CN114419468A - Paddy field segmentation method combining attention mechanism and spatial feature fusion algorithm - Google Patents
Paddy field segmentation method combining attention mechanism and spatial feature fusion algorithm Download PDFInfo
- Publication number
- CN114419468A CN114419468A CN202210091169.7A CN202210091169A CN114419468A CN 114419468 A CN114419468 A CN 114419468A CN 202210091169 A CN202210091169 A CN 202210091169A CN 114419468 A CN114419468 A CN 114419468A
- Authority
- CN
- China
- Prior art keywords
- paddy field
- semantic segmentation
- training
- network model
- attention mechanism
- 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
Images
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
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a paddy field segmentation method combining an attention mechanism and a spatial feature fusion algorithm, which comprises the following steps: constructing a low-altitude paddy field image data set, labeling data in the data set, and dividing the data into a training set and a verification set; building a semantic segmentation network model, and training the semantic segmentation network model based on the training set to obtain a trained semantic segmentation network model; and verifying the trained semantic segmentation network model based on the verification set to obtain a verification result. The method is more efficient and accurate in paddy field prediction results, provides important basis for further obtaining high-precision paddy field boundary positioning information and constructing high-precision maps of a plurality of paddy fields in a larger area, and plays a positive role in promoting efficient and accurate paddy field farmland informatization management.
Description
Technical Field
The invention relates to the technical field of intelligent agriculture, in particular to a paddy field segmentation method combining an attention mechanism and a spatial feature fusion algorithm.
Background
Wisdom agricultural is the must trend of modern agriculture development, understands the farmland environment better in order to help agricultural machine, realizes intelligent agricultural machinery operation, has the important effect to promoting wisdom agricultural development, and the arable land image data based on high resolution unmanned aerial vehicle low latitude remote sensing is mostly that plain area large tracts of land is drawed at present, has great degree of difficulty and the relatively lower condition of interpretation precision to the drawing of hills area paddy field type. The invention provides a deep neural network structure of SA-DeepLabv3+ combined with an attention mechanism and a spatial feature fusion algorithm, which is used for extracting paddy fields and land types in southern hilly areas.
Disclosure of Invention
The invention aims to help agricultural machinery to better understand farmland environment and realize intelligent agricultural machinery operation, and provides a paddy field segmentation method combining an attention mechanism and a spatial feature fusion algorithm.
In order to achieve the purpose, the invention provides the following scheme:
a paddy field segmentation method combining an attention mechanism and a spatial feature fusion algorithm comprises the following steps:
constructing a low-altitude paddy field image data set, labeling data in the data set, and dividing the data into a training set and a verification set;
building a semantic segmentation network model, and training the semantic segmentation network model based on the training set to obtain a trained semantic segmentation network model;
and verifying the trained semantic segmentation network model based on the verification set to obtain a verification result.
Preferably, the low-altitude paddy field image is acquired by using an unmanned aerial vehicle platform, and the low-altitude paddy field image data set is constructed by selecting the paddy field image with the image data width of 4864 pixels and the image data height of 3648 pixels.
Preferably, the collected low-altitude paddy field images are labeled by a labeling tool in a manual visual dotting mode to generate label graphs with the same format, and the label graphs are subjected to data processing to obtain a data format capable of being input into the semantic segmentation network model.
Preferably, the semantic segmentation network model is an SA-DeepLabv3+ semantic segmentation network model and comprises an encoder and a decoder, wherein the encoder comprises a convolutional layer, an Xception-65 feature extraction network, an attention mechanism scSE and a void space pyramid pooling ASPP module; the decoder comprises a feature map connection layer, a deconvolution layer and an Adaptive Spatial Feature Fusion (ASFF) module.
Preferably, the images are classified according to the shapes of paddy fields in the images and the types of surrounding species, the labeled image files are divided according to different types of the images, each type of image is randomly divided into a plurality of equal parts to obtain a training set and a verification set, the training set is input into the semantic segmentation network model for training, the average value of the results is taken as a training result, and the training result is verified based on the verification set.
Preferably, the process of inputting the training set into the semantic segmentation network model for training comprises:
setting training parameters of the semantic segmentation network based on a ReLu activation function;
training and testing the semantic segmentation network model by adopting a five-fold cross verification method;
setting a cross entropy loss function as a loss function of the semantic segmentation network model, calculating the loss number between the feature graph and the label graph, and optimizing network parameters by using a Momentum optimization algorithm and a Momentum factor-based random gradient descent algorithm.
Preferably, the method for calculating the loss number between the feature map and the label map is as follows:
wherein N is the sum of the number of pixels in a batch size, and m is the number of categories; y isijMarking the category j by the pixel point i;
and representing the probability that the pixel point i is in the category j.
Preferably, the process of verifying the training result based on the verification set comprises:
firstly, initializing a network by loading an Xchoice-65 weight pre-trained on the basis of a Pascal VOC 2012 public data set by adopting a transfer learning method, then carrying out scene analysis on the input paddy field data in a verification set, transforming a sample space into a feature space, extracting pixel features by using an encoder and a space pyramid of an Xchoice-65 backbone network added with a scSE module, restoring a feature map to the original size by using a decoder fused with an adaptive spatial feature fusion module ASFF, and outputting the feature map;
comparing the prediction probabilities of the same pixel point in the feature map under different categories, if the prediction probability of the paddy field block is high, judging the pixel as a paddy field pixel, otherwise, outputting a segmentation mask as a background pixel;
and verifying the segmentation mask output by the SA-DeepLabv3+ semantic segmentation model by adopting an image segmentation evaluation index mIoU average cross-over ratio to obtain a verification result.
Preferably, the calculation method for evaluating the segmentation mask output by the SA-deep bv3+ semantic segmentation model by using the average cross-over ratio mlou of image segmentation evaluation index is as follows:
FN is false negative, which means that the reality is 1 and the prediction is 0; TP is true, which means that the reality is 1, and the prediction is 1; FP is false positive, indicating a true 0, predicted to be 1.
The invention has the beneficial effects that:
the method is improved based on a DeepLabv3+ semantic segmentation network model, and the segmentation efficiency and the interpretation precision are improved to a certain extent by applying paddy field image segmentation; the intelligent degree is high, the universality is good, and the technical problems of low efficiency, long time consumption and poor accuracy of the conventional paddy field segmentation method are solved;
the method is more efficient and accurate in paddy field prediction results, provides important basis for further obtaining high-precision paddy field boundary positioning information and constructing high-precision maps of a plurality of paddy fields in a larger area, and plays a positive role in promoting efficient and accurate paddy field farmland informatization management.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of a method of an embodiment of the present invention;
FIG. 2 is a schematic diagram of a SA-DeepLabv3+ network structure according to an embodiment of the present invention;
FIG. 3 is a schematic view of the type 1 paddy field and the division effect according to the embodiment of the present invention;
FIG. 4 is a schematic view of the paddy field type 2 and the division effect according to the embodiment of the present invention;
FIG. 5 is a schematic view of a type 3 paddy field and a partition effect according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of the type 4 paddy field and the segmentation effect according to the embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
The invention provides a paddy field segmentation method combining an attention mechanism and a spatial feature fusion algorithm, which is shown in figure 1 and comprises the following steps:
step one, collecting and labeling low-altitude paddy field picture data;
s11, data acquisition: in this embodiment, a low-altitude farmland picture sample is acquired by using a professional version of the great-Xintom 4RTK unmanned aerial vehicle, the sample scenes are as many as possible, and the sample types include 4 types: type 1 paddy fields have irregular shapes and little vegetation in the field; type 2 paddy fields are irregular in shape and have multiple vegetations in the field; type 3, the paddy field is regular in shape and has no vegetation in the field; type 4 paddy fields are irregular in shape, have no vegetation in the field, and have forest belts outside the field.
In the embodiment, the collected picture is in a JPEG format, the selected width is 4864 pixels, and the selected height is 3648 pixels, because the paddy field boundary detection belongs to a relatively fine image processing work. If some problems are similar to the problems of weeds in ridges, the labeling work is troublesome due to the fact that the image pixels are too low, and the final training accuracy is affected.
S12, data annotation: performing pixel-level labeling on the side ridges of the paddy field by using a Labelme labeling tool in a manual visual dotting mode on an image sample, generating a json format file after labeling is completed, performing data processing on the json format file to generate a label graph, and arranging the label graph into a data format which can be input into a deep Labv3+ semantic segmentation network; because of obstacles such as weeds, trees, illumination shadows and telegraph poles exist in the unmanned aerial vehicle low-altitude collected paddy field image data, the paddy field block is shielded to different degrees, and all obstacles need to be bypassed during marking to the characteristic of the extraneous information outside the paddy field block is extracted in the characteristic extraction process, and then the detection effect of the algorithm on the paddy field area is influenced.
The process of processing the label graph comprises the following steps: saving the marked low-altitude paddy field picture as a json-format label file, and then converting the format of the json-format label file, wherein the specific operations are as follows: outputting the json file to a computer, compiling a Python programming language script, processing the json file input to the computer, wherein the processing process comprises the following steps: establishing a folder and naming the folder as 'field', wherein the folder named as 'field' is used for storing a label file of 'field', and an image of the folder is training data; processing the json file by using a labelme module package of Python, and analyzing a paddy field image sample and a json label graph corresponding to the paddy field image sample; and converting the PNG format of the analyzed paddy field image sample into the JPG format by using a Python PIL module, and adjusting the pixel width and height to 512, so that the size of the original image is properly reduced under the condition of not sacrificing details, and the training speed of a network in the later period can be accelerated.
Labelme labeling software is selected, labelme has the advantages that labeling areas can be freely circled according to needs, labels of paddy field fields are named as 'field', and labeled json files are output to a computer; selecting and writing Python as a programming language for processing json files input to a computer, wherein the processing steps are as follows:
1. the general folder is built and named as 'field', 4 sub folders are arranged below the 'field' folder and named as 'type 1', 'type 2', 'type 3' and 'type 4', and are respectively used for storing: irregular shape, and few vegetation field blocks in the field; irregular shapes, and a plurality of vegetation field blocks in the field; the shape is regular, and no vegetation field block exists in the field; irregular shape, no vegetation in the field, and forest and field blocks outside the field. The image of the folder is used for training the data of paddy field detection;
2. processing the json format file by using a labelme module package of Python, and analyzing the original image and the label graph;
3. the JPEG format of the original image is converted into the PNG format by a Python PIL module, the pixels are changed into 513 pixels in width and 513 pixels in height, the size of the original image is properly reduced under the condition of not sacrificing details, and the training speed of a later deep neural network is accelerated;
4. the pictures are named in a form of 'four-digit picture number', then 80% of the pictures are used as a training set, the rest 20% are used as a verification set, and the accuracy of the model is verified by adopting a five-fold cross-validation method. The picture names of the training set and the verification set are stored in a note file named "train.
Step two, dividing the data set and building an SA-deep Labv3+ semantic segmentation network model
S21, dividing a data set: the data types are divided into 4 types, and the types comprise: type 1 paddy fields are irregular in shape, and have little vegetation in the field (as shown in figure 3); type 2 paddy fields are irregular in shape, and have multiple vegetation in the field (as shown in figure 4); type 3 paddy fields are regular in shape, and have no vegetation in the field (as shown in figure 5); type 4 paddy fields have irregular shapes, no vegetation in the field, and forest belts outside the field (see fig. 6). 500 images of each type are obtained by randomly dividing each type of paddy field image data into 5 equal parts, and taking 4 parts of the paddy field image data as training data and 1 part of the paddy field image data as test data in turn.
S22, building an SA-DeepLabv3+ semantic segmentation network model (as shown in figure 2): the SA-DeepLabv3+ semantic segmentation model comprises: an encoder and a decoder, the encoder comprising: convolutional layer, Xception-65 feature extraction network, attention mechanism (scSE), and spatial pyramid module. The decoder includes: a feature map connection layer, an deconvolution layer, an Adaptive Spatial Feature Fusion (ASFF) module.
Step three, training an SA-deep Labv3+ semantic segmentation network by using the obtained paddy field semantic segmentation data set;
s31, setting semantic segmentation network training parameters and using a ReLu activation function;
s32, training the model by adopting a five-fold cross-validation method and testing the accuracy of the algorithm;
s33, setting a cross entropy loss function as a loss function of the model, and calculating the loss function between the characteristic diagram and the label diagram, wherein the calculation formula is as follows:
wherein N is the sum of the number of pixels in one batch size, and m is the number of categories, and this embodiment divides the paddy field image into a paddy field category and a background category, so m is 2; y isijRepresenting the label of the pixel point i to the category j, wherein the label is 1 if the pixel point i is a paddy field type, and the label is 0 if the pixel point i is a background type;
and representing the probability that the pixel point i is in the category j.
S34, optimizing network parameters by using a Momentum optimization algorithm, wherein the updating mode is poly, the size of an input image is 513 multiplied by 513, the number of batch processing data is set to be 8, and the total iteration number is 20000; performing parameter optimization by using a random gradient descent algorithm with a momentum factor of 0.9, setting an initial learning rate to be 0.001 and a weight attenuation rate to be 0.0004;
s35, training the SA-DeepLabv3+ semantic segmentation network.
Step four, verifying the training result based on the test data set and the trained model;
s41, firstly, an SA-DeepLabv3+ model adopts a transfer learning method, a network is initialized by loading an Xceptation-65 weight pre-trained on the basis of a Pascal VOC 2012 public data set, then scene analysis is carried out on an input paddy field data set, a sample space is converted into a feature space, pixel features are extracted by using an encoder and a space pyramid of the Xceptation-65 backbone network added with an scSE module, a feature map is restored to the original size by a decoder fused with an Adaptive Space Feature Fusion (ASFF) module, feature maps with the number of 2 channels are output, each channel represents different categories, channel 0 represents a background, and channel 1 represents a paddy field;
s42, comparing the prediction probabilities of the same pixel point under the two categories, if the prediction probability of the paddy field block is larger, judging the pixel as a paddy field pixel, otherwise, outputting a segmentation mask as a background pixel;
s43, evaluating the SA-DeepLabv3+ semantic segmentation model by adopting an image segmentation evaluation index mIoU (average cross-over ratio). Wherein, the calculation formula is as follows:
FN is false negative, which means that the reality is 1 and the prediction is 0; TP is true, which means that the reality is 1, and the prediction is 1; FP is false positive, indicating a true 0, predicted to be 1.
The traditional paddy field segmentation method comprises the technical means of manual visual dotting, random forest algorithm optimization of textural features and the like, and has the defects of long time consumption, low interpretation precision, resource waste, insufficient precision and the like when used for extracting paddy field land types; the method is improved based on a DeepLabv3+ semantic segmentation network model, and the segmentation efficiency and the interpretation precision are improved to a certain extent by applying paddy field image segmentation; the intelligent degree is high, the universality is good, and the technical problems of low efficiency, long time consumption and poor accuracy of the conventional paddy field segmentation method are solved;
the method is more efficient and accurate in paddy field prediction results, provides important basis for further obtaining high-precision paddy field boundary positioning information and constructing high-precision maps of a plurality of paddy fields in a larger area, and plays a positive role in promoting efficient and accurate paddy field farmland informatization management.
The above-described embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solutions of the present invention can be made by those skilled in the art without departing from the spirit of the present invention, and the technical solutions of the present invention are within the scope of the present invention defined by the claims.
Claims (9)
1. A paddy field segmentation method combining an attention mechanism and a spatial feature fusion algorithm is characterized by comprising the following steps:
constructing a low-altitude paddy field image data set, labeling data in the data set, and dividing the data into a training set and a verification set;
building a semantic segmentation network model, and training the semantic segmentation network model based on the training set to obtain a trained semantic segmentation network model;
and verifying the trained semantic segmentation network model based on the verification set to obtain a verification result.
2. The paddy field segmentation method combining the attention mechanism and the spatial feature fusion algorithm according to claim 1, wherein a low-altitude paddy field image is acquired by using an unmanned aerial vehicle platform, and a paddy field image with an image data width of 4864 pixels and a height of 3648 pixels is selected to construct the low-altitude paddy field image data set.
3. The paddy field segmentation method combining the attention mechanism and the spatial feature fusion algorithm according to claim 2, wherein the collected low-altitude paddy field images are labeled by a manual visual dotting mode based on a labeling tool to generate label graphs with the same format, and the label graphs are subjected to data processing to obtain a data format capable of being input into the semantic segmentation network model.
4. The paddy field segmentation method combining the attention mechanism and the spatial feature fusion algorithm as claimed in claim 3, wherein the semantic segmentation network model is a SA-DeepLabv3+ semantic segmentation network model, and comprises an encoder and a decoder, wherein the encoder comprises a convolution layer, an Xception-65 feature extraction network, an attention mechanism scSE, and a hole space pyramid pooling ASPP module; the decoder comprises a feature map connection layer, a deconvolution layer and an Adaptive Spatial Feature Fusion (ASFF) module.
5. The paddy field segmentation method combining the attention mechanism and the spatial feature fusion algorithm according to claim 4, wherein the images are classified according to the shape of the paddy field in the images and the types of the peripheral species, the labeled image files are divided according to different types of the images, each type of image is randomly divided into a plurality of equal parts to obtain a training set and a verification set, the training set is input into the semantic segmentation network model for training, the average value of the results is taken as a training result, and the training result is verified based on the verification set.
6. The paddy field segmentation method combining the attention mechanism and the spatial feature fusion algorithm according to claim 5, wherein the process of inputting the training set into the semantic segmentation network model for training comprises:
setting training parameters of the semantic segmentation network based on a ReLu activation function;
training and testing the semantic segmentation network model by adopting a five-fold cross verification method;
setting a cross entropy loss function as a loss function of the semantic segmentation network model, calculating the loss number between the feature graph and the label graph, and optimizing network parameters by using a Momentum optimization algorithm and a Momentum factor-based random gradient descent algorithm.
7. The paddy field segmentation method combining attention mechanism and spatial feature fusion algorithm as claimed in claim 6, wherein the method for calculating the loss number between the feature map and the label map is as follows:
wherein N is the sum of the number of pixels in a batch size, and m is the number of categories; y isijMarking the category j by the pixel point i;
and representing the probability that the pixel point i is in the category j.
8. The paddy field segmentation method combining attention mechanism and spatial feature fusion algorithm according to claim 5, wherein the process of verifying the training result based on the verification set comprises:
firstly, initializing a network by loading an Xchoice-65 weight pre-trained on the basis of a Pascal VOC 2012 public data set by adopting a transfer learning method, then carrying out scene analysis on the input paddy field data in a verification set, transforming a sample space into a feature space, extracting pixel features by using an encoder and a space pyramid of an Xchoice-65 backbone network added with a scSE module, restoring a feature map to the original size by using a decoder fused with an adaptive spatial feature fusion module ASFF, and outputting the feature map;
comparing the prediction probabilities of the same pixel point in the feature map under different categories, if the prediction probability of the paddy field block is high, judging the pixel as a paddy field pixel, otherwise, outputting a segmentation mask as a background pixel;
and verifying the segmentation mask output by the SA-DeepLabv3+ semantic segmentation model by adopting an image segmentation evaluation index mIoU average cross-over ratio to obtain a verification result.
9. The paddy field segmentation method combining the attention mechanism and the spatial feature fusion algorithm as claimed in claim 8, wherein the computation method for evaluating the segmentation mask output by the SA-deep bv3+ semantic segmentation model by using an image segmentation evaluation index average cross ratio mlou is as follows:
FN is false negative, which means that the reality is 1 and the prediction is 0; TP is true, which means that the reality is 1, and the prediction is 1; FP is false positive, indicating a true 0, predicted to be 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210091169.7A CN114419468A (en) | 2022-01-26 | 2022-01-26 | Paddy field segmentation method combining attention mechanism and spatial feature fusion algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210091169.7A CN114419468A (en) | 2022-01-26 | 2022-01-26 | Paddy field segmentation method combining attention mechanism and spatial feature fusion algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114419468A true CN114419468A (en) | 2022-04-29 |
Family
ID=81277807
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210091169.7A Pending CN114419468A (en) | 2022-01-26 | 2022-01-26 | Paddy field segmentation method combining attention mechanism and spatial feature fusion algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114419468A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114662621A (en) * | 2022-05-24 | 2022-06-24 | 灵枭科技(武汉)有限公司 | Agricultural machinery working area calculation method and system based on machine learning |
| CN114723760A (en) * | 2022-05-19 | 2022-07-08 | 北京世纪好未来教育科技有限公司 | Portrait segmentation model training method and device and portrait segmentation method and device |
| CN115222946A (en) * | 2022-09-19 | 2022-10-21 | 南京信息工程大学 | Single-stage example image segmentation method and device and computer equipment |
| CN116630650A (en) * | 2023-07-20 | 2023-08-22 | 武汉理工大学 | Ridge width identification method and unmanned vehicle device for ridges |
-
2022
- 2022-01-26 CN CN202210091169.7A patent/CN114419468A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114723760A (en) * | 2022-05-19 | 2022-07-08 | 北京世纪好未来教育科技有限公司 | Portrait segmentation model training method and device and portrait segmentation method and device |
| CN114662621A (en) * | 2022-05-24 | 2022-06-24 | 灵枭科技(武汉)有限公司 | Agricultural machinery working area calculation method and system based on machine learning |
| CN115222946A (en) * | 2022-09-19 | 2022-10-21 | 南京信息工程大学 | Single-stage example image segmentation method and device and computer equipment |
| CN115222946B (en) * | 2022-09-19 | 2022-11-25 | 南京信息工程大学 | Single-stage instance image segmentation method and device and computer equipment |
| CN116630650A (en) * | 2023-07-20 | 2023-08-22 | 武汉理工大学 | Ridge width identification method and unmanned vehicle device for ridges |
| CN116630650B (en) * | 2023-07-20 | 2023-09-29 | 武汉理工大学 | Ridge width identification method and unmanned vehicle device for ridges |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113449594B (en) | Multilayer network combined remote sensing image ground semantic segmentation and area calculation method | |
| CN110245709B (en) | 3D point cloud data semantic segmentation method based on deep learning and self-attention | |
| CN114419468A (en) | Paddy field segmentation method combining attention mechanism and spatial feature fusion algorithm | |
| CN106909924B (en) | Remote sensing image rapid retrieval method based on depth significance | |
| CN112347970B (en) | Remote sensing image ground object identification method based on graph convolution neural network | |
| CN112232371B (en) | American license plate recognition method based on YOLOv3 and text recognition | |
| CN113705641B (en) | Hyperspectral image classification method based on rich context network | |
| CN112016601B (en) | Network model construction method based on knowledge graph enhanced small sample visual classification | |
| CN113988147B (en) | Multi-label classification method and device for remote sensing image scene based on graph network, and multi-label retrieval method and device | |
| CN114399686A (en) | Remote sensing image ground feature identification and classification method and device based on weak supervised learning | |
| CN111461129B (en) | Context prior-based scene segmentation method and system | |
| CN116206185A (en) | Lightweight small target detection method based on improved YOLOv7 | |
| CN112613548B (en) | User customized target detection method, system and storage medium based on weak supervised learning | |
| CN113378706B (en) | Drawing system for assisting children in observing plants and learning biological diversity | |
| CN114418021B (en) | Model optimization method, device and computer program product | |
| CN115497006B (en) | Urban remote sensing image change depth monitoring method and system based on dynamic mixing strategy | |
| CN113435254A (en) | Sentinel second image-based farmland deep learning extraction method | |
| CN113111716A (en) | Remote sensing image semi-automatic labeling method and device based on deep learning | |
| CN116186594A (en) | Method for realizing intelligent detection of environment change trend based on decision network combined with big data | |
| CN111598460A (en) | Method, device and equipment for monitoring heavy metal content in soil and storage medium | |
| CN109657082A (en) | Remote sensing images multi-tag search method and system based on full convolutional neural networks | |
| CN111242028A (en) | Remote sensing image ground object segmentation method based on U-Net | |
| Sun et al. | Automatic building age prediction from street view images | |
| CN112560925A (en) | Complex scene target detection data set construction method and system | |
| CN117079195A (en) | Wild animal identification method and system based on image video |
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 |