CN110188657A - Corn arid recognition methods based on crimping blade detection - Google Patents
Corn arid recognition methods based on crimping blade detection Download PDFInfo
- Publication number
- CN110188657A CN110188657A CN201910446308.1A CN201910446308A CN110188657A CN 110188657 A CN110188657 A CN 110188657A CN 201910446308 A CN201910446308 A CN 201910446308A CN 110188657 A CN110188657 A CN 110188657A
- Authority
- CN
- China
- Prior art keywords
- corn
- crimping blade
- image
- model
- detection
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to corn arid identification technology fields, are identified by the detection to corn crimping blade under drought stress to corn arid.The present invention is by obtaining corn digital picture under drought stress, corn arid is timely diagnosed and identified by the detection of corn crimping blade, arid identification based on crimping blade detection is a kind of quick, lossless, direct drought monitoring method, algorithm of target detection can not only detect crimping blade, and can be positioned to crimping blade and be partitioned into crimping blade.
Description
Technical field
The present invention relates to crops arids to identify field, and in particular, to a kind of corn based on crimping blade detection is dry
Non-irrigated recognition methods.
Background technique
Arid is to influence one of the principal element of corn yield, and quickly, timely corn draught monitor produces guarantee corn
Amount plays an important role, and traditional drought monitoring method has soil moisture, agrometeorological forecasting and artificial diagnosis etc., though it can be right
Corn arid judges, but has hysteresis quality.Plant shows a series of phenotypic characteristic, such as leaf under drought stress
Here piece withers, curling, and plant strain growth is slow etc., and wherein leaf rolling is one of most significant feature of corn drought stress.With meter
It is raw that the development of calculation machine vision and crop phenotype group, traditional machine learning and deep learning algorithm are widely used in crops
Object or abiotic stress identification (Singh et al., 2016；Singh et al., 2018), and common image recognition and
Sorting algorithm is that the content of an image is identified and classified, however is especially under field condition in agricultural production practice
Piece image may either there are many crops coerced comprising a variety of objects (such as crop and weeds, Different Crop), such as wrap
Containing different pest and disease damages or since soil moisture is unevenly distributed, it is existing suitable in the piece image acquired under field condition to cause
The plant of moisture also has arid to wither plant here (An et al., 2019).It include more objects, multi-target condition in piece image
Under, by influencing each other between complicated image background and more objects, common image recognition algorithm can not carry out image
Accurate identification and classification.Therefore image recognition algorithm still has problem in more object identifications.Computer vision technique is answered
For biology or abiotic stress preventing and reducing natural disasters and in monitoring and warning not only will in image biology or abiotic stress carry out
Correctly identification and classification, and target position is accurately positioned, so that we take suitable and accurately take precautions against natural calamities
Hazard mitigation measure ensures crop normal growth, and it is horizontal to improve crop yield.
Target detection is to be identified using algorithm of target detection in computer vision to target object in piece image, point
Class and positioning.Algorithm of target detection should identify the classification of objects in images, also find out the position of object in the picture
(Felzenszwalb et al.,2010).In recent years, with the development of computer vision and artificial intelligence, object detection method
Gradually it is applied in the application study of crops biology or abiotic stress monitoring and warning.Wu Lulu etc. (Wu Lulu etc., 2014)
Crop disease is detected using edge detection and Hough transform algorithm based on 90 width corps diseases images, detection
Fitting precision is 87.01%, position error 4.44%.Xie Zhonghong etc. (Xie Zhong red etc., 2010) proposes a kind of based on Hough
The fruit detection method of transformation, carries out image segmentation using 2R-G color component, detects fruit profile using template matching method.
The result shows that this method can accurately detect in fact class circle fruit.Xia et al (Xia et al., 2018) uses depth
It practises model to identify insect in 24, application region suggests that network (Region Proposal Network) is right in a model
Image carries out processing and generates target suggestion window, and test result shows that deep learning model can obtain very high precision.Although mesh
It is preceding to be widely used (Li et al., 2016) in terms of face and vehicle identification based on the algorithm of target detection of deep learning, but
Agriculture field, especially agro-ecology or abiotic stress recognition methods research are less,
Hasan et al (Hasan et al., 2018) obtains the digital picture of 10 wheat breeds under field condition,
The wheat wheat head is accurately detected and counted using R-CNN, wheat wheat ear density is calculated and wheat yield is predicted.
Tian et al (Tian et al., 2019) obtains the Apple image of different bearing stage using digital camera, and use is improved
Algorithm of target detection YOLO3 detects the apple under different bearing stage, while comparing YOLO3 and YOLO2 and Faster
The effect that R-CNN detects apple, the results showed that YOLO3 has better detection performance.One of main abiotic stress of arid
(Li Maosong etc., 2005).Under drought stress, corn moisture metabolism and photosynthesis are obstructed, and plant shows blade and withers
Here, it turns to be yellow, curling, the characters such as tarnish, and leaf rolling is phenotypic characteristic the most typical in corn drought stress phenotype
One of (Kadioglu et al., 2012), in the case where other growing environment conditions are certain, maize leaf curling can be considered
Mainly caused by drought stress.Therefore the detection to corn crimping blade is expected that by herein accurately to know corn arid
Not, it is the information provided promptly and accurately of preventing and reducing natural disasters of corn drought stress, to the accuracy and timeliness irrigated is improved, ensures
Corn high and stable yield, meanwhile, this method is equally applicable to other crops biology or abiotic stress detection and identification aspect, such as disease
The accurate detection and identification of insect pest.
Summary of the invention
Leaf rolling is corn drought stress phenotypic characteristic the most typical and sensitive, present invention contemplates that being crimped by corn
The detection of blade carries out timely, accurate, lossless identification to corn arid, to improve the timeliness irrigated.
In order to achieve the object of the present invention, a kind of corn arid recognition methods based on crimping blade detection of the invention, packet
Include following steps:
A) test of corn control drought is carried out by pot-culture method, being divided into is suitable for handling with drought stress two；
B) the vertical view digital picture for being obtained corn at the top of corn at random using slr camera, is established image data set, made
How fine polygon collimation mark is carried out to corn crimping blade with Labelme software to infuse, and data are divided into training sample and survey
Sample sheet, while json file is converted by image data；
C) using the training set image training objective detection model Mask R-CNN marked, train epochs are 50 steps,
Characteristics of image is extracted using ResNet101 in Mask R-CNN model；
D) trained crimping blade detection model Mask R-CNN model is tested with test set, with mean value mean accuracy
MAP, precision, and detect the specific location of crimping blade in the picture, crimping blade is split.
The corn image for obtaining different drought grade by pot experiment first, then pre-processes image, to figure
As size is processed for carrying out image labeling after image preprocessing, the image that will have been marked in storing the input with model
It is divided into training set and test set, training data input Mask R-CNN is trained after image labeling is good, finally obtains corn
Crimping blade detection model；
After training crimping blade detection model, test image input crimping blade detection model is trained into and has been rolled up
Bent crop leaf measuring is as a result, mainly include the quantity of crimping blade, curling confidence level divides mask.
Further, corn crimping blade is labeled using Labelme software, object is carried out using polygon frame
It is fine to mark, a json file is generated after every image labeling, contains the information of image labeling in json file, mainly has
The position coordinates of target object, the class label and image size information of target object carry out the note of figure after image labeling
It releases, each crimping blade is drawn using OpenCV software according to the location information of the crimping blade mark in json file respectively
Bianry image, one of format needed for finally generating Mask R-CNN according to the annotation information of each crimping blade and original image
Json file, this document contain the class label of all images, the coordinate information of crimping blade；It will mark in model training
Good corn image data inputs Mask R-CNN model, and image data set random division is training set and test set, wherein
80% is divided into training set, and 20% is divided into test set, passes through tensorflow visualization tool in the training process
Tensorboard checks the convergent of model, is assessed after model training is good with test the set pair analysis model.
The present invention accurately identifies corn arid by the detection to corn crimping blade, is corn drought stress
The information provided promptly and accurately of preventing and reducing natural disasters ensures corn high and stable yield to the timeliness irrigated is improved, meanwhile, this method
It may be used on other crops biology or abiotic stress detection and identification aspect, such as the accurate detection and identification of pest and disease damage.
Detailed description of the invention
Fig. 1 is the pond RoIAlig and Roipooling operation chart.
Fig. 2 is the flow chart for crimping crop leaf measuring to corn using Mask R-CNN.
Fig. 3 is image labeling.
Fig. 4 is IoU threshold value calculation method.
Fig. 5 is crimping blade detection model to the detection of crimping blade and curling confidence level.
Specific embodiment
In order to make those skilled in the art that the present invention may be better understood, with reference to the accompanying drawings and examples to this hair
Bright technical solution further illustrates.
It is Zheng Dan 958 for examination corn variety, which is the maximum corn variety of the current cultivated area in China, has height
The characteristics of production, stable yields, anti-fall, disease-resistant, wide adaptability, comprehensive agronomy character is good, and Huang-Huai-Hai summer sowing breeding time 96 days or so,
Plant height 240cm, Ear height 100cm or so.
This experiment using pot experiment obtain corn drought stress image, basin high 45cm, base diameter 23cm, on diametrically
30cm obtains corn image at corn jointing-heading stage, 4 plants of plants is put together in image acquisition, including arid
The plant of stress and adequate moisture obtains corn from top using slr camera and overlooks digital picture, obtains image 200 altogether
, slr camera used in pot experiment is EOS700D, valid pixel 18,000,000, CMOS inductor, camera lens real focal length f
=18-135mm, shooting photo are automatically stored in SD card with JPG format.Every 282 signal receiver of camera mounted article color is used for
Automatic to obtain corn image, camera is away from ground 2.5m, daily from 6:00-18:00 between control drought period.Image is obtained to use later
Labelme software is labeled corn crimping blade, and json format needed for being converted into Mask R-CNN.
Mask R-CNN be based on Faster R-CNN framework propose new convolutional network Mask R-CNN be
Mask predicted branches (Mask representation branch) are increased on the basis of Faster R-CNN and use RoIAlign
RoI pooling is replaced to realize that characteristic image is aligned (He et al., 2018) with original image Pixel-level.Pervious target detection
(Faster R-CNN) algorithm is identified and positioned to target in image, in without be added example segmentation, this is to farming
The detection and positioning of object are less accurate.Since object mutually blocks in crops, environmental condition is complicated, and scab in crops
Or other coerce characteristic images often region very little, are irregular figure, such as pest and disease damage, crimping blade very little.Target inspection
Method of determining and calculating not only wants detection object, also carries out finely positioning to object.RoI pooling is used in Faster R-CNN model
Pondization operation is carried out to characteristic pattern, RoI pooling has done quantization operation twice to characteristic pattern during pond, and characteristic pattern exists
It is not to be aligned by pixel during scaling, so being had when being mapped to original image by the characteristic pattern of Chi Huahou larger
Deviation, cause target detection effect poor, this does not influence the classification of object, but influences very big (He to the segmentation of image
et al.,2018).Therefore crimping blade is detected using Mask R-CNN at us.And Mask R-CNN is used
RoIAlign replace RoI pooling to characteristic pattern carry out pondization operation, RoIAlign using two-wire interpolation method realize pixel and
Pixel alignment, characteristic pattern can be preferably mapped on original image in example cutting procedure, one 800 × 800 as shown in Figure 1
Corn image becomes 25 × 25 after reducing 32 times, corn image 665 × 665, be 20.78 after reducing 32 times ×
20.78, RoI pooling take 20.If being mapped to the deviation for having 24.96 pixels in original image.And RoIAlign uses two-wire
Property interpolation method, floor operation is not carried out to pixel, therefore be able to achieve the pixel alignment of original image and characteristic pattern.Mask predicted branches
Target is split using full convolutional network (Fully Convolutional Networks).
In model training we by the corn image data marked input Mask R-CNN model, image data set with
Machine is divided into training set and test set, wherein 80% is divided into training set, 20% is divided into test set.Pass through in the training process
Tensorflow visualization tool tensorboard checks the convergent of model.With test set pair after model training is good
Model is assessed.
1. crimping blade detection model of table judges whether it is the confusion matrix of curling
Crimping blade detection in this experiment can be considered identification of two classification problems i.e. to crimping blade and background.To two points
Class problem can be divided into positive example (Positive) and counter-example (Negetive), and as shown in table x, it is correct that TP, FP respectively represent classification
The positive example of positive class and classification error, FN and TN respectively represent the counter-example of classification error and correct counter-example of classifying.TP in equation 1
Correct positive example of classifying is represented, FP represents the positive example of classification error, and FN represents the counter-example of classification error in formula 2.F1_
Score is the harmonic-mean of precision and recall rate.When assessing model, we are in different IoU threshold values
Model is assessed under (Intersection over union), the mean value mean accuracy of computation model, precision, recall rate
And F1_score, IoU threshold value is respectively set to 0.5 and 0.75 in this experiment.IoU threshold value may be defined as crimping blade detection model
The true mark of intersection and curling of the middle true callout box of crimping blade (Ground truth) model prediction frame (Prediction)
The ratio of the union of frame and model prediction frame is infused, calculation formula is as shown in figure 4, wherein G represents the true frame of target object
Size (Ground truth), the target object frame size (Prediction) of P representative model prediction.
Fig. 5 is that crop leaf measuring model crimps crop leaf measuring and confidence level, as shown in Figure 5, Mask R-CNN model to corn
Preferably corn crimping blade can be detected, and model is all larger than 99% to the confidence level of crimping blade, part crimps
It is 100% that crimping blade, which detects confidence level,.As shown in Table 2, when IoU threshold value is 0.5, the average essence of the mean value of maize leaf detection
Degree is 74.35%, and accuracy, recall rate and F1_score are respectively 88.37%, 45.83% and 60.36.And IoU threshold value is
When 0.75, the mean value mean accuracy of model, accuracy, recall rate and F1_score are respectively 52.25%, 74.92%,
38.48% and 50.85%.Since in the case where other conditions are certain, maize leaf curling is caused by drought stress, therefore is examined
The plant for measuring crimping blade can determine that receive drought stress.
The precision that crop leaf measuring model detects crimping blade is crimped under 2. difference IoU threshold value of table to compare
The preferred embodiment of the present invention above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously
Limitations on the scope of the patent of the present invention therefore cannot be interpreted as.It should be pointed out that for those of ordinary skill in the art
For, without departing from the inventive concept of the premise, several deformations can also be made, improves and substitutes, these belong to this hair
Bright protection scope.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (2)
1. a kind of corn arid recognition methods based on crimping blade detection, which comprises the following steps:
A) test of corn control drought is carried out by pot-culture method, being divided into is suitable for handling with drought stress two；
B) the vertical view digital picture for obtaining corn at the top of corn using slr camera, establishes image data set, and use
Labelme software carries out how fine polygon to corn crimping blade in digital picture and marks, and data are divided into trained sample
Sheet and test sample, while json file is converted by image data；
C) using the training set image training objective detection model Mask R-CNN marked, train epochs are 50 steps, are used
ResNet101 extracts characteristics of image；
D) with test set test trained crimping blade detection model Mask R-CNN model, with mean value mean accuracy mAP,
Precision evaluates model, and detects the specific location of crimping blade in the picture, is split to crimping blade.
2. the method according to claim 1, wherein obtaining corn top view under drought stress using slr camera
Digital picture is labeled corn crimping blade with the polygon frame of Labelme software, and one is generated after every image labeling
Json file contains the information of image labeling in json file, mainly there is the position coordinates of target object, the class of target object
Distinguishing label and image size information carry out the annotation of figure after image labeling, according to the crimping blade mark in json file
Location information draws the bianry image of each crimping blade using OpenCV software respectively, finally according to the note of each crimping blade
One json file of format needed for releasing information and original image generation MaskR-CNN, this document contain the class of all images
Distinguishing label, the coordinate information of crimping blade.The corn image data marked is inputted into MaskR-CNN mould in model training
Type, image data set random division are training set and test set, wherein 80% is divided into training set, 20% is divided into test set,
The convergent for checking model by tensorflow visualization tool tensorboard in the training process, when model is instructed
It is assessed after perfecting with test the set pair analysis model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910446308.1A CN110188657A (en) | 2019-05-27 | 2019-05-27 | Corn arid recognition methods based on crimping blade detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910446308.1A CN110188657A (en) | 2019-05-27 | 2019-05-27 | Corn arid recognition methods based on crimping blade detection |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110188657A true CN110188657A (en) | 2019-08-30 |
Family
ID=67718000
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910446308.1A Pending CN110188657A (en) | 2019-05-27 | 2019-05-27 | Corn arid recognition methods based on crimping blade detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110188657A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111898641A (en) * | 2020-07-01 | 2020-11-06 | 中国建设银行股份有限公司 | Target model detection device, electronic equipment and computer readable storage medium |
| CN112270268A (en) * | 2020-10-29 | 2021-01-26 | 重庆邮电大学 | Fruit picking robot target detection method based on deep learning in unstructured environment |
| CN113283287A (en) * | 2021-04-06 | 2021-08-20 | 安徽农业大学 | CNN algorithm-based using method of corn disease control unmanned aerial vehicle |
| CN115953686A (en) * | 2023-02-07 | 2023-04-11 | 开封市农林科学研究院 | Peanut pest detection method and system based on image processing |
| CN117541924A (en) * | 2023-11-16 | 2024-02-09 | 海南海柠科技有限公司 | Crop growth condition monitoring method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107392892A (en) * | 2017-06-30 | 2017-11-24 | 天津大学 | Maize growth drought stress early stage automatic identifying method based on image |
| CN107894418A (en) * | 2017-10-31 | 2018-04-10 | 华中农业大学 | Field rice leaf roll degree measurement method based on machine vision |
| CN109086826A (en) * | 2018-08-06 | 2018-12-25 | 中国农业科学院农业资源与农业区划研究所 | Wheat Drought recognition methods based on picture depth study |
-
2019
- 2019-05-27 CN CN201910446308.1A patent/CN110188657A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107392892A (en) * | 2017-06-30 | 2017-11-24 | 天津大学 | Maize growth drought stress early stage automatic identifying method based on image |
| CN107894418A (en) * | 2017-10-31 | 2018-04-10 | 华中农业大学 | Field rice leaf roll degree measurement method based on machine vision |
| CN109086826A (en) * | 2018-08-06 | 2018-12-25 | 中国农业科学院农业资源与农业区划研究所 | Wheat Drought recognition methods based on picture depth study |
Non-Patent Citations (8)
| Title |
|---|
| JIANGYONG AN ET AL.: "Identification and Classification of Maize Drought Stress Using Deep Convolutional Neural Network", 《SYMMETRY》 * |
| 勇者归来: "Mask RCNN 简单使用", 《HTTPS://WWW.CNBLOGS.COM/WANGYONG/P/9305347.HTML》 * |
| 周建新 等: "适用于密集人群目标检测的多尺度检测方法", 《系统仿真学报》 * |
| 岳焕然 等: "基于颜色和纹理特征的玉米干旱识别", 《中国农学通报》 * |
| 熊丽婷 等: "基于搜索区域条件概率CNN的精确目标探测方法", 《计算机工程与应用》 * |
| 许庆勇 著: "《基于深度学习理论的纹身图像识别与检测研究》", 31 December 2018 * |
| 郭躬德 等著: "《近邻分类方法及其应用 上册》", 31 December 2013 * |
| 陆游飞 等: "一种结合图像信息的视频行人检测网络研究", 《杭州电子科技大学学报（自然科学版）》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111898641A (en) * | 2020-07-01 | 2020-11-06 | 中国建设银行股份有限公司 | Target model detection device, electronic equipment and computer readable storage medium |
| CN112270268A (en) * | 2020-10-29 | 2021-01-26 | 重庆邮电大学 | Fruit picking robot target detection method based on deep learning in unstructured environment |
| CN113283287A (en) * | 2021-04-06 | 2021-08-20 | 安徽农业大学 | CNN algorithm-based using method of corn disease control unmanned aerial vehicle |
| CN115953686A (en) * | 2023-02-07 | 2023-04-11 | 开封市农林科学研究院 | Peanut pest detection method and system based on image processing |
| CN115953686B (en) * | 2023-02-07 | 2023-07-28 | 开封市农林科学研究院 | Peanut pest detection method and system based on image processing |
| CN117541924A (en) * | 2023-11-16 | 2024-02-09 | 海南海柠科技有限公司 | Crop growth condition monitoring method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Anand et al. | An application of image processing techniques for detection of diseases on brinjal leaves using k-means clustering method | |
| CN110188657A (en) | Corn arid recognition methods based on crimping blade detection | |
| Malambo et al. | Automated detection and measurement of individual sorghum panicles using density-based clustering of terrestrial lidar data | |
| Liu et al. | Automated image-processing for counting seedlings in a wheat field | |
| Tellaeche et al. | A new vision-based approach to differential spraying in precision agriculture | |
| Liu et al. | A vision-based robust grape berry counting algorithm for fast calibration-free bunch weight estimation in the field | |
| Zhou et al. | An integrated skeleton extraction and pruning method for spatial recognition of maize seedlings in MGV and UAV remote images | |
| CN111462058B (en) | Method for rapidly detecting effective rice ears | |
| CN111339912A (en) | Method and system for recognizing cattle and sheep based on remote sensing image | |
| Ji et al. | In-field automatic detection of maize tassels using computer vision | |
| CN114818909A (en) | Weed detection method and device based on crop growth characteristics | |
| Lin et al. | Estimating litchi flower number using a multicolumn convolutional neural network based on a density map | |
| David et al. | Plant detection and counting from high-resolution RGB images acquired from UAVs: comparison between deep-learning and handcrafted methods with application to maize, sugar beet, and sunflower | |
| CN114821268A (en) | Weed and crop identification method based on machine learning | |
| Yan et al. | High-resolution mapping of paddy rice fields from unmanned airborne vehicle images using enhanced-TransUnet | |
| CN116778343B (en) | Target image feature extraction method for comprehensive identification | |
| Behera et al. | Image processing based detection & size estimation of fruit on mango tree canopies | |
| Wang et al. | Detection of corn plant population and row spacing using computer vision | |
| Li et al. | Image processing for crop/weed discrimination in fields with high weed pressure | |
| CN110544237A (en) | Oil tea pest model training method and recognition method based on image analysis | |
| CN114842326B (en) | Calibration-free sandalwood plant seedling missing positioning method | |
| de Ocampo et al. | Integrated Weed Estimation and Pest Damage Detection in Solanum melongena Plantation via Aerial Vision-based Proximal Sensing. | |
| Yihang et al. | Automatic recognition of rape seeding emergence stage based on computer vision technology | |
| Kalo et al. | Detection of irrigation malfunctions based on thermal imaging | |
| Rocha et al. | Detection of Curved Rows and Gaps in Aerial Images of Sugarcane Field Using Image Processing Techniques |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190830 |
|
| RJ01 | Rejection of invention patent application after publication |