CN112036437A - Rice seedling detection model based on improved YOLOV3 network and method thereof - Google Patents
Rice seedling detection model based on improved YOLOV3 network and method thereof Download PDFInfo
- Publication number
- CN112036437A CN112036437A CN202010736191.3A CN202010736191A CN112036437A CN 112036437 A CN112036437 A CN 112036437A CN 202010736191 A CN202010736191 A CN 202010736191A CN 112036437 A CN112036437 A CN 112036437A
- Authority
- CN
- China
- Prior art keywords
- scale
- seedling
- feature
- fusion
- feature extraction
- 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
- 235000007164 Oryza sativa Nutrition 0.000 title claims abstract description 74
- 235000009566 rice Nutrition 0.000 title claims abstract description 74
- 238000001514 detection method Methods 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title abstract description 10
- 240000007594 Oryza sativa Species 0.000 title 1
- 238000000605 extraction Methods 0.000 claims abstract description 138
- 241000209094 Oryza Species 0.000 claims abstract description 73
- 238000012545 processing Methods 0.000 claims abstract description 10
- 230000004927 fusion Effects 0.000 claims description 100
- 238000005070 sampling Methods 0.000 claims description 49
- 238000010276 construction Methods 0.000 claims description 36
- 238000010586 diagram Methods 0.000 claims description 24
- 238000012360 testing method Methods 0.000 claims description 4
- 238000007499 fusion processing Methods 0.000 claims description 2
- 241000196324 Embryophyta Species 0.000 abstract description 6
- 238000005286 illumination Methods 0.000 abstract description 2
- 238000003672 processing method Methods 0.000 abstract description 2
- 230000000694 effects Effects 0.000 abstract 1
- 238000009333 weeding Methods 0.000 description 8
- 239000013043 chemical agent Substances 0.000 description 3
- 235000013305 food Nutrition 0.000 description 2
- 241000209140 Triticum Species 0.000 description 1
- 235000021307 Triticum Nutrition 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 235000013339 cereals Nutrition 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 230000004720 fertilization Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000000447 pesticide residue Substances 0.000 description 1
- 239000007921 spray Substances 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (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)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a rice seedling detection model based on an improved YOLOV3 network and a method thereof, which solve the problems of poor adaptability and robustness of the traditional image processing method to the rice seedling detection and poor image processing algorithm effect; the technical scheme is characterized by comprising a feature extraction module for performing multi-scale feature extraction on an input rice seedling image and a multi-scale prediction module for predicting the position of the rice seedling according to a feature map of the multi-scale feature extraction; the invention discloses a rice seedling detection model based on an improved YOLOV3 network and a method thereof, and solves the problems of poor robustness and accuracy of rice seedling detection in complex paddy field environments, such as high and low weed density, different illumination conditions and seedling loss.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a rice seedling detection model based on an improved YOLOV3 network and a method thereof.
Background
Rice is one of three major food crops in the world, the cultivation area and the total yield of the rice are second to wheat, and the rice is staple food for more than half of the population in the world. The weeding mode that paddy field adopted at present mainly is to spray chemical agent and artifical weeding, and chemical agent's use causes easily that the pesticide residue volume exceeds standard, environmental pollution and ecological chain's destruction scheduling problem, and artifical weeding is consuming time hard, inefficiency and with high costs. Therefore, the demand for mechanized weeding of rice is more and more urgent. The mechanized weeding technology for the rice has great significance for reducing the use of chemical agents, reducing the pollution of ecological environment and improving the grain safety.
The rice seedling detection is one of key links of rice mechanized weeding, and has important guiding function for accurate weeding and fertilization in accurate agriculture. The position information of the rice seedlings is obtained through the detection of the rice seedlings, and the inter-row weeds before the row closing of the rice can be removed by guiding the action of the weeding mechanism.
Crop and weed detection is mainly classified into 2 methods, the first method is based on image processing, the crop is segmented from the background, and then the crop row center line detection is carried out. The second category is crop and weed detection based on deep learning methods. Because the paddy field environment is very complex, the detection influence factors of the rice seedlings are more, and the traditional image processing method has poor adaptability and robustness to the detection of the rice seedlings. Especially when the paddy field image has insignificant variation in R, G and B pixel intensities, the conventional image processing algorithm is not effective when there are many weeds or similar to the size of crops.
Disclosure of Invention
The invention aims to provide a rice seedling detection model based on an improved YOLOV3 network and a method thereof, which solve the problem of poor robustness and accuracy of rice seedling detection under various conditions in a complex paddy field environment.
The technical purpose of the invention is realized by the following technical scheme:
a rice seedling detection model based on an improved YOLOV3 network comprises a feature extraction module for performing multi-scale feature extraction on an input rice seedling image to obtain a multi-scale seedling feature map, and a multi-scale prediction module for predicting the position of a rice seedling according to the multi-scale seedling feature map;
the multi-scale prediction module comprises a multi-scale fusion characteristic construction module for performing fusion construction processing on the multi-scale seedling characteristic diagram to obtain a multi-scale fusion characteristic diagram, and a multi-scale seedling position prediction module for predicting the seedling position according to the corresponding fusion characteristic in the fusion characteristic diagram.
Preferably, the feature extraction module comprises a 52-scale feature extraction submodule, a 26-scale feature extraction submodule and a 13-scale feature extraction submodule which are used for sequentially carrying out magnification sampling extraction processing on the input rice seedling image so as to respectively obtain a 52-scale seedling feature map, a 26-scale seedling feature map and a 13-scale seedling feature map.
Preferably, the multi-scale fusion feature construction module comprises a 52-scale fusion feature construction submodule for performing fusion processing on seedling feature maps with different scales, a 26-scale fusion feature construction submodule and a 13-scale fusion feature construction submodule.
Preferably, the 52-scale feature extraction sub-module performs feature extraction on the input seedling feature map of the 52-scale feature extraction sub-module to obtain a 52-scale seedling feature map containing 52-scale seedling features; the 26-scale feature extraction sub-module performs feature extraction on the input seedling feature map of the 26-scale feature extraction sub-module to obtain a 26-scale seedling feature map containing 26-scale seedling features; and the 13-scale feature extraction sub-module performs feature extraction on the input seedling feature map of the 13-scale feature extraction sub-module to obtain a 13-scale seedling feature map containing 13-scale seedling features.
A detection method of a rice seedling detection model based on an improved YOLOV3 network comprises the following steps:
acquiring an input rice seedling image;
respectively carrying out down-sampling with different multiplying powers on the rice seedling images, and obtaining a 52-scale seedling characteristic diagram, a 26-scale seedling characteristic diagram and a 13-scale seedling characteristic diagram through a characteristic extraction module;
sequentially performing feature fusion on the obtained 13-scale seedling feature map, 26-scale seedling feature map and 52-scale seedling feature map through a multi-scale fusion feature construction module to obtain a 13-scale seedling fusion feature map, a 26-scale seedling fusion feature map and a 52-scale seedling fusion feature map;
and predicting the obtained 13-scale seedling fusion characteristic graph, 26-scale seedling fusion characteristic graph and 52-scale seedling fusion characteristic graph in sequence through a multi-scale seedling position prediction module to obtain a 13-scale seedling prediction result, a 26-scale seedling prediction result and a 52-scale seedling prediction result.
Preferably, the specific steps of the feature extraction module for feature extraction are as follows:
the input convolution layer of the 52-scale feature extraction submodule is obtained by passing the input rice seedling graph through 10 convolution layers and 3 residual error layers, and 52-scale seedling features are obtained by extraction of the 52-scale feature extraction submodule;
convolving the output layer of the 52-scale feature extraction submodule by a convolution kernel of 3 × 3/2 × 512 to obtain an input convolution layer of the 26-scale feature extraction submodule, and extracting by the 26-scale feature extraction submodule to obtain 26-scale seedling features;
and (3) convolving the output layer of the 26-scale feature extraction submodule by a convolution kernel of 3 × 3/2 × 512 to obtain an input convolution layer of the 13-scale feature extraction submodule, and extracting by the 13-scale feature extraction submodule to obtain 13-scale seedling features.
Preferably, the multi-scale fusion feature construction module performs feature fusion construction by the following specific steps:
carrying out down-sampling on the output layer of the 26-scale feature extraction sub-module to obtain a 13 x 512 feature map, and carrying out tensor splicing on the feature map and the output layer of the 13-scale feature extraction sub-module to obtain a 13-scale seedling fusion feature map;
carrying out down-sampling on an output layer of the 52-scale feature extraction sub-module to obtain 26 × 256 feature maps, carrying out up-sampling on an output layer of the 13-scale feature extraction sub-module to obtain 26 × 512 feature maps, and carrying out tensor splicing on the 26 × 256 feature maps obtained by down-sampling, the 26 × 512 feature maps obtained by up-sampling and the output layer of the 26-scale feature extraction sub-module to obtain 26-scale seedling fusion feature maps;
and carrying out down-sampling on 104-scale output seedling feature maps obtained by carrying out feature extraction on the input rice seedling pictures to obtain 52 x 128 feature maps, carrying out up-sampling on output layers of the 26-scale feature extraction sub-modules to obtain 52 x 512 feature maps, and carrying out tensor splicing on the 52 x 128 feature maps obtained by down-sampling, the 52 x 512 feature maps obtained by up-sampling and the output layers of the 52-scale feature extraction sub-modules to obtain 52-scale seedling fusion feature maps.
Preferably, the specific prediction steps of the multi-scale seedling position prediction model are as follows:
the 13-scale seedling fusion features constructed by the 13-scale fusion feature construction submodule are subjected to convolution of a set of convolution kernels, 13 × 512 convolution kernels and 1 × 18 convolution kernel to obtain a 13-scale prediction result 13 × 18;
the 26-scale seedling fusion features constructed by the 26-scale fusion feature construction submodule are subjected to convolution of a set of convolution kernels, 13 × 512 convolution kernels and 1 × 18 convolution kernel to obtain a 26-scale prediction result 26 × 18;
the 52-scale seedling fusion features constructed by the 52-scale fusion feature construction submodule are subjected to convolution of a set of convolution kernels, 13 × 256 convolution kernel and 1 × 18 convolution kernel to obtain a 52-scale prediction result 52 × 18.
In conclusion, the invention has the following beneficial effects:
the rice seedling detection model based on the improved Yolov3 network can automatically extract rice seedling characteristics, the model adopts a 52-scale characteristic extraction submodule to replace an original residual error unit to widen the network on a 52-scale characteristic graph close to the rice seedling size in an original characteristic extraction network, sampling characteristics with different scales are added when fusion characteristics of all scales are constructed, the combination of position information of a low layer and semantic information of a high layer is added to the fused characteristics, the obtained rice seedling fusion characteristic levels of all scales are richer, and the improved rice seedling detection model has better accuracy and robustness on high and low weed density distribution in a complex paddy field environment, different illumination conditions and rice seedling detection under the condition of rice seedling loss.
Drawings
FIG. 1 is a block diagram schematically illustrating the structure of a detection model;
FIG. 2 is a network architecture diagram of a detection model;
FIG. 3 is a schematic flow chart of the detection method of the present invention;
FIG. 4 is a test p-r plot based on the modified Yolov3 and Yolov3 rice seedling detection models.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
According to one or more embodiments, a rice seedling detection model based on an improved YOLOV3 network is disclosed, as shown in fig. 1, which includes a feature extraction module for performing multi-scale rice seedling feature extraction on an input rice seedling image to obtain a multi-scale seedling feature map, and a multi-scale prediction module for predicting the position of a rice seedling according to the multi-scale seedling feature map extracted by the feature extraction module.
As shown in fig. 1, the feature extraction module includes a 52-scale feature extraction submodule, a 26-scale feature extraction submodule, and a 13-scale feature extraction submodule, which sequentially perform magnification sampling extraction processing on an input rice seedling image. For clarity, the rice seedling image input by way of example is an image of 416 × 416, and when sampling extraction is performed, 8-time, 16-time and 32-time down-sampling are performed respectively.
The multi-scale prediction module comprises a multi-scale fusion characteristic construction module for performing fusion construction processing on seedling characteristic graphs of different scales extracted by the characteristic extraction module to obtain a multi-scale fusion characteristic graph, and a multi-scale seedling position prediction module for predicting the seedling position according to the multi-scale fusion characteristic in the fusion characteristic graph. The multi-scale fusion feature construction module specifically comprises a 52-scale fusion feature construction submodule, a 26-scale fusion feature construction submodule and a 13-scale fusion feature construction submodule, and the multi-scale seedling position prediction module specifically comprises a 52-scale position prediction submodule, a 26-scale position prediction submodule and a 13-scale position prediction submodule.
The input seedling characteristic diagram extracted by the 52-scale characteristic extraction submodule specifically comprises: and (3) passing the input rice seedling image through 10 convolution layers and 3 residual layers to obtain an input convolution layer of the 52-scale feature extraction submodule as an input rice seedling feature map of the 52-scale feature extraction submodule. And the 52-scale feature extraction sub-module performs feature extraction on the input seedling feature map to obtain a 52-scale seedling feature map containing 52-scale seedling features.
The input seedling characteristic diagram extracted by the 26-scale characteristic extraction submodule specifically comprises: and (3) convolving the output layer of the 52-scale feature extraction submodule, namely the output 52-scale seedling feature map, by the convolution kernel of 3 × 3/2 × 512 to obtain the input convolution layer of the 26-scale feature extraction submodule, so as to obtain the input seedling feature map of the 26-scale feature extraction submodule. And the 26-scale feature extraction sub-module performs feature extraction on the input seedling feature map to obtain a 26-scale seedling feature map containing 26-scale seedling features.
The input seedling characteristic diagram extracted by the 13-scale characteristic extraction submodule specifically comprises: and (3) convolving the output layer of the 26-scale feature extraction submodule, namely the output 26-scale seedling feature map, by using a convolution kernel of 3 × 3/2 × 512 to obtain an input convolution layer of the 13-scale feature extraction submodule, so as to obtain the input seedling feature map of the 13-scale feature extraction submodule. And extracting a 13-scale seedling feature map containing 13-scale seedling features by using a 13-scale feature extraction submodule. And the feature extraction modules of the three scales sequentially extract.
The feature map constructed by the fusion of the 13-scale fusion feature construction submodule comprises a 13-scale seedling feature map obtained by extraction of an output layer of the 13-scale feature extraction submodule, a 13 x 512 feature map obtained by down-sampling of a 26-scale seedling feature map obtained by extraction of an output layer of the 26-scale feature extraction submodule, and the 13-scale seedling feature map and the 13 x 512 feature map obtained by tensor splicing are subjected to tensor splicing to obtain the 13-scale seedling fusion feature map.
Similarly, the seedling feature map constructed by the 26-scale fusion feature construction submodule in a fusion mode comprises an output layer of the 26-scale feature extraction submodule, the output layer of the 52-scale feature extraction submodule, namely the 52-scale seedling feature map obtained by extraction, is subjected to down-sampling to obtain 26 x 256 feature maps, the output layer of the 13-scale feature extraction submodule is subjected to up-sampling to obtain 26 x 512 feature maps, and the 26 x 256 feature maps obtained by down-sampling, the 26 x 512 feature maps obtained by up-sampling and the output layer of the 26-scale feature extraction submodule are spliced to obtain the 26-scale seedling fusion feature map.
Similarly, the seedling feature map constructed by the fusion of the 52-scale fusion feature construction submodule comprises an output layer of the 52-scale feature extraction submodule, a 104-scale output seedling feature map obtained by performing feature extraction on the input rice seedling picture is subjected to down sampling to obtain a 52 x 128 feature map, an output layer of the 26-scale feature extraction submodule is subjected to up sampling to obtain a 52 x 512 feature map, and the 52 x 128 feature map obtained by down sampling, a 52 x 512 feature map obtained by up sampling and an output layer of the 52-scale feature extraction submodule are spliced to obtain a 52-scale seedling fusion feature map.
The 13-scale location prediction sub-module of the multi-scale seedling location prediction module obtains 13-scale prediction results 13 x 18 by convolving the 13-scale seedling fusion features with 13 x 512 convolution kernel and 1 x 18 convolution kernel through a set of convolution kernels.
The 26-scale location prediction sub-module obtains 26 x 18 prediction results at 26 scales by convolving the 26-scale seedling fusion features with 13 x 512 convolution kernel and 1 x 18 convolution kernel.
The 52-scale location prediction sub-module yields 52-scale prediction results 52 x 18 by convolving the 52-scale seedling fusion features with 13 x 256 convolution kernel and 1 x 18 convolution kernel.
The original Yolov3 network model adopts a mode of sampling on a small-scale feature map and fusing the small-scale feature map and the scale feature map when each scale feature is fused, the invention adopts a mode of sampling on the small-scale feature map, sampling on a large-scale feature map and fusing the scale feature map based on the improved Yolov3 model, the fused features increase the combination of the position information of a lower layer and the semantic information of a higher layer, and the obtained rice seedling fusion feature levels of all scales are richer.
According to one or more embodiments, a method for detecting a rice seedling detection model based on an improved YOLOV3 network is disclosed, as shown in fig. 2 and 3, and comprises the following steps:
acquiring an input rice seedling image;
respectively carrying out down-sampling with different multiplying powers on the rice seedling images, and obtaining a 52-scale seedling characteristic diagram, a 26-scale seedling characteristic diagram and a 13-scale seedling characteristic diagram through a characteristic extraction module;
sequentially performing feature fusion on the obtained 13-scale seedling feature map, 26-scale seedling feature map and 52-scale seedling feature map through a multi-scale fusion feature construction module to obtain a 13-scale seedling fusion feature map, a 26-scale seedling fusion feature map and a 52-scale seedling fusion feature map;
and predicting the obtained 13-scale seedling fusion characteristic graph, 26-scale seedling fusion characteristic graph and 52-scale seedling fusion characteristic graph in sequence through a multi-scale seedling position prediction module to obtain a 13-scale seedling prediction result, a 26-scale seedling prediction result and a 52-scale seedling prediction result.
Specifically, the specific steps of the feature extraction module for feature extraction are as follows:
the input convolution layer of the 52-scale feature extraction submodule is obtained by passing the input rice seedling graph through 10 convolution layers and 3 residual error layers, and 52-scale seedling features are obtained by extraction of the 52-scale feature extraction submodule;
convolving the output layer of the 52-scale feature extraction submodule by a convolution kernel of 3 × 3/2 × 512 to obtain an input convolution layer of the 26-scale feature extraction submodule, and extracting by the 26-scale feature extraction submodule to obtain 26-scale seedling features;
and (3) convolving the output layer of the 26-scale feature extraction submodule by a convolution kernel of 3 × 3/2 × 512 to obtain an input convolution layer of the 13-scale feature extraction submodule, and extracting by the 13-scale feature extraction submodule to obtain 13-scale seedling features.
Specifically, the multi-scale fusion feature construction module performs feature fusion construction specifically including the following steps:
carrying out down-sampling on the output layer of the 26-scale feature extraction sub-module to obtain a 13 x 512 feature map, and carrying out tensor splicing on the feature map and the output layer of the 13-scale feature extraction sub-module to obtain a 13-scale seedling fusion feature map;
carrying out down-sampling on an output layer of the 52-scale feature extraction sub-module to obtain 26 × 256 feature maps, carrying out up-sampling on an output layer of the 13-scale feature extraction sub-module to obtain 26 × 512 feature maps, and carrying out tensor splicing on the 26 × 256 feature maps obtained by down-sampling, the 26 × 512 feature maps obtained by up-sampling and the output layer of the 26-scale feature extraction sub-module to obtain 26-scale seedling fusion feature maps;
and carrying out down-sampling on 104-scale output seedling feature maps obtained by carrying out feature extraction on the input rice seedling pictures to obtain 52 x 128 feature maps, carrying out up-sampling on output layers of the 26-scale feature extraction sub-modules to obtain 52 x 512 feature maps, and carrying out tensor splicing on the 52 x 128 feature maps obtained by down-sampling, the 52 x 512 feature maps obtained by up-sampling and the output layers of the 52-scale feature extraction sub-modules to obtain 52-scale seedling fusion feature maps.
The 52-scale feature extraction sub-module is an inclusion module, and a three-way convolution broadening network is adopted: the first convolution reduces the number of channels of the input seedling feature map by adopting 1 × 64 convolution kernels, and then convolution extraction features are carried out by adopting 3 × 128 and 3 × 64 convolution kernels; in the second path, 1 × 64 convolution kernels are adopted to reduce the number of channels of the input seedling characteristic graph, and then 3 × 128 convolution kernels are adopted to perform convolution to extract characteristics; the third path reduces the number of paths for inputting the seedling characteristic diagram by 1 x 64 convolution. And carrying out tensor splicing on the 52-scale seedling feature map obtained after the three-way convolution processing by the inclusion module and the 52 x 128 feature map and the 52 x 512 feature map obtained by up/down sampling to complete 52-scale feature fusion. The Inception modules are preferably set to be 4, the original residual error unit is replaced to widen the network, more features can be extracted, and parameters and training time can be effectively reduced.
Further, the specific prediction steps of the multi-scale seedling position prediction model are as follows:
performing convolution on the 13-scale seedling fusion features of the 13-scale seedling fusion feature map through a group of convolution kernels, 1 convolution kernel of 3 × 512 and 1 convolution kernel of 1 × 18 to obtain a prediction result of 13 × 18 of the 13 scales;
performing convolution on the 26-scale seedling fusion features of the 26-scale seedling fusion feature map through a group of convolution kernels, 1 convolution kernel of 3 × 512 and 1 convolution kernel of 1 × 18 to obtain a 26-scale prediction result of 26 × 18;
and (3) performing convolution on the 52-scale seedling fusion features of the 52-scale seedling fusion feature map through a group of convolution kernels, 1 convolution kernel of 3 × 256 and 1 convolution kernel of 1 × 18 to obtain a 52-scale prediction result 52 × 18.
For clarity, taking an example, the rice seedling detection models based on the improved Yolov3 and Yolov3 were tested using a rice seedling test set, and after the test, p-r curves were plotted, as shown in fig. 4(a), (b), (a) being a p-r curve of the improved Yolov3 and (b) being a p-r curve of the Yolov3, when the confidence threshold was 0.5, the accuracies of the rice seedling detection models based on the improved Yolov3 and Yolov3 networks were 0.82 and 0.48, respectively, so that the rice seedling detection algorithm based on the improved Yolov3 was improved in accuracy by about 34%.
The present embodiment is only for explaining the present invention, and it is not limited to the present invention, and those skilled in the art can make modifications of the present embodiment without inventive contribution as needed after reading the present specification, but all of them are protected by patent law within the scope of the claims of the present invention.
Claims (8)
1. A rice seedling detection model based on an improved YOLOV3 network is characterized in that: the system comprises a feature extraction module for performing multi-scale feature extraction on an input rice seedling image to obtain a multi-scale seedling feature map, and a multi-scale prediction module for predicting the position of a rice seedling according to the multi-scale seedling feature map;
the multi-scale prediction module comprises a multi-scale fusion characteristic construction module for performing fusion construction processing on the multi-scale seedling characteristic diagram to obtain a multi-scale fusion characteristic diagram, and a multi-scale seedling position prediction module for predicting the seedling position according to the corresponding fusion characteristic in the fusion characteristic diagram.
2. The improved YOLOV3 network-based rice seedling detection model of claim 1, wherein: the characteristic extraction module comprises a 52-scale characteristic extraction submodule, a 26-scale characteristic extraction submodule and a 13-scale characteristic extraction submodule which are used for sequentially carrying out multiplying factor sampling extraction processing on the input rice seedling image so as to respectively obtain a 52-scale seedling characteristic diagram, a 26-scale seedling characteristic diagram and a 13-scale seedling characteristic diagram.
3. The improved YOLOV3 network-based rice seedling detection model of claim 2, wherein: the multi-scale fusion feature construction module comprises a 52-scale fusion feature construction submodule for performing fusion processing on seedling feature maps with different scales, a 26-scale fusion feature construction submodule and a 13-scale fusion feature construction submodule.
4. The rice seedling detection model based on the improved YOLOV3 network as claimed in claim 3, wherein: the 52-scale feature extraction sub-module performs feature extraction on the input seedling feature map of the 52-scale feature extraction sub-module to obtain a 52-scale seedling feature map containing 52-scale seedling features; the 26-scale feature extraction sub-module performs feature extraction on the input seedling feature map of the 26-scale feature extraction sub-module to obtain a 26-scale seedling feature map containing 26-scale seedling features; and the 13-scale feature extraction sub-module performs feature extraction on the input seedling feature map of the 13-scale feature extraction sub-module to obtain a 13-scale seedling feature map containing 13-scale seedling features.
5. A detection method of a rice seedling detection model based on an improved YOLOV3 network is characterized by comprising the following steps:
acquiring an input rice seedling image;
respectively carrying out down-sampling with different multiplying powers on the rice seedling images, and obtaining a 52-scale seedling characteristic diagram, a 26-scale seedling characteristic diagram and a 13-scale seedling characteristic diagram through a characteristic extraction module;
sequentially performing feature fusion on the obtained 13-scale seedling feature map, 26-scale seedling feature map and 52-scale seedling feature map through a multi-scale fusion feature construction module to obtain a 13-scale seedling fusion feature map, a 26-scale seedling fusion feature map and a 52-scale seedling fusion feature map;
and predicting the obtained 13-scale seedling fusion characteristic graph, 26-scale seedling fusion characteristic graph and 52-scale seedling fusion characteristic graph in sequence through a multi-scale seedling position prediction module to obtain a 13-scale seedling prediction result, a 26-scale seedling prediction result and a 52-scale seedling prediction result.
6. The improved YOLOV3 network-based rice seedling detection method as claimed in claim 5, wherein the feature extraction module performs the following steps:
the input convolution layer of the 52-scale feature extraction submodule is obtained by passing the input rice seedling image through 10 convolution layers and 3 residual error layers, and the 52-scale seedling features are obtained by extraction of the 52-scale feature extraction submodule;
convolving the output layer of the 52-scale feature extraction submodule by a convolution kernel of 3 × 3/2 × 512 to obtain an input convolution layer of the 26-scale feature extraction submodule, and extracting by the 26-scale feature extraction submodule to obtain 26-scale seedling features;
and (3) convolving the output layer of the 26-scale feature extraction submodule by a convolution kernel of 3 × 3/2 × 512 to obtain an input convolution layer of the 13-scale feature extraction submodule, and extracting by the 13-scale feature extraction submodule to obtain 13-scale seedling features.
7. The improved YOLOV3 network-based rice seedling detection method as claimed in claim 6, wherein the multi-scale fusion feature construction module performs the specific steps of feature fusion construction as follows:
carrying out down-sampling on the output layer of the 26-scale feature extraction sub-module to obtain a 13 x 512 feature map, and carrying out tensor splicing on the feature map and the output layer of the 13-scale feature extraction sub-module to obtain a 13-scale seedling fusion feature map;
carrying out down-sampling on an output layer of the 52-scale feature extraction sub-module to obtain 26 × 256 feature maps, carrying out up-sampling on an output layer of the 13-scale feature extraction sub-module to obtain 26 × 512 feature maps, and carrying out tensor splicing on the 26 × 256 feature maps obtained by down-sampling, the 26 × 512 feature maps obtained by up-sampling and the output layer of the 26-scale feature extraction sub-module to obtain 26-scale seedling fusion feature maps;
and carrying out down-sampling on 104-scale output seedling feature maps obtained by carrying out feature extraction on the input rice seedling pictures to obtain 52 x 128 feature maps, carrying out up-sampling on output layers of the 26-scale feature extraction sub-modules to obtain 52 x 512 feature maps, and carrying out tensor splicing on the 52 x 128 feature maps obtained by down-sampling, the 52 x 512 feature maps obtained by up-sampling and the output layers of the 52-scale feature extraction sub-modules to obtain 52-scale seedling fusion feature maps.
8. The improved YOLOV3 network-based rice seedling testing method as claimed in claim 7, wherein the multi-scale seedling position prediction model comprises the following steps:
the 13-scale seedling fusion features constructed by the 13-scale fusion feature construction submodule are subjected to convolution of a set of convolution kernels, 13 × 512 convolution kernels and 1 × 18 convolution kernel to obtain a 13-scale prediction result 13 × 18;
the 26-scale seedling fusion features constructed by the 26-scale fusion feature construction submodule are subjected to convolution of a set of convolution kernels, 13 × 512 convolution kernels and 1 × 18 convolution kernel to obtain a 26-scale prediction result 26 × 18;
the 52-scale seedling fusion features constructed by the 52-scale fusion feature construction submodule are subjected to convolution of a set of convolution kernels, 13 × 256 convolution kernel and 1 × 18 convolution kernel to obtain a 52-scale prediction result 52 × 18.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010736191.3A CN112036437B (en) | 2020-07-28 | 2020-07-28 | Rice seedling detection model based on improved YOLOV network and method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010736191.3A CN112036437B (en) | 2020-07-28 | 2020-07-28 | Rice seedling detection model based on improved YOLOV network and method thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112036437A true CN112036437A (en) | 2020-12-04 |
| CN112036437B CN112036437B (en) | 2024-06-07 |
Family
ID=73583325
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010736191.3A Active CN112036437B (en) | 2020-07-28 | 2020-07-28 | Rice seedling detection model based on improved YOLOV network and method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112036437B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113435309A (en) * | 2021-06-24 | 2021-09-24 | 农业农村部南京农业机械化研究所 | Rice seedling row identification method based on row vector grid classification |
| CN114119640A (en) * | 2022-01-27 | 2022-03-01 | 广东皓行科技有限公司 | Model training method, image segmentation method and image segmentation system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019144575A1 (en) * | 2018-01-24 | 2019-08-01 | 中山大学 | Fast pedestrian detection method and device |
| CN110648316A (en) * | 2019-09-07 | 2020-01-03 | 创新奇智(成都)科技有限公司 | Steel coil end face edge detection algorithm based on deep learning |
| WO2020047738A1 (en) * | 2018-09-04 | 2020-03-12 | 安徽中科智能感知大数据产业技术研究院有限责任公司 | Automatic pest counting method based on combination of multi-scale feature fusion network and positioning model |
-
2020
- 2020-07-28 CN CN202010736191.3A patent/CN112036437B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019144575A1 (en) * | 2018-01-24 | 2019-08-01 | 中山大学 | Fast pedestrian detection method and device |
| WO2020047738A1 (en) * | 2018-09-04 | 2020-03-12 | 安徽中科智能感知大数据产业技术研究院有限责任公司 | Automatic pest counting method based on combination of multi-scale feature fusion network and positioning model |
| CN110648316A (en) * | 2019-09-07 | 2020-01-03 | 创新奇智(成都)科技有限公司 | Steel coil end face edge detection algorithm based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 孟本成;: "基于YOLOV3算法的行人检测方法", 电视技术, no. 09 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113435309A (en) * | 2021-06-24 | 2021-09-24 | 农业农村部南京农业机械化研究所 | Rice seedling row identification method based on row vector grid classification |
| CN113435309B (en) * | 2021-06-24 | 2022-04-26 | 农业农村部南京农业机械化研究所 | Rice seedling row identification method based on row vector grid classification |
| CN114119640A (en) * | 2022-01-27 | 2022-03-01 | 广东皓行科技有限公司 | Model training method, image segmentation method and image segmentation system |
| CN114119640B (en) * | 2022-01-27 | 2022-04-22 | 广东皓行科技有限公司 | Model training method, image segmentation method and image segmentation system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112036437B (en) | 2024-06-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109800736B (en) | Road extraction method based on remote sensing image and deep learning | |
| Arya et al. | Deep learning: As the new frontier in high-throughput plant phenotyping | |
| CN111340141A (en) | Crop seedling and weed detection method and system based on deep learning | |
| Wen et al. | Wheat spike detection and counting in the field based on SpikeRetinaNet | |
| CN109344738A (en) | The recognition methods of crop diseases and pest crop smothering and device | |
| CN114841961B (en) | Wheat scab detection method based on image enhancement and improved YOLOv5 | |
| CN112036437A (en) | Rice seedling detection model based on improved YOLOV3 network and method thereof | |
| Agarwal et al. | Efficient NetB3 for Automated Pest Detection in Agriculture | |
| CN115797781A (en) | Crop identification method and device, computer equipment and storage medium | |
| CN113051983B (en) | Method for training field crop disease recognition model and field crop disease recognition | |
| Ahmed et al. | Recognizing a spatial extreme dependence structure: A deep learning approach | |
| CN114299379A (en) | Shadow area vegetation coverage extraction method based on high dynamic image | |
| CN114120359A (en) | Method for measuring body size of group-fed pigs based on stacked hourglass network | |
| CN117557914A (en) | Crop pest identification method based on deep learning | |
| Wang et al. | High‐throughput measurement of plant fitness traits with an object detection method using Faster R‐CNN | |
| Menezes et al. | Pseudo-label semi-supervised learning for soybean monitoring | |
| Chadœuf et al. | Modelling unbiased dispersal kernels over continuous space by accounting for spatial heterogeneity in marking and observation efforts | |
| Wang et al. | Sugarcane stem node detection and localization for cutting using deep learning | |
| CN117392668A (en) | Wheat scab state evaluation method and system and electronic equipment | |
| Jurado-Ruiz et al. | LeTra: a leaf tracking workflow based on convolutional neural networks and intersection over union | |
| CN114549958B (en) | Night and camouflage target detection method based on context information perception mechanism | |
| CN116188855A (en) | Multi-scale plant disease identification method, device, storage medium and apparatus | |
| CN116206210A (en) | NAS-Swin-based remote sensing image agricultural greenhouse extraction method | |
| Kamei | Effect of image resolution on automatic detection of whitefly (Hemiptera: Aleyrodidae) species on tomato leaflets using deep learning | |
| CN114882362A (en) | Natural gas pipeline high consequence area building change detection method based on remote sensing image |
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 |