CN115035313A - Black-neck crane identification method, device, equipment and storage medium - Google Patents
Black-neck crane identification method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN115035313A CN115035313A CN202210678921.8A CN202210678921A CN115035313A CN 115035313 A CN115035313 A CN 115035313A CN 202210678921 A CN202210678921 A CN 202210678921A CN 115035313 A CN115035313 A CN 115035313A
- Authority
- CN
- China
- Prior art keywords
- black
- neck
- crane
- neck crane
- image
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a black-neck crane identification method, a device, equipment and a storage medium. The method comprises the steps of combining technologies such as image classification, image segmentation, feature point extraction, machine learning and target detection with refined intelligent recognition of the black-neck crane, establishing an accurate recognition model for the black-neck crane, recognizing the ID of the black-neck crane by a method of point extraction of the facial features of the black-neck crane and facial depth convolution network machine learning based on the recognition model, recognizing the ID of each black-neck crane on the basis of judging the species of the black-neck crane, and providing data support for propagation, migration and individual behavior research of the black-neck crane.
Description
Technical Field
The invention relates to the technical field of black-neck crane identification, in particular to a black-neck crane identification method, a device, equipment and a storage medium.
Background
The black-neck crane is a national primary important protection wild animal and also an internationally recognized rare or endangered animal, and is praised as 'plateau spirit' or 'bird panda'. With the popularization of technologies such as informatization, artificial intelligence, machine learning and the like, accurate identification for specific species becomes a demand for fine management in an actual application scenario.
At present, the technology proposes that a machine learning technology is applied to animal face identification, animal face verification image data is input into a convolution network nerve, feature point extraction is carried out through the convolution network nerve, face abstract features of different dimensionalities are obtained, and therefore identification of species is identified. (application No. 201811591828.3 animal face recognition method, apparatus, computer device, and storage medium). The technology provides that the existing face recognition algorithm is applied to the animal face algorithm recognition, but in the practical application, the difference between the facial features of the human face and the facial features of the animal is very large, and the animal classification methods are various and can be classified according to the living environment, the shape, the food habits and the lactation mode. The body types and the face differences of different animals are large, the body type difference between the largest animal and the smallest animal can be as high as 20-30 meters, the technology is difficult to be applied to positioning the face of a certain animal, the face needs to be positioned, and the face identification is difficult to realize.
Aiming at a scheme of bird identification, 202110344587.8 discloses a bird fine-grained identification method based on second-order characteristics; 201810348415.6A bird identification method based on deep learning; 201811091554.1A bird recognition method based on improved convolution neural network, which is based on public data set training, model establishment, and target species recognition through target image comparison model, wherein the model scheme mainly forms a second or third order equal fine-grained feature model through basic model or model improvement, and the essence is to add special fine-grained feature for comparison recognition according to bird image classification. For example: 202110344587.8A bird fine-grained identification method based on second-order characteristics is added with the establishment of a machine learning mode to the bird fine-grained identification model result based on bird image classification technology, thereby realizing the purpose of bird identification, and the identification accuracy of the type depends on a multi-dimensional quality and quantity library of various bird species data sets.
However, in the natural world, birds have a large number of birds and large size difference, and then, in the research of bird identification algorithms, the same set of algorithm is applied to the identification of large species, so that the problem of low identification accuracy is caused, and when a data set changes or training base models of the data set are few in the identification process of specific species, the problem of inaccurate identification is caused. The black-neck crane has small crane face and unique morphological characteristics such as feather and beak, so the recognition accuracy is not ideal when the conventional universal bird recognition model is applied for recognition.
Therefore, how to improve the identification accuracy of the black-neck crane and realize the identification of the individual ID of the black-neck crane is a technical problem which needs to be solved urgently. The above is only for the purpose of assisting understanding of the technical aspects of the present invention, and does not represent an admission that the above is prior art.
Disclosure of Invention
The invention mainly aims to provide a black-neck crane identification method, a black-neck crane identification device, black-neck crane identification equipment and a black-neck crane storage medium, and aims to solve the technical problem that the identification precision of the conventional bird identification model is low when the conventional bird identification model identifies a black-neck crane with the characteristics of small crane face, unique feather and unique beak.
In order to achieve the purpose, the invention provides a black-neck crane identification method, which comprises the following steps:
acquiring a bird image data set, and performing data annotation on each bird image to acquire a training data set;
constructing a pre-training model, and training the pre-training model by using the training data set to obtain a black-neck crane identification model;
acquiring a black-neck crane image data set, and performing key region segmentation on each black-neck crane image;
extracting characteristic points of a plurality of key areas, and training a matching model corresponding to each key area by using the characteristic points to obtain a black-neck crane matching model;
when a target image is received, inputting the target image into a black-neck crane identification model, and judging whether the target image is a black-neck crane or not;
and if so, matching the target image with the black-neck crane characteristic matrix in the black-neck crane database by using the black-neck crane matching model to obtain the corresponding label of the black-neck crane in the target image in the black-neck crane database.
Optionally, the acquiring bird image dataset step specifically includes:
acquiring a bird picture; the bird pictures comprise a black-neck crane picture and a non-black-neck crane picture which are captured in a protected area and extracted by frame extraction in a bird video;
performing image preprocessing on the black-neck crane picture and the non-black-neck crane picture to obtain a bird image dataset; wherein the image pre-processing comprises one or more of rotation, mirroring, Gaussian noise, brightness adjustment, or contrast adjustment.
Optionally, the step of training a pre-training model by using the training data set to obtain a black-neck crane recognition model specifically includes:
extracting a feature map in a training data set, and establishing a feature pyramid with output feature proportions of 1/4, 1/8, 1/16, 1/32 and 1/64 respectively;
processing the training data set by using a region generation network to obtain a candidate region recommendation box;
determining a region feature map based on the feature pyramid and the candidate region recommendation box;
and obtaining a classification result according to the regional characteristic diagram, and adjusting a cross entropy loss function according to the classification result to obtain a first-order black-neck crane identification model.
Optionally, after the step of obtaining a classification result according to the region feature map and adjusting the cross entropy loss function according to the classification result to obtain a first-order black-neck crane identification model, the method further includes:
classifying the black-neck crane behavior in the black-neck crane picture to obtain N-1 black-neck crane behavior image sets;
and respectively taking the N-1 black-neck crane behavior and action image sets as training data sets for training a first-order black-neck crane recognition model, returning to the step of extracting the characteristic diagram in the training data sets, and performing N-1 times of training on the first-order black-neck crane recognition model until the N-order black-neck crane recognition model is obtained.
Optionally, the step of performing key region segmentation on each black-neck crane image specifically includes:
classifying the black-neck crane images into a left black-neck crane image, a right black-neck crane image and a front black-neck crane image according to the position angle of the black-neck crane;
and performing matrix scanning learning on each classified black-neck crane image, and performing key region segmentation on each black-neck crane image according to a learning result.
Optionally, the step of extracting feature points of a plurality of key regions, training a matching model corresponding to each key region by using the feature points, and obtaining a black-neck crane matching model specifically includes:
extracting the characteristic points of each key area, and establishing a corresponding matching model for each key area; the key regions comprise a left face neck region, a left eye region, a left coracoid region, a right face neck region, a right eye region, a right coracoid region and a front face neck region;
and training the matching model pair corresponding to each key area by using the characteristic point of each key area to obtain a black-neck crane matching model.
Optionally, the step of matching the target image with the black-neck crane feature matrix in the black-neck crane database by using the black-neck crane matching model to obtain the corresponding label of the black-neck crane in the target image in the black-neck crane database specifically includes:
matching the target image with a black-neck crane characteristic matrix in a black-neck crane database by using a black-neck crane matching model, and calculating the Euclidean distance value between the characteristic of each black-neck crane image in the black-neck crane database and the characteristic of the target image;
judging whether a black-neck crane with an Euclidean distance value smaller than a preset value exists in the black-neck crane database, and if so, outputting the label of the black-neck crane in the black-neck crane database;
otherwise, adding the target image into the black-neck crane image data set, and training the black-neck crane matching model.
In addition, in order to achieve the above object, the present invention also provides a black-neck crane recognition apparatus including:
the marking module is used for acquiring a bird image data set, and performing data marking on each bird image to acquire a training data set;
the first training module is used for constructing a pre-training model and training the pre-training model by using the training data set to obtain a black-neck crane identification model;
the segmentation module is used for acquiring a black-neck crane image data set and performing key region segmentation on each black-neck crane image;
the second training module is used for extracting feature points of a plurality of key areas and training a matching model corresponding to each key area by using the feature points to obtain a black-neck crane matching model;
the judging module is used for inputting the target image into a black-neck crane identification model when the target image is received, and judging whether the target image is a black-neck crane or not;
and the matching module is used for matching the target image with the black-neck crane characteristic matrix in the black-neck crane database by using the black-neck crane matching model if so, and acquiring the corresponding label of the black-neck crane in the target image in the black-neck crane database.
In addition, in order to achieve the above object, the present invention also provides a black-neck crane identifying apparatus, including: the device comprises a memory, a processor and a black-neck crane identification method program which is stored on the memory and can run on the processor, wherein the black-neck crane identification method program realizes the steps of the black-neck crane identification method when being executed by the processor.
In order to achieve the above object, the present invention further provides a storage medium storing a black-neck crane identification method program, which when executed by a processor, implements the steps of the black-neck crane identification method described above.
The method comprises the steps of obtaining a bird image data set, constructing a pre-training model and training to obtain a black-neck crane recognition model, obtaining the black-neck crane image data set, executing key region segmentation and training to obtain a black-neck crane matching model, and judging whether a target image is a black-neck crane and a label of the corresponding black-neck crane in a black-neck crane database when the target image is received. The method comprises the steps of establishing an accurate black-neck crane identification model by combining technologies such as image classification, image segmentation, feature point extraction, machine learning and target detection with the refined intelligent identification of the black-neck crane, establishing the identification model aiming at the black-neck crane, realizing the identification of the ID of the black-neck crane by the methods of point extraction of the facial features of the black-neck crane and facial depth convolution network machine learning based on the identification model, identifying the ID of each black-neck crane on the basis of judging the species of the black-neck crane, and providing data support for the breeding, migration and individual behavior research of the black-neck crane.
Drawings
Fig. 1 is a schematic structural diagram of a black-neck crane identification device in an embodiment of the present invention;
FIG. 2 is a schematic flow chart illustrating a black-neck crane identification method according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an embodiment of the present invention illustrating a training data set enhanced by image preprocessing;
FIG. 4 is a diagram illustrating the learning of a pre-training model according to an embodiment of the present invention;
FIG. 5 is a schematic diagram illustrating a principle of establishing a black-neck crane recognition model according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a black-neck crane identification model according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of a black-neck crane target detection in an embodiment of the present invention;
FIG. 8 is a diagram illustrating a black-neck crane target detection according to an embodiment of the present invention;
FIG. 9 is a schematic diagram of a black-neck crane target classification according to an embodiment of the present invention;
FIG. 10 is a diagram of a black-neck crane target classification according to an embodiment of the present invention;
FIG. 11 is a diagram illustrating segmentation of key regions and feature point extraction of a black-neck crane according to an embodiment of the present disclosure;
FIG. 12 is a diagram illustrating convolutional net learning of key regions of a black-neck crane according to an embodiment of the present invention;
fig. 13 is a block diagram of a black-neck crane recognition apparatus according to an embodiment of the present invention.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
At present, the traditional bird recognition model has low recognition precision when recognizing a black-neck crane with small crane face, unique feather and unique coracoid characteristics.
To solve this problem, various embodiments of the black-neck crane identification method of the present invention are proposed. The black-neck crane identification method provided by the invention is characterized in that technologies such as image classification, image segmentation, feature point extraction, machine learning and target detection are combined with the refined intelligent identification of the black-neck crane to establish an accurate black-neck crane identification model, the identification model is established for the black-neck crane, the identification of the black-neck crane ID is realized by the methods of point extraction of the facial features of the black-neck crane and facial deep convolutional network machine learning based on the identification model, the ID of each black-neck crane is identified on the basis of judging the black-neck crane species, and data support is provided for the black-neck crane propagation, migration and individual behavior research.
Referring to fig. 1, fig. 1 is a schematic structural diagram of a black-neck crane identification device according to an embodiment of the present invention.
The device may be a User Equipment (UE) such as a Mobile phone, smart phone, laptop, digital broadcast receiver, Personal Digital Assistant (PDA), tablet computer (PAD), handheld device, vehicular device, wearable device, computing device or other processing device connected to a wireless modem, Mobile Station (MS), or the like. The device may be referred to as a user terminal, portable terminal, desktop terminal, etc.
Generally, the apparatus comprises: at least one processor 301, a memory 302 and a black-neck crane identification method program stored on the memory and executable on the processor, the black-neck crane identification method program being configured to implement the steps of the black-neck crane identification method as described before.
The processor 301 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and so on. The processor 301 may be implemented in at least one hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), and a PLA (Programmable Logic Array). The processor 301 may also include a main processor and a coprocessor, where the main processor is a processor for processing data in an awake state, and is also called a Central Processing Unit (CPU); a coprocessor is a low power processor for processing data in a standby state. In some embodiments, the processor 301 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the display screen. Processor 301 may also include an AI (Artificial Intelligence) processor for processing operations related to the black-neck crane recognition method, such that the black-neck crane recognition method model may be trained and learned autonomously, improving efficiency and accuracy.
In some embodiments, the terminal may further include: a communication interface 303 and at least one peripheral device. The processor 301, the memory 302 and the communication interface 303 may be connected by a bus or signal lines. Various peripheral devices may be connected to communication interface 303 via a bus, signal line, or circuit board. Specifically, the peripheral device includes: at least one of radio frequency circuitry 304, a display screen 305, and a power source 306.
The communication interface 303 may be used to connect at least one peripheral device related to I/O (Input/Output) to the processor 301 and the memory 302. The communication interface 303 is used for receiving the movement tracks of the plurality of mobile terminals uploaded by the user and other data through the peripheral device. In some embodiments, processor 301, memory 302, and communication interface 303 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 301, the memory 302 and the communication interface 303 may be implemented on a single chip or circuit board, which is not limited in this embodiment.
The Radio Frequency circuit 304 is used for receiving and transmitting RF (Radio Frequency) signals, also called electromagnetic signals. The radio frequency circuit 304 communicates with a communication network and other communication devices through electromagnetic signals, so as to obtain the movement tracks and other data of a plurality of mobile terminals. The rf circuit 304 converts an electrical signal into an electromagnetic signal to transmit, or converts a received electromagnetic signal into an electrical signal. Optionally, the radio frequency circuit 304 comprises: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a subscriber identity module card, and so forth. The radio frequency circuitry 304 may communicate with other terminals via at least one wireless communication protocol. The wireless communication protocols include, but are not limited to: metropolitan area networks, various generation mobile communication networks (2G, 3G, 4G, and 5G), Wireless local area networks, and/or WiFi (Wireless Fidelity) networks. In some embodiments, the rf circuit 304 may further include NFC (Near Field Communication) related circuits, which are not limited in this application.
The display screen 305 is used to display a UI (User Interface). The UI may include graphics, text, icons, video, and any combination thereof. When the display screen 305 is a touch display screen, the display screen 305 also has the ability to capture touch signals on or over the surface of the display screen 305. The touch signal may be input to the processor 301 as a control signal for processing. At this point, the display screen 305 may also be used to provide virtual buttons and/or a virtual keyboard, also referred to as soft buttons and/or a soft keyboard. In some embodiments, the display screen 305 may be one, the front panel of the electronic device; in other embodiments, the display screens 305 may be at least two, which are respectively disposed on different surfaces of the electronic device or in a foldable design; in still other embodiments, the display screen 305 may be a flexible display screen disposed on a curved surface or a folded surface of the electronic device. Even further, the display screen 305 may be arranged in a non-rectangular irregular figure, i.e. a shaped screen. The Display screen 305 may be made of LCD (liquid crystal Display), OLED (Organic Light-Emitting Diode), and the like.
The power supply 306 is used to power various components in the electronic device. The power source 306 may be alternating current, direct current, disposable or rechargeable. When the power source 306 includes a rechargeable battery, the rechargeable battery may support wired or wireless charging. The rechargeable battery may also be used to support fast charge technology.
Those skilled in the art will appreciate that the configuration shown in fig. 1 does not constitute a limitation of the black-neck crane identification apparatus and may include more or fewer components than those shown, or some components in combination, or a different arrangement of components.
The embodiment of the invention provides a black-neck crane identification method, and referring to fig. 2, fig. 2 is a schematic flow diagram of a first embodiment of the black-neck crane identification method.
In this embodiment, the black-neck crane identification method includes the following steps:
and S100, acquiring a bird image data set, performing data annotation on each bird image, and acquiring a training data set.
In this embodiment, acquiring a bird image dataset includes acquiring a bird picture, and performing image preprocessing on the bird picture to obtain a bird image dataset.
Specifically, the bird pictures comprise black-neck crane pictures and non-black-neck crane pictures which are captured in the protected area and extracted by frame extraction in the bird video, the non-black-neck crane pictures comprise common birds in the protected areas such as gray cranes, bar-headed gooses, sheldrake ducks and crows, after the bird pictures are obtained, the quality of the collected bird pictures can be detected and screened, and non-target or over-fuzzy pictures can be deleted.
It should be noted that, compared with a graph database shot by a macro camera, the graph database that is used for model training through pictures or videos collected by a camera has the advantages that real shooting scenes are used for training, and the recognition accuracy of a training algorithm can be enhanced.
Specifically, as shown in FIG. 3, the image pre-processing includes one or more of rotation, mirroring, Gaussian noise, brightness adjustment, or contrast adjustment, by which the training data set is enhanced.
In a preferred embodiment, photos of the black-neck crane under different angles, different features and different environments can be extracted independently from the bird image data set, and the photos are used for performing subsequent training of a multi-stage black-neck crane recognition model with higher accuracy on the basis of building the black-neck crane recognition model.
And S200, constructing a pre-training model, and training the pre-training model by using the training data set to obtain a black-neck crane identification model.
In this embodiment, as shown in fig. 4, the constructed pre-training model is obtained by pre-training in an open data set based on a deep learning architecture and a target detection model. The deep learning architecture can adopt a deep residual error network, a VGG Net and the like, the target detection model can adopt FasterR-CNN, R-FCN, SDD and the like, the public data set can adopt coco2017, and the method has good robustness and precision.
And after the constructed pre-training model is obtained, training the pre-training model by using a training data set to obtain a black-neck crane identification model.
Specifically, a feature pyramid with output feature proportions of 1/4, 1/8, 1/16, 1/32 and 1/64 is established by extracting a feature map in a training data set, a region generation network is utilized to process the training data set to obtain a candidate region recommendation box, a region feature map is determined based on the feature pyramid and the candidate region recommendation box, a classification result is obtained according to the region feature map, and a cross entropy loss function is adjusted according to the classification result to obtain a first-order black-neck crane recognition model.
After that, the extracted pictures of the black-neck crane under different angles, different characteristics and different environments can be used for establishing a higher-order black-neck crane identification model on the first-order black-neck crane identification model so as to improve the accuracy of the black-neck crane identification model in identifying the bird species of the black-neck crane.
Specifically, black-neck crane behaviors in the black-neck crane picture are classified to obtain N-1 black-neck crane behavior and action image sets, the N-1 black-neck crane behavior and action image sets are respectively used as training data sets for training a first-order black-neck crane recognition model, the step of extracting feature maps in the training data sets is returned, and the first-order black-neck crane recognition model is trained for N-1 times until the N-order black-neck crane recognition model is obtained.
In order to more clearly illustrate this step, the principle of establishing the black-neck crane identification model is described in detail below.
As shown in fig. 5, when the black-neck crane recognition model is established, the pre-training model is used as an initial parameter to train the obtained bird data set of the marked protection area, and the preliminary bird first-order pre-training model is used. And repeating the steps to carry out a more accurate second-order to N-order black-neck crane training model on the black-neck crane species of the specific species. The pre-training model process is as follows:
(1) the input image, as shown in fig. 6, uses a set of underlying conv (convolution) + relu (linear rectification) + pooling) layers to extract feature maps of the image. FPNs (feature pyramids) with output features of proportions 1/4, 1/8, 1/16, 1/32 and 1/64 are respectively extracted, input images are changed into pictures with different scales through an image pyramid, and the feature pyramids are used for a subsequent RPN (region generation network) layer and a full connection layer.
(2) Each image is processed using RPN (region generation network), and 1000 to 2000 candidate region recommendation frames are screened using NMS (non-maximum suppression). And sending the prediction frame into a P-Net network to obtain species prediction frame prediction, and sending the species frame obtained by P-Net prediction into R-Net and 0-Net for multiple judgment and prediction.
(3) Inputting the feature pyramid in (1) and the candidate region recommendation box in (2), deleting redundant prediction boxes through a maximum suppression algorithm, and extracting a generic feature maps (region feature map) after integrating the information.
(4) And (3) obtaining a fine-tuned box position and classification result through a ROIPool (region of interest pooling) full link layer by using the region feature map obtained in the step (3), and finally, filtering out 100 boxes at the maximum (default) by using NMS (non-maximum suppression). And obtaining the category of the detection frame and the final accurate position of the detection frame. Target detection and classification is completed, and as shown in fig. 7-10, black-neck crane species/non-black-neck crane species are trained by predicting the cross entropy loss function during training.
Compared with the existing bird recognition model which is mainly based on image recognition, the algorithm is simple, the recognition accuracy is low, in the embodiment, the black-neck crane recognition model is provided, the main advantage is that the pre-training is carried out on the public data set coco2017 through the fast RCNN to obtain the bird pre-training model, the model can be used for machine learning, sample data is continuously collected, a large number of machine learning samples of key video frames such as the behavior and the action of the black-neck crane of a specific species are added, and the more accurate second-order, third-order and N-order black-neck crane training model is established.
And step S300, acquiring a black-neck crane image data set, and performing key region segmentation on each black-neck crane image.
In this embodiment, a black-neck crane image dataset is acquired, and a high-resolution black-neck crane image dataset is acquired, where the high-resolution photo set is a black-neck crane photo set whose image definition and quality meet the requirements of the algorithm through quality detection and screening. The concrete requirements are as follows: the image resolution is higher than 1920 x 1080 and the body proportion is larger than 80%.
In a preferred embodiment, data enhancement can be performed on the acquired black-neck crane image data set, specifically including left-right turning of a photo, angle adjustment, image RGB result output under different light rays, and the like.
After the black-neck crane image data set is acquired, classifying the black-neck crane images into a left black-neck crane image, a right black-neck crane image and a front black-neck crane image according to the position angle of the black-neck crane; and performing matrix scanning learning on each classified black-neck crane image, and performing key region segmentation on each black-neck crane image according to a learning result.
Specifically, the key regions include a left face neck region, a left eye region, a left rostral region, a right face neck region, a right eye region, a right rostral region, and a frontal face neck region. Therefore, the image parts of 7 key areas are extracted from one black-neck crane image.
And step S400, extracting characteristic points of a plurality of key areas, and training a matching model corresponding to each key area by using the characteristic points to obtain a black-neck crane matching model.
In this embodiment, feature points of each key region are extracted, a corresponding matching model is established for each key region, and the matching model pair corresponding to the key region is trained by using the feature points of each key region, so as to obtain a black-neck crane matching model.
Specifically, as shown in fig. 11, when extracting the feature points, the extraction of the key feature points is performed for 7 keys segmented for each black-neck crane. And extracting 64 feature points from each of 2 side pictures, extracting 32 feature points from the front picture, totaling 160 feature points, and establishing a black-neck crane three-dimensional recognition model.
It should be noted that, in the conventional machine learning, a unified convolutional network neural algorithm is established to train the model, and in order to improve the accuracy and fineness of the algorithm, in the identification application of the black-neck crane, the embodiment performs convolutional neural network learning on 7 segmented key regions and feature points respectively, as shown in fig. 12, and establishes a black-neck crane high-dimensional feature algorithm model.
In the embodiment, the fine research is mainly carried out on the crane face (including the neck) of the black-neck crane, on the basis of the traditional machine learning convolution neural algorithm and on the basis of the general facial recognition machine learning algorithm, a fine recognition algorithm model of the black-neck crane is established according to the characteristics of the black-neck crane, and the fine recognition of the black-neck crane id is realized by establishing a high-dimensional characteristic database of the black-neck crane and fully considering the small crane face, unique morphological characteristics such as feathers and beaks and the like in the high-dimensional recognition algorithm model of the black-neck crane.
Step S500, when a target image is received, inputting the target image into a black-neck crane identification model, and judging whether the target image is a black-neck crane.
Specifically, when a target image to be recognized is received, the target image may be input into a black-neck crane recognition model to determine whether the bird species in the target image is a black-neck crane species or a non-black-neck crane species.
And step S600, if so, matching the target image with the black-neck crane characteristic matrix in the black-neck crane database by using the black-neck crane matching model to obtain the corresponding label of the black-neck crane in the target image in the black-neck crane database.
Specifically, when the bird species in the obtained target image is judged to be black-neck crane species, matching the target image with a black-neck crane feature matrix in a black-neck crane database by using a black-neck crane matching model, and calculating the Euclidean distance value of the feature of each black-neck crane image in the black-neck crane database and the feature of the target image; judging whether a black-neck crane with an Euclidean distance value smaller than a preset value exists in the black-neck crane database, and if so, outputting the label of the black-neck crane in the black-neck crane database; otherwise, adding the target image into the black-neck crane image data set, and training the black-neck crane matching model.
It should be noted that, in this embodiment, the preset value is 1.25, that is, the collected picture of the black-neck crane is compared with the data set in the database by an algorithm, and if the euclidean distance is less than or equal to 1.25, the matching is successful, and it can be determined that the picture is the same black-neck crane, and if the matching is unsuccessful, the black-neck crane matching model training step is repeated, and the database is increased by machine learning comparison, so as to improve the recognition success rate.
In this embodiment, a black-neck crane identification method is provided, which is to perform 160 feature point extractions on the front, the face, the beak part, and the eye region of the left and right sides of the black-neck crane on the basis of identifying the species of the black-neck crane, establish a multi-dimensional feature algorithm model of the black-neck crane through high-dimensional feature point extraction and multi-dimensional convolution network learning, and perform a matching result of euclidean distances through a target object and the multi-dimensional feature algorithm model in a database, so as to identify the black-neck crane, that is, identify the ID of the target black-neck crane, thereby achieving the purpose of fine identification. Species identification of the black-neck crane and fine identification of the black-neck crane ID are realized by constructing a black-neck crane algorithm model. The identification of the black-neck crane ID can identify the black-neck crane a and the black-neck crane b through the constructed black-neck crane algorithm, the black-neck crane n can be accurately positioned to a certain black-neck crane through the algorithm, the fine research of the black-neck crane is realized, and the method has very important value in the research of black-neck crane groups and individual species.
Referring to fig. 13, fig. 13 is a block diagram of the black-neck crane recognition device according to an embodiment of the present invention.
As shown in fig. 13, the black-neck crane recognition apparatus according to the embodiment of the present invention includes:
the marking module 10 is used for acquiring a bird image data set, performing data marking on each bird image, and acquiring a training data set;
the first training module 20 is configured to construct a pre-training model, and train the pre-training model by using the training data set to obtain a black-neck crane recognition model;
the segmentation module 30 is used for acquiring a black-neck crane image data set and performing key region segmentation on each black-neck crane image;
the second training module 40 is used for extracting feature points of a plurality of key areas, and training a matching model corresponding to each key area by using the feature points to obtain a black-neck crane matching model;
the judging module 50 is configured to, when a target image is received, input the target image into a black-neck crane recognition model, and judge whether the target image is a black-neck crane;
and the matching module 60 is used for matching the target image with the black-neck crane characteristic matrix in the black-neck crane database by using the black-neck crane matching model if so, so as to obtain the corresponding label of the black-neck crane in the target image in the black-neck crane database.
Other embodiments or specific implementation manners of the black-neck crane identification device of the present invention can refer to the above method embodiments, and are not described herein again.
Furthermore, an embodiment of the present invention further provides a storage medium, where a black-neck crane identification method program is stored, and when being executed by a processor, the black-neck crane identification method program implements the steps of the black-neck crane identification method described above. Therefore, a detailed description thereof will be omitted. In addition, the beneficial effects of the same method are not described in detail. For technical details not disclosed in embodiments of the computer-readable storage medium referred to in the present application, reference is made to the description of embodiments of the method of the present application. It is determined that, by way of example, the program instructions may be deployed to be executed on one computing device or on multiple computing devices at one site or distributed across multiple sites and interconnected by a communication network.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
It should be noted that the above-described embodiments of the apparatus are merely schematic, where the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. In addition, in the drawings of the embodiment of the apparatus provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, and may be specifically implemented as one or more communication buses or signal lines. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that the present invention may be implemented by software plus necessary general hardware, and may also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components and the like. Generally, functions performed by computer programs can be easily implemented by corresponding hardware, and specific hardware structures for implementing the same functions may be various, such as analog circuits, digital circuits, or dedicated circuits. However, the software program implementation is a better implementation mode for the present invention in more cases. Based on such understanding, the technical solutions of the present invention or portions thereof contributing to the prior art may be embodied in the form of a software product, where the computer software product is stored in a readable storage medium, such as a floppy disk, a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk of a computer, and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network device) to execute the methods according to the embodiments of the present invention.
Claims (10)
1. A black-neck crane identification method is characterized by comprising the following steps:
acquiring a bird image data set, and performing data annotation on each bird image to obtain a training data set;
constructing a pre-training model, and training the pre-training model by using the training data set to obtain a black-neck crane recognition model;
acquiring a black-neck crane image data set, and performing key region segmentation on each black-neck crane image;
extracting characteristic points of a plurality of key areas, and training a matching model corresponding to each key area by using the characteristic points to obtain a black-neck crane matching model;
when a target image is received, inputting the target image into a black-neck crane identification model, and judging whether the target image is a black-neck crane or not;
and if so, matching the target image with the black-neck crane characteristic matrix in the black-neck crane database by using the black-neck crane matching model to obtain the corresponding label of the black-neck crane in the target image in the black-neck crane database.
2. The blackneck crane identification method of claim 1, wherein the step of acquiring a bird image dataset specifically comprises:
acquiring a bird picture; the bird pictures comprise a black-neck crane picture and a non-black-neck crane picture which are captured in a protected area and extracted by frame extraction in a bird video;
performing image preprocessing on the black-neck crane picture and the non-black-neck crane picture to obtain a bird image dataset; wherein the image pre-processing comprises one or more of rotation, mirroring, Gaussian noise, brightness adjustment, or contrast adjustment.
3. The method for identifying a black-neck crane according to claim 2, wherein the step of training a pre-training model by using the training data set to obtain the black-neck crane identification model specifically comprises:
extracting a feature map in a training data set, and establishing a feature pyramid with output feature proportions of 1/4, 1/8, 1/16, 1/32 and 1/64 respectively;
processing the training data set by using a region generation network to obtain a candidate region recommendation box;
determining a region feature map based on the feature pyramid and the candidate region recommendation box;
and obtaining a classification result according to the regional characteristic diagram, and adjusting a cross entropy loss function according to the classification result to obtain a first-order black-neck crane identification model.
4. The blackneck crane identification method according to claim 3, wherein after the step of obtaining the classification result according to the region feature map and adjusting the cross entropy loss function according to the classification result to obtain the first-order blackneck crane identification model, the method further comprises:
classifying the black-neck crane behavior in the black-neck crane picture to obtain N-1 black-neck crane behavior and action image sets;
and respectively taking the N-1 black-neck crane behavior and action image sets as training data sets for training a first-order black-neck crane recognition model, returning to the step of extracting the characteristic diagram in the training data sets, and performing N-1 times of training on the first-order black-neck crane recognition model until the N-order black-neck crane recognition model is obtained.
5. The black-neck crane identification method according to claim 1, wherein the step of performing key region segmentation on each black-neck crane image specifically comprises:
classifying the black-neck crane images into a left black-neck crane image, a right black-neck crane image and a front black-neck crane image according to the position angle of the black-neck crane;
and performing matrix scanning learning on each classified black-neck crane image, and performing key region segmentation on each black-neck crane image according to a learning result.
6. The method for identifying the black-neck crane according to claim 5, wherein the step of extracting feature points of a plurality of key regions and training a matching model corresponding to each key region by using the feature points to obtain the black-neck crane matching model specifically comprises:
extracting the characteristic points of each key area, and establishing a corresponding matching model for each key area; the key regions comprise a left face neck region, a left eye region, a left coracoid region, a right face neck region, a right eye region, a right coracoid region and a front face neck region;
and training the matching model pair corresponding to each key area by using the characteristic point of each key area to obtain a black-neck crane matching model.
7. The black-neck crane identification method according to claim 6, wherein the step of matching the target image with the black-neck crane feature matrix in the black-neck crane database by using the black-neck crane matching model to obtain the corresponding label of the black-neck crane in the target image in the black-neck crane database specifically comprises:
matching the target image with a black-neck crane characteristic matrix in a black-neck crane database by using a black-neck crane matching model, and calculating the Euclidean distance value between the characteristic of each black-neck crane image in the black-neck crane database and the characteristic of the target image;
judging whether a black-neck crane with an Euclidean distance value smaller than a preset value exists in the black-neck crane database, and if so, outputting the label of the black-neck crane in the black-neck crane database;
otherwise, adding the target image into the black-neck crane image data set, and training the black-neck crane matching model.
8. A black-neck crane recognition device, characterized in that, black-neck crane recognition device includes:
the marking module is used for acquiring a bird image data set, and performing data marking on each bird image to acquire a training data set;
the first training module is used for constructing a pre-training model and training the pre-training model by using the training data set to obtain a black-neck crane identification model;
the segmentation module is used for acquiring a black-neck crane image data set and performing key region segmentation on each black-neck crane image;
the second training module is used for extracting feature points of a plurality of key areas and training a matching model corresponding to each key area by using the feature points to obtain a black-neck crane matching model;
the judging module is used for inputting the target image into a black-neck crane identification model when the target image is received, and judging whether the target image is a black-neck crane or not;
and the matching module is used for matching the target image with the black-neck crane characteristic matrix in the black-neck crane database by using the black-neck crane matching model if so, and acquiring the corresponding label of the black-neck crane in the target image in the black-neck crane database.
9. A black-neck crane recognition apparatus, comprising: memory, processor and black-neck crane identification method program stored on the memory and executable on the processor, which when executed by the processor implements the steps of the black-neck crane identification method as claimed in any one of claims 1 to 7.
10. A storage medium, characterized in that the storage medium has stored thereon a blackneck crane identification method program which, when executed by a processor, implements the steps of the blackneck crane identification method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210678921.8A CN115035313B (en) | 2022-06-15 | 2022-06-15 | Black-neck crane identification method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210678921.8A CN115035313B (en) | 2022-06-15 | 2022-06-15 | Black-neck crane identification method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115035313A true CN115035313A (en) | 2022-09-09 |
| CN115035313B CN115035313B (en) | 2023-01-03 |
Family
ID=83125262
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210678921.8A Active CN115035313B (en) | 2022-06-15 | 2022-06-15 | Black-neck crane identification method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115035313B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116229518A (en) * | 2023-03-17 | 2023-06-06 | 百鸟数据科技(北京)有限责任公司 | Bird species observation method and system based on machine learning |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105261109A (en) * | 2015-07-24 | 2016-01-20 | 广州广电运通金融电子股份有限公司 | Identification method of prefix letter of banknote |
| CN106864351A (en) * | 2017-01-17 | 2017-06-20 | 盐城师范学院 | A kind of vehicle at night traveling auxiliary lighting system based on computer vision |
| US20190095764A1 (en) * | 2017-09-26 | 2019-03-28 | Panton, Inc. | Method and system for determining objects depicted in images |
| CN110580024A (en) * | 2019-09-17 | 2019-12-17 | Oppo广东移动通信有限公司 | workshop auxiliary operation implementation method and system based on augmented reality and storage medium |
| CN111738062A (en) * | 2020-05-09 | 2020-10-02 | 广智微芯(扬州)有限公司 | Automatic re-identification method and system based on embedded platform |
| CN111931592A (en) * | 2020-07-16 | 2020-11-13 | 苏州科达科技股份有限公司 | Object recognition method, device and storage medium |
| CN112561893A (en) * | 2020-12-22 | 2021-03-26 | 平安银行股份有限公司 | Picture matching method and device, electronic equipment and storage medium |
| CN113283359A (en) * | 2021-06-02 | 2021-08-20 | 万达信息股份有限公司 | Authentication method and system for handheld certificate photo and electronic equipment |
| US20210286977A1 (en) * | 2019-02-26 | 2021-09-16 | Tencent Technology (Shenzhen) Company Limited | Computer application method and apparatus for generating three-dimensional face model, computer device, and storage medium |
| CN113420690A (en) * | 2021-06-30 | 2021-09-21 | 平安科技（深圳）有限公司 | Vein identification method, device and equipment based on region of interest and storage medium |
-
2022
- 2022-06-15 CN CN202210678921.8A patent/CN115035313B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105261109A (en) * | 2015-07-24 | 2016-01-20 | 广州广电运通金融电子股份有限公司 | Identification method of prefix letter of banknote |
| CN106864351A (en) * | 2017-01-17 | 2017-06-20 | 盐城师范学院 | A kind of vehicle at night traveling auxiliary lighting system based on computer vision |
| US20190095764A1 (en) * | 2017-09-26 | 2019-03-28 | Panton, Inc. | Method and system for determining objects depicted in images |
| US20210286977A1 (en) * | 2019-02-26 | 2021-09-16 | Tencent Technology (Shenzhen) Company Limited | Computer application method and apparatus for generating three-dimensional face model, computer device, and storage medium |
| CN110580024A (en) * | 2019-09-17 | 2019-12-17 | Oppo广东移动通信有限公司 | workshop auxiliary operation implementation method and system based on augmented reality and storage medium |
| CN111738062A (en) * | 2020-05-09 | 2020-10-02 | 广智微芯(扬州)有限公司 | Automatic re-identification method and system based on embedded platform |
| CN111931592A (en) * | 2020-07-16 | 2020-11-13 | 苏州科达科技股份有限公司 | Object recognition method, device and storage medium |
| CN112561893A (en) * | 2020-12-22 | 2021-03-26 | 平安银行股份有限公司 | Picture matching method and device, electronic equipment and storage medium |
| CN113283359A (en) * | 2021-06-02 | 2021-08-20 | 万达信息股份有限公司 | Authentication method and system for handheld certificate photo and electronic equipment |
| CN113420690A (en) * | 2021-06-30 | 2021-09-21 | 平安科技（深圳）有限公司 | Vein identification method, device and equipment based on region of interest and storage medium |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116229518A (en) * | 2023-03-17 | 2023-06-06 | 百鸟数据科技(北京)有限责任公司 | Bird species observation method and system based on machine learning |
| CN116229518B (en) * | 2023-03-17 | 2024-01-16 | 百鸟数据科技(北京)有限责任公司 | Bird species observation method and system based on machine learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115035313B (en) | 2023-01-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11830230B2 (en) | Living body detection method based on facial recognition, and electronic device and storage medium | |
| WO2021077984A1 (en) | Object recognition method and apparatus, electronic device, and readable storage medium | |
| US8750573B2 (en) | Hand gesture detection | |
| US8792722B2 (en) | Hand gesture detection | |
| CN112633144A (en) | Face occlusion detection method, system, device and storage medium | |
| CN111914812B (en) | Image processing model training method, device, equipment and storage medium | |
| CN104063686B (en) | Crop leaf diseases image interactive diagnostic system and method | |
| CN109559362B (en) | Image subject face replacing method and device | |
| CN112749609B (en) | Human body image segmentation method, device, computer equipment and storage medium | |
| CN111144284B (en) | Method and device for generating depth face image, electronic equipment and medium | |
| CN112419326B (en) | Image segmentation data processing method, device, equipment and storage medium | |
| CN112052186A (en) | Target detection method, device, equipment and storage medium | |
| CN112989995B (en) | Text detection method and device and electronic equipment | |
| CN115699082A (en) | Defect detection method and device, storage medium and electronic equipment | |
| CN114332530A (en) | Image classification method and device, computer equipment and storage medium | |
| CN108198130A (en) | Image processing method, device, storage medium and electronic equipment | |
| CN112651333B (en) | Silence living body detection method, silence living body detection device, terminal equipment and storage medium | |
| CN112949507A (en) | Face detection method and device, computer equipment and storage medium | |
| CN112836625A (en) | Face living body detection method and device and electronic equipment | |
| CN110866900A (en) | Water body color identification method and device | |
| CN115035313B (en) | Black-neck crane identification method, device, equipment and storage medium | |
| CN113947613B (en) | Target area detection method, device, equipment and storage medium | |
| CN110991325A (en) | Model training method, image recognition method and related device | |
| CN116580211B (en) | Key point detection method, device, computer equipment and storage medium | |
| CN114511877A (en) | Behavior recognition method and device, storage medium and terminal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: Black necked crane identification method, device, equipment, and storage medium Effective date of registration: 20230811 Granted publication date: 20230103 Pledgee: Hua Xia Bank Co.,Ltd. Kunming Branch Pledgor: Yunnan Information Technology Co.,Ltd. Registration number: Y2023980051902 |