CN114758357A - Animal species identification method based on neural network and improved K-SVD algorithm - Google Patents
Animal species identification method based on neural network and improved K-SVD algorithm Download PDFInfo
- Publication number
- CN114758357A CN114758357A CN202210394606.2A CN202210394606A CN114758357A CN 114758357 A CN114758357 A CN 114758357A CN 202210394606 A CN202210394606 A CN 202210394606A CN 114758357 A CN114758357 A CN 114758357A
- Authority
- CN
- China
- Prior art keywords
- animal
- image
- algorithm
- neural network
- improved
- 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
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)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
An animal species identification method based on a neural network and an improved K-SVD algorithm belongs to the field of image identification. An animal species identification method based on a neural network and an improved K-SVD algorithm is characterized in that video image data are preprocessed to obtain picture data, and an animal individual identity identification data set is manufactured; step two, improving a YOLOv4 method, and detecting an animal target in image data according to an improved YOLOv4 algorithm; the improvement of YOLOv4 means that a data set is used for model pre-training, and then Fine-tuning is carried out through the constructed target detection data set to train proper weight; increasing three-color component information of the image based on a K-SVD algorithm, and improving the K-SVD algorithm to de-noise the animal individual image; and identifying the individual animal identity based on the deep convolutional neural network. The method of the invention has high accuracy in identifying the animal individuals.
Description
Technical Field
The invention relates to an animal identification method, in particular to an animal species identification method based on a neural network and an improved K-SVD algorithm.
Background
In terms of the development of animal husbandry in China, the large-scale cultivation can effectively improve the production efficiency of animal husbandry cultivation, is beneficial to increasing both production and income of farmers, effectively improves the food safety guarantee and the epidemic disease prevention and control capacity, and becomes an effective way for realizing economic growth and environmental sustainable development in the field of animal husbandry. However, realizing large-scale cultivation also means that the method faces greater challenges, and higher requirements are put on the cultivation mode and the management system to a certain extent. In large-scale breeding, the difference of the yield of the individual in livestock groups due to the difference of information such as age, physical signs, epidemic prevention and the like, for example, the difference of the milk yield of sheep and cows in different individuals is larger than the difference of the breed. Therefore, aiming at the precise livestock breeding, especially the livestock such as sheep with large economic value of the monomer, individual difference needs to be considered in the breeding process, the growth condition of the individual is obtained according to collection, and then different breeding schemes are made for different individuals, so as to realize refined breeding. On one hand, the identification effect of the traditional animal individual identification based on the ear tag type is limited by the distance, and the ear tag is easy to lose effectiveness due to damage; on the other hand, the piercing ear tag can cause physical damage to the animal body when being mounted on the ear of the animal, even the ear of the animal is torn due to improper mounting, the ear of the animal falls off and is lost due to certain probability, and the electronic ear tag can also influence the growth condition of the animal. Therefore, the identification mode of manual inspection, information acquisition, analysis and arrangement is high in labor and material cost, and individual growth information cannot be fed back timely and accurately, so that managers cannot be effectively guided to optimize aiming at occurring problems, and the breeding productivity is restricted to a great extent.
The identification of individual livestock and the intelligent perception of behaviour are the core of accurate animal husbandry. Cattle, pigs and sheep are closely related to our lives as common livestock, and have certain economic value, so that various scholars at home and abroad carry out extensive research on the livestock. In view of the current combination of future development trend and market demand of precise animal husbandry, the development directions of intelligent perception and behavior detection of animal individual information are more focused on non-contact, high-precision and high-degree automation. Therefore, as the premise and the foundation of accurate animal husbandry, establishing a livestock individual identification algorithm based on non-contact type, low cost and high identification precision has important practical significance. The invention provides the cattle individual identity recognition method based on the deep convolutional neural network by constructing a data set and combining a target detection algorithm and an individual recognition algorithm, so that the individual recognition accuracy and the individual recognition rate are effectively improved, and an effective solution is provided for the individual identity recognition of livestock.
The realization of the rapid and efficient individual identification of livestock is the basis of accurate animal husbandry, which is beneficial to improving the quality of animal products and enhancing the competitive power of the animal products in the international market. Realizes the economic growth of animal husbandry and sustainable ecological agriculture, improves the production efficiency and greatly reduces the cost of manpower and material resources. Accurate livestock raising requires realization of individual information acquisition of livestock and analysis of behaviors, so individual identification is the premise and the basis of automatic information acquisition and processing. The method comprises the steps of establishing individual livestock archives in an accurate breeding system, acquiring information, managing breeding, tracing livestock products and the like, and is based on the premise of individual identification of livestock. The individual identification based on the neural network can realize intelligent acquisition of livestock information and improvement of the working efficiency of a farm, is favorable for analyzing animal behaviors based on monitoring videos subsequently, analyzes the growth process of the animal behaviors according to the animal behavior analysis rule and the change of the animal behavior analysis rule, and then formulates corresponding feeding schemes for different individuals, so that the breeding yield is improved. The development of livestock breeding industry promotes the development of the agricultural market to a certain extent. However, animal husbandry is relatively distributed in operation, and the current agricultural risk technology is still immature, so that cheating and insurance phenomena generally exist in many areas, the operation effect of agricultural risk is not ideal, and farmers also face the problems of low guarantee, difficult claim settlement and the like. Compared with the traditional manual method, the method has the advantages of high identification accuracy and low cost through the images, and can effectively avoid the occurrence of cheat insurance. The method can also be popularized to other application scenes, such as finding lost pets, managing intelligent farms and the like, so that the development of the project research not only has important theoretical value, but also has wide market prospect and application value.
The method takes the identification problem of the cattle in the animal husbandry breeding process as a research target, is based on the deep convolutional neural network, combines methods such as deep learning and machine learning, realizes quick and effective identification aiming at the identification problem of the individual cattle, and has great practical significance.
As an indispensable part of the agricultural field, the animal husbandry plays an important role in national economy and social development of China, and especially as a big agricultural country, the animal husbandry puts higher requirements on the development of the animal husbandry. With the continuous development of economy and society in China in recent years, the traditional animal husbandry is actively upgraded and modified while achieving great results. At present, the intelligent livestock identification is still in the development stage in China, and the traditional livestock identification method based on ear tags is still used in many areas. Therefore, it is necessary to develop a livestock individual identification algorithm with high reliability, low cost and high identification precision aiming at the development of animal husbandry at the present stage.
Disclosure of Invention
The invention provides an animal species identification method based on a neural network and an improved K-SVD algorithm.
An animal species identification method based on a neural network and an improved K-SVD algorithm is realized by the following steps:
firstly, making an animal individual identification data set;
preprocessing video image data to obtain picture data, calibrating by using a LabelImg calibration tool, and making a target detection data set for individual detection;
step two, improving a YOLOv4 method, and detecting whether an animal target exists in the image data according to an improved YOLOv4 algorithm;
if yes, carrying out the next image denoising operation;
if not, classifying the image into an image set without an animal target;
the improvement of YOLOv4 means that a data set is used for model pre-training, and then Fine-tuning is carried out through the constructed target detection data set to train proper weight;
increasing three-color component information of the image based on a K-SVD algorithm, and improving the K-SVD algorithm to de-noise the animal individual image;
fourthly, identifying the identity of the animal individual based on the deep convolutional neural network;
and taking TensorFlow as a deep learning framework, taking the training set image after detection as the input of the IncepotionV 3 network, and training to obtain a more optimized cow identity recognition model.
Preferably, the process of making an animal individual identification data set of the pair of the step one comprises:
labeling of data samples: and (3) manually marking the data sample by using a LabelImg marking tool, marking the position of the target in the original image and the category of the target, and generating a corresponding xml file for each image.
Preferably, the process of detecting whether an animal target exists in the image data according to the modified YOLOv4 algorithm in the step two comprises:
design the network structure of YOLOv 4:
CSPDarknet53 is used as a backbone network, SPP is used as an additional module of Neck, PANET is used as a feature fusion module of Neck, and YOLOv3 is used as Head; wherein the Darknet53 contains 5 large residual blocks, and the number of small residual units contained in the 5 large residual blocks is 1, 2, 8 and 4 respectively; CSPDarknet53 adds CSPNet on each large residual block of Darknet53, integrates the large residual blocks into a feature map through the change of gradient, divides the feature map into two parts, one part carries out convolution operation, and the other part is combined with the convolution result of the last time; in the target detection process, the CSP improves the learning capacity of the CNN and reduces the calculated amount at the same time; the PANet fully utilizes feature fusion, and the fusion method in YOLOv4 is changed from addition to multiplication;
The network model of SSD is based on a feed-forward convolutional neural network VGG 16: replacing the FC6 layer and the FC7 layer in the VGG16 full connection layer with a convolution layer Conv6 of 3x3 and a convolution layer Conv7 of 1x 1; removing all Dropout layers and FC8 layers in the VGG 16; changing the pool layer pool5 in VGG16 from 2x2 with original stride =2 to 3x3 with stride = 1; an Atrous algorithm is added; a convolution layer is newly added on the basis of VGC 16;
the invention has the beneficial effects that:
the method adopts a YOLOV4 method to detect whether an animal target exists in the image, if so, the next image denoising operation is carried out, otherwise, the image is classified as an image set without the animal target. Aiming at an image with an animal target, a multi-channel-based K-SVD algorithm is adopted to finish fine separation of training data and testing data, a denoised training data set is used as the input of an inclusion-V3 network, an animal identity recognition model is obtained through training, the model is applied to animal identity recognition in a testing set to obtain model feedback information, training parameters are adjusted and optimized according to feedback, and a more optimized animal identity recognition model is obtained.
The invention relates to an individual recognition algorithm based on a deep learning convolution neural network, wherein the neural network adopts an initiation network model of Google. The concept network model of Google is mainly used for species classification, the method is applied to individual identification of cows, the concept model is trained, and the generalization capability of the concept model on the individual identification of cows is verified through the method. The algorithm has the following advantages: 1. in the process of animal identification, the input of the neural network is the original information of the animal image, compared with the method of performing early-stage processing by main element analysis, the method of supporting a vector machine (SVM) to perform two-classification identification retains the animal individual data information; 2. a large amount of data information (such as background pixel information of the environment where the dairy cow individual is located) which is irrelevant to identification and contained in the animal image increases the difficulty of identification, and the algorithm can perform high-dimensional extraction on local information (animal trunk line information) in the image to acquire image commonalities for classification and identification.
Particularly, the invention provides an image multi-channel based K-SVD algorithm on the basis of improvement of the K-SVD algorithm, and the accuracy of individual identification of animal identity is improved by acquiring more detailed information of an original image by using multiple channels. In addition, the dairy cow data are abstractly learned by using an inclusion V3 model of Google, and learning of a network is accelerated by using three inclusion models used by an inclusion-V3 model. The influence of noise on the identification accuracy is verified in a mode of superposing different levels of noise on the original image, and the K-SVD algorithm is proved to have a promoting effect on the identification accuracy of the dairy cows. Through the recognition rate test of the test set and the training set of the animal individual recognition method, the condition that the network is not over-fitted in the training process is proved, and the feasibility of the animal individual recognition technology under the complex background is verified.
Drawings
FIG. 1 is a flow chart of the present invention;
fig. 2 is a network structure diagram of YOLOV4 according to the present invention.
Detailed Description
The first specific implementation way is as follows:
in the embodiment, as shown in fig. 1, the animal species identification method based on the neural network and the improved K-SVD algorithm is implemented by the following steps:
Firstly, making an animal individual identification data set;
preprocessing video image data to obtain picture data, calibrating by using a LabelImg calibration tool, and making a target detection data set for individual detection;
step two, improving a YOLOv4 method, and detecting whether an animal target exists in image data according to an improved YOLOv4 algorithm;
if yes, performing the next image denoising operation;
if the image does not exist, classifying the image as an image set without an animal target;
the improvement of YOLOv4 refers to the steps of pre-training a model by using a data set, then performing Fine-tuning on the constructed target detection data set, training appropriate weight, further improving the recognition effect, and sending the detected individual image into an individual identity recognition model;
thirdly, based on a K-SVD algorithm, increasing three-color component information of the image to enhance the extraction capability of image features, and improving the K-SVD algorithm to denoise the animal individual image;
the individual identification of the cattle is realized based on a sparse representation theory of dictionary learning, and an overcomplete dictionary is constructed by utilizing image samples to carry out sparse representation on an original input signal. The classical K-SVD algorithm is improved, and the three-color component information of the image is added to enhance the extraction capability of the image features;
The system mainly comprises a classical dictionary learning algorithm with a K-SVD algorithm as a main part, and on the basis, provides an image multi-channel based K-SVD algorithm, and improves the individual identification precision of livestock by acquiring more detailed information of an original image by using multiple channels. Although the K-SVD algorithm is simple and effective, the K-SVD algorithm cannot achieve satisfactory performance in some practical applications due to the fact that the K-SVD algorithm has insufficient capacity of extracting original image information. Therefore, aiming at the problems, the invention constructs a dictionary learning algorithm based on image multi-channel K-SVD, extracts three-color components of the image on RGB three channels, contains more detailed information and hidden information of the image, and effectively improves the classification and identification precision of the image;
fourthly, identifying the identity of the animal individual based on the deep convolutional neural network;
the YOLOV4 target detection method, the improved k-SVD algorithm and the Incepion V3 convolutional neural network are combined, and the method is applied to target detection and individual identification of dairy cows in a complex visual scene. And taking TensorFlow as a deep learning framework, taking the training set image after detection as the input of the IncepotionV 3 network, and training to obtain a more optimized cow identity recognition model. The model performs abstract learning on milk cow data, the effectiveness of the algorithm is verified through testing the accuracy of the model, ten-fold cross validation and an overfitting experiment, and the effect of improving a K-SVD denoising network on the milk cow identification accuracy is verified through performing algorithm testing by using milk cow images polluted by different levels of noise.
The second embodiment is as follows:
different from the first embodiment, in the animal species identification method based on the neural network and the improved K-SVD algorithm of the present embodiment, the first step of the process of creating the animal individual identification data set includes:
the data set is an integral part of the model training process. The quality of the data samples directly affects the results of the model.
The data sample collection process comprises the following steps: the size of the data sample influences the model training effect, the larger the data sample is, the better the model training effect of target detection is, and the main ways of data set sources include searching of pictures on the internet, cutting of videos and providing of projects.
Labeling of data samples: the marking of the data sample is an important premise for generating the executable file, the more accurate the marking of the data sample is, the accuracy of the model can be greatly improved, the marking of the data sample is carried out by utilizing a LabelImg marking tool, the position of a target in an original image is marked, the category of the target is marked, and a corresponding xml file is generated for each image.
The third concrete implementation mode:
different from the first or second embodiment, in the animal species identification method based on the neural network and the improved K-SVD algorithm of the present embodiment, the process of detecting whether the animal target exists in the image data according to the improved YOLOv4 algorithm in the second step includes:
The YOLO network is a regression-based target detection algorithm, has high detection speed and achieves good effect in many target detection tasks. The YOLO network divides an input picture into 32 × 32 meshes, and if the center position of an object falls into a certain mesh, the mesh is responsible for detecting a target. The Yolov4 target detection algorithm is more complex in network structure than Yolov3, and uses a plurality of training skills to improve the accuracy of the neural network.
Design the network structure of YOLOv 4:
CSPDarknet53 is used as a backbone network, SPP is used as an additional module of Neck, PANET is used as a feature fusion module of Neck, and YOLOv3 is used as Head; wherein the Darknet53 contains 5 large residual blocks, and the number of small residual units contained in the 5 large residual blocks is 1, 2, 8 and 4 respectively; CSPDarknet53 adds CSPNet on each large residual block of Darknet53, integrates the large residual blocks into a feature map through gradient change, and divides the feature map into two parts, wherein one part is subjected to convolution operation, and the other part is combined with the convolution result of the last time; in the target detection process, the CSP improves the learning capacity of the CNN and reduces the calculated amount at the same time; the PANet fully utilizes feature fusion, and the fusion method is changed from addition to multiplication in the YOLOv4, so that the network can obtain more accurate target detection capability; the structure of the YOLOV4 network is shown in fig. 2.
The network model of SSD is based on a feed-forward convolutional neural network VGG16, which is modified from VGG 16: replacing the FC6 layer and the FC7 layer in the VGG16 full connection layer with a convolution layer Conv6 of 3x3 and a convolution layer Conv7 of 1x 1; removing all Dropout layers and FC8 layers in the VGG 16; changing the pool layer pool5 in the VGG16 from 2x2 with original stride =2 to 3x3 with stride = 1; the Atrous algorithm (hole algorithm) is added to obtain denser score mapping; convolution layers are added on the basis of the VGC16 to obtain more characteristic maps for detection.
The embodiments of the present invention are disclosed as the preferred embodiments, but not limited thereto, and those skilled in the art can easily understand the spirit of the present invention and make various extensions and changes without departing from the spirit of the present invention.
Claims (3)
1. An animal species identification method based on a neural network and an improved K-SVD algorithm is characterized in that: the method is realized by the following steps:
firstly, making an animal individual identification data set;
preprocessing video image data to obtain picture data, calibrating by using a LabelImg calibration tool, and making a target detection data set for individual detection;
Step two, improving a YOLOv4 method, and detecting whether an animal target exists in the image data according to an improved YOLOv4 algorithm;
if yes, carrying out the next image denoising operation;
if not, classifying the image into an image set without an animal target;
the improvement of YOLOv4 means that a data set is used for model pre-training, and then Fine-tuning is carried out through the constructed target detection data set to train proper weight;
increasing three-color component information of the image based on a K-SVD algorithm, and improving the K-SVD algorithm to de-noise the animal individual image;
fourthly, identifying the identity of the animal individual based on the deep convolutional neural network;
and taking TensorFlow as a deep learning framework, taking the training set image after detection as the input of the IncepotionV 3 network, and training to obtain a more optimized cow identity recognition model.
2. The animal species identification method based on the neural network and the improved K-SVD algorithm as claimed in claim 1, wherein: step one the process of making an individual identification data set of an animal of the pair comprises:
labeling of data samples: and (3) manually marking the data sample by using a LabelImg marking tool, marking the position of the target in the original image and the category of the target, and generating a corresponding xml file for each image.
3. The animal species identification method based on the neural network and the improved K-SVD algorithm as claimed in claim 1 or 2, wherein: the process of detecting whether an animal target exists in the image data according to the improved YOLOv4 algorithm in the second step comprises the following steps:
design the network structure of YOLOv 4:
CSPDarknet53 is used as a backbone network, SPP is used as an additional module of Neck, PANET is used as a feature fusion module of Neck, and YOLOv3 is used as Head; wherein the Darknet53 contains 5 large residual blocks, and the number of small residual units contained in the 5 large residual blocks is 1, 2, 8 and 4 respectively; CSPDarknet53 adds CSPNet on each large residual block of Darknet53, integrates the large residual blocks into a feature map through gradient change, and divides the feature map into two parts, wherein one part is subjected to convolution operation, and the other part is combined with the convolution result of the last time; in the target detection process, the CSP improves the learning capacity of the CNN and reduces the calculated amount at the same time; the PANet fully utilizes feature fusion, and a fusion method is changed from addition to multiplication in YOLOv 4;
the network model of SSD is based on a feed-forward convolutional neural network VGG 16: replacing the FC6 and FC7 layers in the VGG16 full connectivity layer with convolution layers Conv6 of 3x3 and convolution layers Conv7 of 1x 1; removing all Dropout layers and FC8 layers in the VGG 16; changing the pool layer pool5 in VGG16 from 2x2 with original stride =2 to 3x3 with stride = 1; an Atrous algorithm is added; a convolution layer is added on the basis of VGC 16.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210394606.2A CN114758357A (en) | 2022-04-14 | 2022-04-14 | Animal species identification method based on neural network and improved K-SVD algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210394606.2A CN114758357A (en) | 2022-04-14 | 2022-04-14 | Animal species identification method based on neural network and improved K-SVD algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114758357A true CN114758357A (en) | 2022-07-15 |
Family
ID=82330187
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210394606.2A Pending CN114758357A (en) | 2022-04-14 | 2022-04-14 | Animal species identification method based on neural network and improved K-SVD algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114758357A (en) |
-
2022
- 2022-04-14 CN CN202210394606.2A patent/CN114758357A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yang et al. | Development of image recognition software based on artificial intelligence algorithm for the efficient sorting of apple fruit | |
CN109949316A (en) | A kind of Weakly supervised example dividing method of grid equipment image based on RGB-T fusion | |
CN107292298A (en) | Ox face recognition method based on convolutional neural networks and sorter model | |
CN108596038B (en) | Method for identifying red blood cells in excrement by combining morphological segmentation and neural network | |
CN111727457B (en) | Cotton crop row detection method and device based on computer vision and storage medium | |
CN112131927B (en) | Sow delivery time prediction system based on gesture conversion characteristics in late gestation period | |
Rong et al. | Pest identification and counting of yellow plate in field based on improved mask r-cnn | |
CN112613428B (en) | Resnet-3D convolution cattle video target detection method based on balance loss | |
CN112528823B (en) | Method and system for analyzing batcharybus movement behavior based on key frame detection and semantic component segmentation | |
CN116052222A (en) | Cattle face recognition method for naturally collecting cattle face image | |
CN115830490A (en) | Multi-target tracking and behavior statistical method for herd health pigs | |
Xiao et al. | Group-housed pigs and their body parts detection with Cascade Faster R-CNN | |
CN113780207A (en) | System and method for goat face recognition | |
CN117456358A (en) | Method for detecting plant diseases and insect pests based on YOLOv5 neural network | |
CN115226650B (en) | Sow oestrus state automatic detection system based on interaction characteristics | |
CN114758357A (en) | Animal species identification method based on neural network and improved K-SVD algorithm | |
CN116295022A (en) | Pig body ruler measurement method based on deep learning multi-parameter fusion | |
CN116563205A (en) | Wheat spike counting detection method based on small target detection and improved YOLOv5 | |
CN115019162A (en) | Silkworm detection method based on deep learning | |
Huang et al. | A survey of deep learning-based object detection methods in crop counting | |
CN113392927A (en) | Animal target detection method based on single-order deep neural network | |
CN112508012B (en) | Orchard pest intelligent positioning and identifying method suitable for small target sample | |
CN115309941B (en) | AI-based intelligent tag retrieval method and system | |
CN116798066A (en) | Sheep individual identity recognition method and system based on deep measurement learning | |
CN113326804A (en) | Animal individual identification method |
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 |