattern overlaid on a corresponding square pre-selected frame, leaving only white areas and not black areas of the mask pattern.
And 5: and setting the size of the preselected frame according to the size distribution of the apple fruits in the image so as to be matched with the size of the fruits. The method comprises the steps of firstly marking the size of fruits in an image in a large scale, then analyzing and drawing a size distribution diagram of sample fruits by adopting a clustering algorithm, and finally selecting a proper value in the diagram as a preselected frame size. The specific clustering algorithm is as follows:
first in the size distribution range [ x ]1,x2]Selecting 3 seed points with the values of x1+ h, x1+3h and x1+5h respectively, wherein h is (x)2-x1) 6; then using k-meansThe clustering algorithm performs clustering with the seed point as the center to determine a suitable preselected box size.
Step 6: and (3) replacing the branch for performing rectangular box position regression in the fast RCNN framework by adopting a branch structure capable of performing oval shape prediction and position regression.
Firstly, pixel-level segmentation of a local area is realized on the basis of a full convolution network, and then the central coordinate, the major and minor axis radius, the inclination angle and the offset of the ellipse are predicted through an inserted parameter regression module. An example of the final recognition effect is shown in fig. 6.
Claims (7)
1. An apple identification method fusing fruit features and a deep convolutional neural network is characterized in that: the method takes the Faster RCNN as a basic network, and realizes the customization of the apple identification frame by integrating the fruit characteristics into an input layer, an RPN and a position regression branch of the Faster RCNN, and comprises the following steps:
step 1: taking a fast RCNN target detection frame as a basic network frame, and fusing fruit characteristics into an input layer, an RPN and a position regression branch of the fast RCNN;
step 2: expressing the color, shape, texture, edge, spectrum and three-dimensional characteristics of the apple in a graphical mode, and integrating the color, shape, texture, edge, spectrum and three-dimensional characteristics of the apple and a fruit image into a multi-channel image as input of a target detection framework;
and step 3: determining a region where the fruit can exist according to the color characteristics of the fruit, and limiting the generation position of a preselected frame in the RPN to the region;
and 4, step 4: the shape of a preselection frame in the RPN is set to be circular, so that the preselection frame is adaptive to the shape of an apple, and partial background interference on the characteristics of the purified fruit is effectively shielded;
and 5: setting the size of a preselected frame according to the size distribution of the apple fruits in the image so as to enable the size of the preselected frame to be matched with the size of the apple fruits;
step 6: and (3) replacing the branch for performing rectangular box position regression in the fast RCNN framework by adopting a branch structure capable of performing oval shape prediction and position regression.
2. The apple identification method fusing fruit features and the deep convolutional neural network as claimed in claim 1, wherein: and 2, graphically expressing the fruit characteristics, wherein the color characteristics are expressed by using a color difference graph or a cluster graph, the shape characteristics are expressed by using a gradient graph, the texture characteristics are expressed by using a coding graph, the edge characteristics are expressed by using a side line graph, the spectrum characteristics are expressed by using a multilayer spectrum reflection graph, and the three-dimensional shape characteristics are expressed by using a point cloud graph or an intensity graph.
3. The apple identification method fusing fruit features and the deep convolutional neural network as claimed in claim 2, wherein: integrating the fruit image and the multichannel image into a multichannel image as the input of a target detection framework in the step 2, wherein the specific integration mode is as follows: and superposing the graphical characteristics of the fruits on the third dimension of the fruit image to form an image containing a plurality of channels, and modifying the number of channels of the fast RCNN input layer according to the number of the image channels.
4. The apple identification method fusing fruit features and the deep convolutional neural network as claimed in claim 1, wherein: determining the possible existing area of the fruit according to the color characteristics of the fruit in the step 3, wherein the specific operation steps are as follows:
firstly, generating a color difference graph Gray of the apple image by adopting an R-G operator, wherein the specific R-G operator is as follows:
then 1/10 with the best threshold value calculated by adopting a maximum inter-class variance method is used for image segmentation, and expansion is carried out on the connected region to obtain a region where the apple may exist; the maximum inter-class variance method classifies the pixels in the image into 2 classes by finding the optimal threshold Thr, and the specific formula is as follows:
whereinAndinter-class variance and intra-class variance, respectively, for the 2 classes of pixel values; mu.s1And mu2Respectively, the mean values of the 2 types of pixel values; p is a radical of1And p2The proportion of the number of the 2 types of pixels to the total number of the pixels is respectively;andrespectively, the variance of the class 2 pixel values.
5. The apple identification method fusing fruit features and the deep convolutional neural network as claimed in claim 1, wherein: setting the shape of the preselected frame in the RPN to be a circle in step 4, specifically, the following steps are performed:
firstly, the shape of a preselected frame in the RPN is set to be square and used for participating in operation, and the specific operation mode is as follows: generating 3 square pre-selection frames with different sizes by taking each pixel of a possible fruit existing area as a center, and then mapping the pre-selection frames onto the feature map according to the size proportion of the input image and the feature map;
and then, when the features in the preselected frame are extracted, the shape of the preselected frame is modified into a circle by adopting a mask pattern, so that the preselected frame is adaptive to the shape of the apple.
6. The apple identification method fusing fruit features and the deep convolutional neural network as claimed in claim 1, wherein: setting the size of a preselected frame in the step 5, firstly marking the size of fruits in the image in a large scale, then analyzing and drawing a size distribution diagram of sample fruits by adopting a clustering algorithm, and finally selecting a proper value in the diagram as the size of the preselected frame; the specific clustering algorithm is as follows: first in the size distribution range [ x ]1,x2]Selecting 3 seed points with the values of x1+ h, x1+3h and x1+5h respectively, wherein h is (x)2-x1) 6; and then clustering by using a k-means clustering algorithm and taking the seed point as a center to determine a proper preselected box size.
7. The apple identification method fusing fruit features and the deep convolutional neural network as claimed in claim 1, wherein: and 6, the ellipse regression branch firstly realizes the pixel level segmentation of a local area on the basis of a full convolution network, and then predicts the central coordinate, the major and minor axis radius, the inclination angle and the offset of the ellipse through an inserted parameter regression module.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110376545.2A CN113255434B (en) | 2021-04-08 | 2021-04-08 | Apple identification method integrating fruit characteristics and deep convolutional neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110376545.2A CN113255434B (en) | 2021-04-08 | 2021-04-08 | Apple identification method integrating fruit characteristics and deep convolutional neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113255434A true CN113255434A (en) | 2021-08-13 |
CN113255434B CN113255434B (en) | 2023-12-19 |
Family
ID=77220466
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110376545.2A Active CN113255434B (en) | 2021-04-08 | 2021-04-08 | Apple identification method integrating fruit characteristics and deep convolutional neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113255434B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114612769A (en) * | 2022-03-14 | 2022-06-10 | 电子科技大学 | Integrated sensing infrared imaging ship detection method integrated with local structure information |
CN114913179A (en) * | 2022-07-19 | 2022-08-16 | 南通海扬食品有限公司 | Apple skin defect detection system based on transfer learning |
CN116021526A (en) * | 2023-02-07 | 2023-04-28 | 台州勃美科技有限公司 | Agricultural robot control method and device and agricultural robot |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105718945A (en) * | 2016-01-20 | 2016-06-29 | 江苏大学 | Apple picking robot night image identification method based on watershed and nerve network |
CN108470339A (en) * | 2018-03-21 | 2018-08-31 | 华南理工大学 | A kind of visual identity of overlapping apple and localization method based on information fusion |
WO2019040866A2 (en) * | 2017-08-25 | 2019-02-28 | The Board Of Trustees Of The University Of Illinois | Apparatus and method for agricultural data collection and agricultural operations |
CN109948444A (en) * | 2019-02-19 | 2019-06-28 | 重庆理工大学 | Method for synchronously recognizing, system and the robot of fruit and barrier based on CNN |
CN111325764A (en) * | 2020-02-11 | 2020-06-23 | 广西师范大学 | Fruit image contour recognition method |
JP2020205048A (en) * | 2019-06-18 | 2020-12-24 | 富士通株式会社 | Object detection method based on deep learning network, apparatus, and electronic device |
CN112395972A (en) * | 2020-11-16 | 2021-02-23 | 中国科学院沈阳自动化研究所 | Electric power system insulator string identification method based on unmanned aerial vehicle image processing |
-
2021
- 2021-04-08 CN CN202110376545.2A patent/CN113255434B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105718945A (en) * | 2016-01-20 | 2016-06-29 | 江苏大学 | Apple picking robot night image identification method based on watershed and nerve network |
WO2019040866A2 (en) * | 2017-08-25 | 2019-02-28 | The Board Of Trustees Of The University Of Illinois | Apparatus and method for agricultural data collection and agricultural operations |
CN108470339A (en) * | 2018-03-21 | 2018-08-31 | 华南理工大学 | A kind of visual identity of overlapping apple and localization method based on information fusion |
CN109948444A (en) * | 2019-02-19 | 2019-06-28 | 重庆理工大学 | Method for synchronously recognizing, system and the robot of fruit and barrier based on CNN |
JP2020205048A (en) * | 2019-06-18 | 2020-12-24 | 富士通株式会社 | Object detection method based on deep learning network, apparatus, and electronic device |
CN111325764A (en) * | 2020-02-11 | 2020-06-23 | 广西师范大学 | Fruit image contour recognition method |
CN112395972A (en) * | 2020-11-16 | 2021-02-23 | 中国科学院沈阳自动化研究所 | Electric power system insulator string identification method based on unmanned aerial vehicle image processing |
Non-Patent Citations (4)
Title |
---|
XIAOYANG LIU: ""A Detection Method for Apple Fruits Based on Color and Shape Features"", 《IEEE ACCESS》 * |
刘晓洋: ""多功能果蔬采摘机器人的图像识别算法研究"", 《中国博士学位论文全文数据库》 * |
樊泽泽;柳倩;柴洁玮;杨晓峰;李海芳;: "基于颜色与果径特征的苹果树果实检测与分级", 计算机工程与科学, no. 09 * |
田玉宇: ""复杂环境下苹果采摘机器人的目标识别算法"", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114612769A (en) * | 2022-03-14 | 2022-06-10 | 电子科技大学 | Integrated sensing infrared imaging ship detection method integrated with local structure information |
CN114913179A (en) * | 2022-07-19 | 2022-08-16 | 南通海扬食品有限公司 | Apple skin defect detection system based on transfer learning |
CN116021526A (en) * | 2023-02-07 | 2023-04-28 | 台州勃美科技有限公司 | Agricultural robot control method and device and agricultural robot |
Also Published As
Publication number | Publication date |
---|---|
CN113255434B (en) | 2023-12-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113255434B (en) | Apple identification method integrating fruit characteristics and deep convolutional neural network | |
CN105718945B (en) | Apple picking robot night image recognition method based on watershed and neural network | |
Liu et al. | A detection method for apple fruits based on color and shape features | |
CN108009542B (en) | Weed image segmentation method in rape field environment | |
Bai et al. | Crop segmentation from images by morphology modeling in the CIE L* a* b* color space | |
Xiong et al. | The recognition of litchi clusters and the calculation of picking point in a nocturnal natural environment | |
CN111798467B (en) | Image segmentation method, device, equipment and storage medium | |
CN104598908B (en) | A kind of crops leaf diseases recognition methods | |
CN110610506B (en) | Image processing technology-based agaricus blazei murill fruiting body growth parameter detection method | |
CN106127735B (en) | A kind of facilities vegetable edge clear class blade face scab dividing method and device | |
CN113674226A (en) | Tea leaf picking machine tea leaf bud tip detection method based on deep learning | |
CN102214306A (en) | Leaf disease spot identification method and device | |
CN109145964B (en) | Method and system for realizing image color clustering | |
CN109740721A (en) | Wheat head method of counting and device | |
CN114241326B (en) | Progressive intelligent production method and system for ground feature elements of remote sensing images | |
CN111783693A (en) | Intelligent identification method of fruit and vegetable picking robot | |
CN111008642A (en) | High-resolution remote sensing image classification method and system based on convolutional neural network | |
CN111798470A (en) | Crop image entity segmentation method and system applied to intelligent agriculture | |
CN115862004A (en) | Corn ear surface defect detection method and device | |
CN115861686A (en) | Litchi key growth period identification and detection method and system based on edge deep learning | |
CN111079637A (en) | Method, device and equipment for segmenting rape flowers in field image and storage medium | |
CN111832508B (en) | DIE _ GA-based low-illumination target detection method | |
CN107239761B (en) | Fruit tree branch pulling effect evaluation method based on skeleton angular point detection | |
CN110084820A (en) | Purple soil image adaptive division and extracting method based on improved FCM algorithm | |
CN111191659B (en) | Multi-shape clothes hanger identification method for clothing production system |
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 |