CN112419202A - Wild animal image automatic identification system based on big data and deep learning - Google Patents
Wild animal image automatic identification system based on big data and deep learning Download PDFInfo
- Publication number
- CN112419202A CN112419202A CN202011442143.XA CN202011442143A CN112419202A CN 112419202 A CN112419202 A CN 112419202A CN 202011442143 A CN202011442143 A CN 202011442143A CN 112419202 A CN112419202 A CN 112419202A
- Authority
- CN
- China
- Prior art keywords
- image
- database
- wild animal
- result
- identification
- 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
- 241001465754 Metazoa Species 0.000 title claims abstract description 171
- 238000013135 deep learning Methods 0.000 title claims abstract description 33
- 238000004458 analytical method Methods 0.000 claims abstract description 63
- 238000012544 monitoring process Methods 0.000 claims abstract description 55
- 238000000034 method Methods 0.000 claims abstract description 32
- 238000012545 processing Methods 0.000 claims abstract description 25
- 238000003860 storage Methods 0.000 claims abstract description 25
- 238000011084 recovery Methods 0.000 claims abstract description 22
- 238000005516 engineering process Methods 0.000 claims abstract description 19
- 230000000007 visual effect Effects 0.000 claims abstract description 7
- 238000013079 data visualisation Methods 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 19
- 238000007781 pre-processing Methods 0.000 claims description 11
- 230000005540 biological transmission Effects 0.000 claims description 10
- 238000013527 convolutional neural network Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 7
- 238000003672 processing method Methods 0.000 claims description 7
- 238000009826 distribution Methods 0.000 claims description 6
- 230000011218 segmentation Effects 0.000 claims description 6
- 238000003707 image sharpening Methods 0.000 claims description 5
- 238000012800 visualization Methods 0.000 claims description 5
- 230000006399 behavior Effects 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 230000033764 rhythmic process Effects 0.000 claims description 3
- 238000003708 edge detection Methods 0.000 claims description 2
- 238000004422 calculation algorithm Methods 0.000 abstract description 15
- 238000007726 management method Methods 0.000 abstract description 6
- 230000008030 elimination Effects 0.000 abstract description 2
- 238000003379 elimination reaction Methods 0.000 abstract description 2
- 238000001514 detection method Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 7
- 238000011176 pooling Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 239000013598 vector Substances 0.000 description 5
- 238000007405 data analysis Methods 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 241000894007 species Species 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000005520 cutting process Methods 0.000 description 2
- 238000013523 data management Methods 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 230000008676 import Effects 0.000 description 2
- 230000006698 induction Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 208000012661 Dyskinesia Diseases 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G06T5/73—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/27—Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
Abstract
The embodiment of the invention discloses a wild animal image automatic identification system based on big data and deep learning, which takes a wild animal monitoring image database as a main body, combines a super-resolution image recovery technology and a wild animal automatic identification algorithm to realize the processing and analysis of wild animal data, and builds a front-end interface of the system through data visualization. Specifically, the wild animal monitoring image database is used for storage management of wild animal images and related information, the super-resolution image restoration technology is used for restoration and reconstruction of low-quality compressed images, and the automatic identification algorithm is used for realizing automatic identification and monitoring analysis of the wild animal images through deep learning, so that classification processing of the images and elimination of invalid images are realized. The method can realize the high-efficiency classified storage of the wild animal images and the automatic analysis and visual display of the wild animal data, and improve the intelligent management level of the wild animal images.
Description
Technical Field
The invention relates to the technical field of artificial intelligence-based wild animal monitoring image processing, in particular to a wild animal image automatic identification system based on big data and deep learning.
Background
Ecological civilization construction is the basic national policy of China, and wild animal resource protection is an important means for maintaining ecological balance of the system. In order to assist the science and effective protection of wild animals, a large amount of abundant wild animal monitoring information needs to be collected, the efficiency of the traditional monitoring means and data processing mode is low, and a large amount of manpower, material resources and financial resources need to be consumed. Systematic and intelligent wild animal monitoring systems are the main development trend in the future.
The monitoring means based on the wireless sensor network has the characteristics of wide application range, high use efficiency and small environmental invasiveness, and is the most important monitoring means at present. However, in a field environment, wild animal monitoring has certain specificity, and the collected wild animal monitoring image has the following characteristics: weather changes or occasional abnormal movements of the natural environment can cause false triggering of monitoring equipment, and further, more invalid images without wild animals are generated; the monitored image contains complex background information, which influences the detection and identification of wild animals; the activity of the wild animals is random and hidden, so that the target positions and sizes of the wild animals in the monitored images are not fixed; considering that the power consumption of the sensor node is limited and the signal transmission environment in the protection area is not good enough, the quality of the monitoring image acquired through the wireless sensor network is low, and the information is lost. The above-mentioned problems present challenges to the automatic identification and classification of wild animals.
Disclosure of Invention
Aiming at the technical defects in the prior art, the embodiment of the invention aims to provide a wild animal image automatic identification system, method and electronic equipment based on big data and deep learning, and mainly solves the following related technical problems:
1. in the face of mass wild animal data acquired by a wild animal monitoring system, the distributed data management system is used for realizing efficient data storage and management, and facilitating later retrieval, analysis and application of the data.
2. The wild animal image data after compression and transmission has the characteristics of low image quality and low reference value, and the method realizes low-resolution image reconstruction through a super-resolution image recovery technology and provides a data basis with practical reference value for data analysis.
3. In the face of the characteristics that the manual identification difficulty of mass images is high, the proportion of invalid images is large, and background areas of wild animal monitoring images are complex, the method applies deep learning to the field of wild animal monitoring image identification, and provides possibility for realizing efficient wild animal detection and identification.
Therefore, the technical scheme provided by the invention is as follows:
in a first aspect, an embodiment of the present invention provides a wild animal image automatic identification system based on big data and deep learning, including:
wild animal image monitoring database: adopting a distributed storage architecture to store and manage mass wild animal image data; the wild animal image monitoring database comprises an original image database, a historical image database, an image classification database and an analysis result database;
an image reconstruction module: based on the super-coding resolution image restoration technology, image reconstruction is carried out on the image data in the original image database;
the image recognition analysis module: performing deep convolutional neural network model training according to the image data in the historical image database to obtain an image recognition analysis model; adopting the image identification analysis model to identify and analyze the output result of the image reconstruction module or the image data in the image classification database so as to obtain an identification result and an analysis result; and storing the identification result into the image classification database, and storing the analysis result into the analysis result database.
In some embodiments of the present application, the image classification database includes an invalid image database to be deleted, and the image reconstruction module is specifically configured to:
image quality grading: based on the error rate in image transmission, performing quality grade division on image data in the original image database to obtain a low-quality image and an invalid sample image, deleting the invalid sample image, and storing the invalid sample image to the invalid image database to be deleted;
low-quality image preprocessing: performing pixel expansion preprocessing on the low-quality image;
image enhancement processing: improving the definition of the preprocessed low-quality image in an image sharpening mode, and carrying out denoising treatment;
and (3) image recovery processing: and performing super-pixel segmentation on the low-quality image subjected to image enhancement processing to obtain a sparse depth map, and performing image recovery on the sparse depth map.
In some embodiments of the present application, the image recognition analysis module is specifically configured to:
training historical images: importing an original image sample library from the historical image database; extracting an interested region from the original image sample library, and establishing an interested region image sample library; simultaneously inputting the image sample library of the region of interest and the image which is not extracted from the region of interest into a dual-channel VGG16 network for model training to obtain an image recognition analysis model;
original image identification: importing an image to be recognized from the original image database or the classified image database; extracting the region of interest of the image to be identified to obtain a region of interest image; and inputting the image of the region of interest into an image recognition analysis model for recognition, and outputting a recognition result.
In certain embodiments of the present application, the image classification database further comprises a wildlife image database set and a to-be-processed image database;
if the image to be identified comes from the image reconstruction module and the identification result meets the following conditions, executing corresponding operation: if the identification result picture does not contain the wild animal image, marking the picture as a false trigger image and transferring the false trigger image into an invalid image database to be deleted in the image classification database; if the wild animal image is identified according to the identification result and the result precision is more than or equal to the set threshold value, transferring the wild animal image into a sub-database of the label corresponding to the wild animal image database set in the image classification database according to the identification result; if the wild animal image is identified and the result precision is smaller than the set threshold value, transferring the wild animal image into an image database to be processed in an image classification database;
if the image to be recognized is from the image database to be processed, and the recognition result meets the following conditions, executing corresponding operation: if the wild animal image is identified according to the identification result and the result precision is more than or equal to the set threshold value, transferring the wild animal image into a sub-database of the label corresponding to the wild animal image database set in the image classification database according to the identification result; and if the wild animal image is identified and the result precision is less than the set threshold, transferring the wild animal image into an image database to be processed in the image classification database.
Further, in certain preferred embodiments of the present application, the system further comprises a display module: building a front-end visual interface based on data visualization;
the visualization interface is to:
displaying a single monitoring image led into the automatic wild animal image identification system and an identification result of automatic identification of the single monitoring image;
displaying batch monitoring images led into the automatic wild animal image identification system and statistical data for automatic identification of the batch monitoring images;
displaying the analysis result, wherein the analysis result comprises animal habitat distribution, behavior rhythm or population distribution.
In a second aspect, an embodiment of the present invention provides a wild animal image processing method based on big data and deep learning, including:
acquiring historical images from a historical image database, and training according to the historical images to obtain an image recognition analysis model;
acquiring an original image from an original image database, and performing image reconstruction on the original image based on a super-coding resolution image recovery technology to obtain an image to be recognized;
adopting the image identification analysis model to perform identification analysis on the image to be identified to obtain an identification result and an analysis result;
and storing the identification result into an image classification database, and storing the analysis result into an analysis result database.
In a third aspect, an embodiment of the present invention provides an electronic device, including a processor, an input device, an output device, and a memory, where the processor, the input device, the output device, and the memory are connected to each other, where the memory is used to store a computer program, and the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method according to the second aspect.
Compared with the prior art, the wild animal image automatic identification system based on big data and deep learning has the following beneficial effects:
1. based on the characteristics of large data volume and complexity of wild animal images, when a wild animal image database is established, the distributed storage model architecture is adopted to realize efficient storage and management of mass data, and data retrieval and analysis are facilitated.
2. The invention realizes the recovery and reconstruction of the image based on the super-resolution image recovery technology, can greatly improve the identification accuracy of the automatic identification system, and provides the image data with practical application value.
3. According to the wild animal automatic identification algorithm based on the region of interest and the convolutional neural network, disclosed by the invention, the wild animal automatic identification algorithm is developed based on the characteristics of the wild animal monitoring image, so that the influence of a complex background on image identification can be reduced, and the accuracy of image identification is improved.
4. According to the wild animal protection system, big data and deep learning are applied to the wild animal monitoring system, the automation and intelligence level of the system is greatly improved, macroscopic data analysis can be provided for wild animal protection, and a data basis is provided for making a wild animal protection policy.
Drawings
In order to more clearly illustrate the detailed description of the invention or the technical solutions in the prior art, the drawings that are needed in the detailed description of the invention or the prior art will be briefly described below.
FIG. 1 is a block diagram of an overall system of an automatic wild animal image identification system based on big data and deep learning according to the present invention;
FIG. 2 is a block diagram of a wildlife image surveillance database according to the present invention;
FIG. 3 is a flowchart of an overall method of the super-resolution image restoration technique of the present invention;
FIG. 4 is a schematic diagram of weight determination of pixel values;
FIG. 5 is an image restoration flow diagram of the super-resolution image restoration technique of the present invention;
FIG. 6 is a flow chart of the deep learning wild animal automatic identification method of the present invention;
FIG. 7 is a diagram of a deep learning two-channel network model according to the present invention;
FIG. 8 is a flow chart of a wildlife image processing method based on big data and deep learning according to the present invention;
fig. 9 is a block diagram of an electronic device of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention has the following inventive concept: the wild animal image automatic identification system based on big data and deep learning is provided. The system takes a wild animal monitoring image database as a main body, realizes the processing and analysis of wild animal data by combining a super-resolution image recovery technology and a neural network wild animal automatic identification algorithm, and builds a front-end interface of the system through data visualization.
As shown in fig. 1, the system mainly includes:
a1, wild animal image monitoring database: and a distributed storage architecture is adopted to store and manage mass wild animal image data.
The original data of the wild animal image monitoring database is from a wireless sensor network deployed in a wild animal monitoring area; the wireless sensing network consists of a terminal node for acquiring wild animal monitoring images and a coordination node for network relay transmission, wherein the terminal node triggers and shoots the wild animal monitoring images by utilizing infrared induction of a camera and transmits the wild animal monitoring images to the coordination node in a multi-stage mode through the network for image return; and the wild animal image monitoring data adopts a distributed data storage framework according to the characteristics of the stored image.
A1, image reconstruction module: and based on the super-coding resolution image restoration technology, image reconstruction is carried out on the image data in the original image database.
The super-resolution image restoration technology is mainly used for automatically restoring pixels of low-quality images, enhancing the detail information of the images, expanding the resolution of the images and restoring image information lost areas caused by transmission error codes.
A3, image recognition analysis module: carrying out deep convolutional neural network model training according to image data in a historical image database to obtain an image recognition analysis model; adopting the image identification analysis model to identify and analyze the output result of the image reconstruction module or the image data in the image classification database so as to obtain an identification result and an analysis result; and storing the identification result into the image classification database, and storing the analysis result into the analysis result database.
The deep learning is realized based on a deep convolutional neural network, in particular to a wild animal automatic identification algorithm based on an interested region and the convolutional neural network; the automatic wild animal image identification result comprises information such as species and number of wild animals, and the region where the animals are located is marked; the automatic wild animal image identification result can be used for eliminating invalid images, identifying and analyzing single and batch images, classifying and storing image data and the like.
A4, display module: and building a system front-end visual interface based on data visualization.
The system front-end visualization interface is mainly used for realizing the result visualization of automatic identification and analysis of wild animal images based on a man-machine friendly interactive interface.
Further, as shown in fig. 2, the wild animal image monitoring database in the embodiment of the present invention mainly includes an original image database, a historical image database, an image classification database, and an analysis result database.
The original image database is mainly used for storing all original data collected by the data receiving end, wherein the original data refers to unprocessed compressed and encoded image data which contains tags such as shooting time and shooting place.
The historical image database is mainly used for storing historical wild animal monitoring image data acquired from a plurality of natural protection areas in China, is attached with image shooting time and shooting place information, and carries out manual calibration work on database images under the guidance of professional wild animal protection experts. The database is mainly used as a training data set of the deep learning automatic identification system.
The image classification database comprises a plurality of sub-databases, and specifically, the plurality of sub-databases comprise an invalid image database to be deleted, an image database to be processed and a wild animal image database set.
The invalid image database to be deleted is mainly used for storing images with high transmission error rate and false triggering images identified by the deep learning automatic identification system in a short time, and the database data is cleared regularly to reduce occupied storage space; the image database to be processed comprises a database needing secondary identification and a database needing manual identification, wherein the database needing secondary identification is used for storing wild animal images which are not high in identification precision after primary identification by a deep learning automatic identification system and need secondary identification of non-false trigger images, and the database needing manual identification is used for storing wild animal images which are not high in identification precision after secondary identification by the deep learning automatic identification system and need manual identification; the wild animal image database set is a set of a plurality of sub-databases which take factors such as wild animal species attributes as labels, and mainly stores image data identified and determined by the deep learning automatic identification system.
The analysis result database is mainly used for storing result information obtained by analyzing data of the wild animal monitoring image database, and comprises related data such as wild animal habitat analysis and group behavior identification analysis.
Referring to fig. 3 again, the overall process of the super-resolution image restoration technique in this embodiment includes:
a2-1, image quality grading.
Importing image data in an original image database, and dividing the quality grade of an image based on the error rate in image transmission. Said bit error rate emsThe judgment is as follows:
the total number of transmitted codes is the total number of pixel points of the image, the pixel value at the position (i, j) of the originally acquired image is assumed to be represented as T (i, j), the pixel value at the image position finally obtained through multi-stage transmission is represented as R (i, j), and if R (i, j) -T (i, j) ≠ 0, the number of error codes is increased by one.
According to the analysis of the collected monitoring images, the monitoring images are divided into the following 4 quality levels according to the error rate: e.g. of the typemsIf the image quality is less than 5%, the quality grade of the image is A; e is more than or equal to 5%msIf the image quality is less than 10%, the image quality grade is B; e is more than or equal to 10%msIf the image quality is less than 30%, the image quality grade is C; e.g. of the typemsAnd if the image quality is more than or equal to 30 percent, the quality grade of the image is D.
Preferably, the original image with the image quality level D is regarded as an invalid sample, and is transferred to an invalid image database to be deleted.
A2-2, low-quality image preprocessing.
Respectively carrying out image preprocessing on the low-quality wild animal monitoring images with the quality grades of A, B, C, wherein the image preprocessing is to carry out pixel expansion on the images, and specifically comprises the following steps:
and (5) expanding the image by adopting a bicubic interpolation method. The cubic interpolation uses the gray values of 16 pixel points around a target point to perform cubic interpolation, so that an amplification effect closer to a high-resolution image can be obtained, and an interpolation basis function needs to be selected to fit data, wherein the constructed interpolation basis function is as follows:
the weight of each pixel value is determined by the distance from the target point to the pixel point to be solved, and the distance comprises the distance in the horizontal direction and the vertical direction. As shown in fig. 4, taking the pixel a00 as an example, the distances between the target point and the pixel a00 in the vertical and horizontal directions are 1+ u and 1+ v, respectively, and then the weight of the pixel a00 is w ═ w (1+ u) × w (1+ v). The target point pixel value f (i + u, j + v) is calculated as follows.
A=[w(1+u)w(u)w(1-u)w(2-u)]T
C=[w(1+v)w(v)w(1-v)w(2-v)]
f(i+u,j+v)=C×B×A
W (x) represents an interpolation basis function, (i, j) represents coordinate values of pixel points of a target point in the original image, u, v represent distances between the target point and the pixel points in the original image in the vertical and horizontal directions, respectively, f (i + u, j + v) represents pixel values of the target point, a represents a weight vector of the original pixel point in the vertical direction, C represents a weight direction of the original pixel point in the horizontal direction, and B represents pixel values of 16 surrounding pixel points in the original image.
A2-3, image enhancement. And improving the image definition by adopting an image sharpening mode, and simultaneously carrying out denoising treatment on the image. The method comprises the following specific steps:
the method comprises the steps of firstly, carrying out denoising processing on an image by adopting median filtering, wherein the median filtering is to replace the value of a certain pixel point by the median of the pixel points in a specified range, so that the effect of eliminating isolated noise points can be achieved, and meanwhile, the edge of the image can be reserved. And then sharpening the denoised image by adopting an edge detection method based on a Sobel operator, detecting the edge of the image by using a Sobel algorithm through two templates in horizontal and vertical directions, and outputting the edge information of the image. And synthesizing the detected edge image with the original image to finish sharpening the image.
A2-4, image recovery. Fig. 5 is an image restoration flow chart of the super-resolution image restoration technology, and the super-resolution image restoration technology is used for realizing restoration and reconstruction of an image. The method comprises the following specific steps:
firstly, performing super-pixel segmentation on an RGB image; then registering the segmented RGB image with the original depth image; then, the pixel blocks are divided into two categories, namely hole pixels containing edge pixels and hole pixels not containing edge pixels according to the areas where the pixel blocks are located. And calculating the defocus blur amount of the edge hole pixel, wherein the defocus blur amount of the hole pixel module including the edge pixel is defined as the average value of the defocus blur amounts of the edge pixel.
And then calculating the defocus blur amount of the hole pixel module without the edge pixel according to the hole pixel module which is the nearest neighbor of the hole pixel module and contains the edge pixel through a correlation model. The correlation model is as follows:
then, a hole pixel module sp which does not include the edge pixel is selected according to the following formulaiAssociated aperture pixel module including pixels at edgesThe defocus blur amount of (1) is spiThe amount of defocus blur.
And obtaining a sparse depth map at a superpixel level according to the defocus blur amount of each superpixel module, performing sparse representation on each pixel block, then performing optimization processing on the sparse depth map, reducing errors, and finally performing image recovery.
Wherein σspiRepresenting a super-pixel module spiThe defocus blur amount of (1);representing the defocus blur amount of the kth pixel at the edge; n isspiRepresenting the number of pixels at the edge; spi,spjRespectively representing a passive super pixel module and an active super pixel module; h isi,hjThe centers of the passive super pixel module and the active super pixel module are respectively; c. Ci,cjThe color mean values of the passive super pixel module and the active super pixel module are respectively; rhoh,ρcRespectively are preset parameter values in the calculation process of Euclidean distance and color weight between the passive super pixel module and the active super pixel module;a similar correlation model representing a passive super-pixel module and an active super-pixel module;defocused mode representing passive superpixel modulePaste amount.
Referring to fig. 6 again, the method for automatically identifying wild animals based on deep learning in this embodiment specifically includes:
aiming at a wild animal monitoring image shot by an infrared automatic induction camera in a field environment, firstly, extracting an interested region in the monitoring image based on target detection of a regression algorithm, and then, combining a global-local double-channel convolutional neural network to realize automatic identification of the wild animal monitoring image.
Specifically, firstly, detecting the region where the wild animal is located in the image, and realizing automatic segmentation and cutting to obtain an image set (region-of-interest image) taking the animal as a main body; and a two-channel model is provided, and an image without region of interest extraction and a region of interest image are simultaneously input into a VGG16 network for training. And the feature extraction layer performs fine adjustment on the basis of parameters obtained by pre-training of an image database. And finally, inputting the features into a classifier and outputting a final recognition result. The method can overcome the influence of background information on image recognition.
The wild animal automatic identification algorithm is divided into two parts of historical image training and image identification. The training of the historical images comprises the following steps:
a3-1-1, importing the original image sample library.
A3-1-2, detecting and extracting the interested area by the target.
A3-1-3, establishing a region-of-interest image sample library.
A3-1-4, establishing a dual-channel VGG16 network.
A3-1-5, model training is carried out.
The identification of the image comprises the steps of:
a3-2-1, importing the image to be recognized.
A3-2-2, detecting and extracting the interested area.
A3-2-3, region of interest image.
A3-2-4, the recognition system carries out recognition of the image.
A3-2-5, outputting the recognition result.
Specifically, the extraction of the region of interest is realized by adopting a target detection technology based on a regression algorithm, firstly, the image size is modified to 448 × 448, a network is input for prediction, and the network output is a tensor of S × S × (B × 5+ C), wherein S is the number of divided grids, B is the number of frames predicted by the network, and C is the number of categories. Then dividing the whole image into S multiplied by S grids, predicting B detection boundary frames by each grid, and calculating the confidence coefficient of each frame according to the following formula:
Confidence=Pr(Object)×IOU
wherein pr (object) is the probability that the bounding box contains the target object; BBgtA reference standard box based on training labels; BBdtDetecting a bounding box; area represents an area.
The present embodiment uses S-7, that is, the input image is firstly divided into 7 × 7-49 meshes, each mesh predicts B-2 detection bounding boxes, and each detection bounding box includes four parameters x, y, w, and h in the algorithm in addition to the confidence. Where x, y represent the specific location of the detection bounding box and w, h are the width and height of the detection bounding box. Here, only regional regression of wild animals was done, and no specific class prediction of species was done, so C was taken as 1 and the final output was a tensor of 7X (2X 5+ 1). Using the sum of squared errors as a loss function, the calculation formula is as follows:
wherein the content of the first and second substances,represents the jth predictor responsible for prediction in the ith trellis. S is the number of divided grids, and B is the number of predicted frames of the network. x is the number ofi,yiFor the position of the center of the ith detection bounding box with respect to its parent grid, wi,hiThe values of (A) represent the detection bounding box respectivelyHigh and wide.
In each image, there are many grids that do not contain the object, and the confidence of these grids is 0, so that their training gradient is large, and this problem also affects the stability of the whole detection network, and even causes the network divergence. To solve this problem, the present embodiment gives a larger weight to the positioning error in the detection process, and takes λcoord=5。
According to the coordinates of the detected area where the wild animal is (including the coordinates [ x ] at the upper left corner)1,y1]And the coordinate of the lower right corner [ x ]2,y2]) And automatically cutting the image, outputting an interested region image, establishing an interested region data set of the wild animal monitoring image, and providing a data basis for the automatic identification of the wild animal in the next step.
Specifically, the two-channel convolutional neural network adopts a VGG16 network, namely a network structure comprising 16 weight layers. The first five parts of the network are all provided with a plurality of convolutional layers which are connected in series, and finally a maximum pooling layer is accessed, wherein the convolutional layers adopt 3 multiplied by 3 small convolutional kernels. The last part is three fully connected layers and one Softmax layer. The image is input to the model, and a multidimensional feature map is obtained by operations such as convolution and pooling of the above layers. The profile output by each layer is typically expressed as:
in the formula (I), the compound is shown in the specification,andrespectively the nth and mth characteristic diagrams of the l-1 layer (the upper layer) of the l layer (the current layer),a convolution kernel is shown operating between the two graphs. f (x) is the activation function,representing the bias term.
The pooling layer is used for feature mapping and dimensionality reduction, maximum pooling is adopted by the invention, and the convolution feature map is downsampled to form a pooling feature map, which is generally expressed as:
where s is the selected downsampling template,is the weight of the template. After multilayer cascade, the full connection layer carries out vector transformation on the feature graph output by the last pooling layer to form a feature vector and transmits the feature vector to the classification layer.
Constructing a two-channel network model based on VGG16 is shown in FIG. 7, which includes a network channel G1And network channel G2Respectively inputting G training samples which are an image sample library not extracted by the region of interest and an image sample library of the region of interest extracted by the region of interest1And G2Respectively outputting prediction probability matrixes P by taking Softmax as a classifierG1And PG2。
Wherein, Pi_jThe prediction probability of the network output is represented, i belongs to {1, 2}, which respectively corresponds to the network channels G1 and G2, j belongs to {1, 2 … n }, which respectively correspond to n predicted species. PG1And PG2For n row vectors, the probability matrix satisfies that the sum of various prediction probabilities is 1, namely satisfies
Finally, the probability matrix P is fused through classification probabilityG1And PG2And adding and averaging to obtain a final recognition probability matrix Q, and outputting the species corresponding to the probability maximum value in the matrix as a recognition result.
Preferably, the image sample database is from a historical image database; the image to be recognized is from the image obtained after the super-resolution image restoration technology processing and a secondary recognition database required under the image database to be processed in the image classification database;
as a specific complement to this embodiment, the following corresponding operations are taken for the result according to the image recognition:
if the image to be identified is from the image obtained after the super-resolution image restoration technology processing, and the result identified according to the steps meets the following conditions, executing corresponding operation: if the identification result picture does not contain the wild animal image, marking the picture as a false trigger image and transferring the false trigger image into an invalid image database to be deleted in an image classification database; if the wild animal image is identified according to the identification result and the result precision is more than or equal to the set threshold value, transferring the wild animal image into a sub-database of the label corresponding to the wild animal image database set in the image classification database according to the identification result; and if the wild animal image is identified and the result precision is less than the set threshold value, transferring the wild animal image into an image database to be secondarily processed in the image classification database.
If the image to be identified is from the image data of the secondary identification database required under the image database to be processed in the image classification database in the wild animal image monitoring database, and the result after identification according to the steps meets the following conditions, executing corresponding operation: if the wild animal image is identified according to the identification result and the result precision is more than or equal to the set threshold value, transferring the wild animal image into a sub-database of the label corresponding to the wild animal image database set in the image classification database according to the identification result; and if the wild animal image is identified and the result precision is smaller than a set threshold value, the wild animal image is transferred into an image database to be processed in the image classification database, and the database needs to be manually identified.
Further, in this embodiment, the front-end visualization interface of the automatic identification system based on data visualization is specifically planned as follows:
the visual interface comprises result display of the deep learning automatic identification system and result display based on wild animal data analysis. The former comprises the import of single monitoring image and the display of automatic identification result, and the import of batch monitoring image data and the statistical display of batch identification data; the later analyzes the information of habitat distribution, behavior rhythm, population distribution and the like of animals on the basis of acquiring sufficient videos and images, and displays the result in a visual mode of data in the form of charts and the like.
Compared with the prior art, the wild animal image automatic identification system based on big data and deep learning has the following beneficial effects:
1. based on the characteristics of large data volume and complexity of wild animal images, when a wild animal image database is established, the distributed storage model architecture is adopted to realize efficient storage and management of mass data, and data retrieval and analysis are facilitated.
2. The invention realizes the recovery and reconstruction of the image based on the super-resolution image recovery technology, can greatly improve the identification accuracy of the automatic identification system, and provides the image data with practical application value.
3. According to the wild animal automatic identification algorithm based on the region of interest and the convolutional neural network, disclosed by the invention, the wild animal automatic identification algorithm is developed based on the characteristics of the wild animal monitoring image, so that the influence of a complex background on image identification can be reduced, and the accuracy of image identification is improved.
4. According to the wild animal protection system, big data and deep learning are applied to the wild animal monitoring system, the automation and intelligence level of the system is greatly improved, macroscopic data analysis can be provided for wild animal protection, and a data basis is provided for making a wild animal protection policy.
Based on the same inventive concept, an embodiment of the present invention further provides a wild animal image processing method based on big data and deep learning, as shown in fig. 8, including:
s101, acquiring a historical image from a historical image database, and training according to the historical image to obtain an image recognition analysis model;
s102, acquiring an original image from an original image database, and performing image reconstruction on the original image based on a super-coding resolution image recovery technology to obtain an image to be recognized;
s103, carrying out identification analysis on the image to be identified by adopting the image identification analysis model to obtain an identification result and an analysis result;
and S104, storing the identification result into an image classification database, and storing the analysis result into an analysis result database.
Further, performing image reconstruction on the original image specifically includes:
image quality grading: based on the error rate in image transmission, performing quality grade division on the original image to obtain a low-quality image and an invalid sample image, deleting the invalid sample image, and storing the invalid sample image to an invalid image database to be deleted in the image classification database;
low-quality image preprocessing: performing pixel expansion preprocessing on the low-quality image;
image enhancement processing: improving the definition of the preprocessed low-quality image in an image sharpening mode, and carrying out denoising treatment;
and (3) image recovery processing: and performing super-pixel segmentation on the low-quality image subjected to image enhancement processing to obtain a sparse depth map, and performing image recovery on the sparse depth map.
Further, the method further comprises:
if the identification result picture does not contain the wild animal image, marking the picture as a false trigger image and transferring the false trigger image into an invalid image database to be deleted in the image classification database;
if the wild animal image is identified according to the identification result and the result precision is more than or equal to the set threshold value, transferring the wild animal image into a sub-database of the label corresponding to the wild animal image database set in the image classification database according to the identification result;
and if the wild animal image is identified and the result precision is less than the set threshold, transferring the wild animal image into an image database to be processed in the image classification database.
It should be noted that, the above method embodiment only describes the case where the image to be identified comes from the reconstructed image. If the image to be recognized is from the image database to be processed in the image classification database, and the recognition result meets the following conditions, executing corresponding operation: if the wild animal image is identified according to the identification result and the result precision is more than or equal to the set threshold value, transferring the wild animal image into a sub-database of the label corresponding to the wild animal image database set in the image classification database according to the identification result; and if the wild animal image is identified and the result precision is less than the set threshold, transferring the wild animal image into an image database to be processed in the image classification database.
It should be noted that, for a more detailed description of the method embodiment, please refer to the system embodiment described above, which is not repeated herein.
In the embodiment of the invention, the wild animal monitoring image database is used for storage management of wild animal images and related information, the super-resolution image recovery technology is used for recovery and reconstruction of low-quality compressed images, and the automatic identification algorithm is used for realizing automatic identification and monitoring analysis of the wild animal images through deep learning, so that classification processing of the images and elimination of invalid images are realized. The method can realize the high-efficiency classified storage of the wild animal images and the automatic analysis and visual display of the wild animal data, and improve the intelligent management level of the wild animal images.
Corresponding to the method embodiment, the embodiment of the invention also provides the electronic equipment. As shown in fig. 9, the electronic device may include: one or more processors 101, one or more input devices 102, one or more output devices 103, and memory 104, the processors 101, input devices 102, output devices 103, and memory 104 being interconnected via a bus 105. The memory 104 is used to store a computer program comprising program instructions, the processor 101 being configured to invoke the program instructions to perform the methods of the above-described wildlife image processing method embodiment based on big-data and deep learning.
It should be understood that, in the embodiment of the present invention, the Processor 101 may be a Central Processing Unit (CPU), and the Processor may also be other general processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, and the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The input device 102 may include a keyboard or the like, and the output device 103 may include a display (LCD or the like), a speaker, or the like.
The memory 104 may include read-only memory and random access memory, and provides instructions and data to the processor 101. A portion of the memory 104 may also include non-volatile random access memory. For example, the memory 104 may also store device type information.
In specific implementation, the processor 101, the input device 102, and the output device 103 described in the embodiment of the present invention may execute the implementation manner described in the embodiment of the wild animal image processing method based on big data and deep learning provided in the embodiment of the present invention, and details are not described here again.
Further, an embodiment of the present invention also provides a readable storage medium, in which a computer program is stored, where the computer program includes program instructions, and the program instructions, when executed by a processor, implement: the wild animal image processing method based on big data and deep learning is disclosed.
The computer readable storage medium may be an internal storage unit of the electronic device described in the foregoing embodiment, for example, a hard disk or a memory of a system. The computer readable storage medium may also be an external storage device of the system, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) Card, Flash memory Card (Flash Card), etc. provided on the system. Further, the computer readable storage medium may also include both an internal storage unit and an external storage device of the system. The computer-readable storage medium is used for storing the computer program and other programs and data required by the system. The computer readable storage medium may also be used to temporarily store data that has been output or is to be output.
Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the examples described in connection with the embodiments disclosed herein may be embodied in electronic hardware, computer software, or combinations of both, and that the components and steps of the examples have been described in a functional general in the foregoing description for the purpose of illustrating clearly the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention essentially or partially contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
While the invention has been described with reference to specific embodiments, the invention is not limited thereto, and various equivalent modifications and substitutions can be easily made by those skilled in the art within the technical scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A wild animal image automatic identification system based on big data and deep learning is characterized by comprising:
wild animal image monitoring database: adopting a distributed storage architecture to store and manage mass wild animal image data; the wild animal image monitoring database comprises an original image database, a historical image database, an image classification database and an analysis result database;
an image reconstruction module: based on the super-coding resolution image restoration technology, image reconstruction is carried out on the image data in the original image database;
the image recognition analysis module: performing deep convolutional neural network model training according to the image data in the historical image database to obtain an image recognition analysis model; adopting the image identification analysis model to identify and analyze the output result of the image reconstruction module or the image data in the image classification database so as to obtain an identification result and an analysis result; and storing the identification result into the image classification database, and storing the analysis result into the analysis result database.
2. The system of claim 1, wherein the image classification database comprises a database of invalid images to be deleted, and wherein the image reconstruction module is specifically configured to:
image quality grading: based on the error rate in image transmission, performing quality grade division on image data in the original image database to obtain a low-quality image and an invalid sample image, deleting the invalid sample image, and storing the invalid sample image to the invalid image database to be deleted;
low-quality image preprocessing: performing pixel expansion preprocessing on the low-quality image;
image enhancement processing: improving the definition of the preprocessed low-quality image in an image sharpening mode, and carrying out denoising treatment;
and (3) image recovery processing: and performing super-pixel segmentation on the low-quality image subjected to image enhancement processing to obtain a sparse depth map, and performing image recovery on the sparse depth map.
3. The system according to claim 2, characterized in that said image enhancement processing comprises in particular:
carrying out denoising treatment on the preprocessed low-quality image by adopting median filtering;
and sharpening the denoised low-quality image by adopting an edge detection method based on a Sobel operator.
4. The system of claim 2, wherein the image recognition analysis module is specifically configured to:
training historical images: importing an original image sample library from the historical image database; extracting an interested region from the original image sample library, and establishing an interested region image sample library; simultaneously inputting the image sample library of the region of interest and the image which is not extracted from the region of interest into a dual-channel VGG16 network for model training to obtain an image recognition analysis model;
original image identification: importing an image to be recognized from the original image database or the classified image database; extracting the region of interest of the image to be identified to obtain a region of interest image; and inputting the image of the region of interest into an image recognition analysis model for recognition, and outputting a recognition result.
5. The system of claim 4, wherein the image classification database further comprises a wildlife image database set and a pending image database;
if the image to be identified comes from the image reconstruction module and the identification result meets the following conditions, executing corresponding operation: if the identification result picture does not contain the wild animal image, marking the picture as a false trigger image and transferring the false trigger image into an invalid image database to be deleted in the image classification database; if the wild animal image is identified according to the identification result and the result precision is more than or equal to the set threshold value, transferring the wild animal image into a sub-database of the label corresponding to the wild animal image database set in the image classification database according to the identification result; if the wild animal image is identified and the result precision is smaller than the set threshold value, transferring the wild animal image into an image database to be processed in an image classification database;
if the image to be recognized is from the image database to be processed, and the recognition result meets the following conditions, executing corresponding operation: if the wild animal image is identified according to the identification result and the result precision is more than or equal to the set threshold value, transferring the wild animal image into a sub-database of the label corresponding to the wild animal image database set in the image classification database according to the identification result; and if the wild animal image is identified and the result precision is less than the set threshold, transferring the wild animal image into an image database to be processed in the image classification database.
6. The system of any one of claims 1-5, further comprising a display module: building a front-end visual interface based on data visualization;
the visualization interface is to:
displaying a single monitoring image led into the automatic wild animal image identification system and an identification result of automatic identification of the single monitoring image;
displaying batch monitoring images led into the automatic wild animal image identification system and statistical data for automatic identification of the batch monitoring images;
displaying the analysis result, wherein the analysis result comprises animal habitat distribution, behavior rhythm or population distribution.
7. A wild animal image processing method based on big data and deep learning is characterized by comprising the following steps:
acquiring historical images from a historical image database, and training according to the historical images to obtain an image recognition analysis model;
acquiring an original image from an original image database, and performing image reconstruction on the original image based on a super-coding resolution image recovery technology to obtain an image to be recognized;
adopting the image identification analysis model to perform identification analysis on the image to be identified to obtain an identification result and an analysis result;
and storing the identification result into an image classification database, and storing the analysis result into an analysis result database.
8. The method of claim 7, wherein the image reconstructing the original image comprises:
image quality grading: based on the error rate in image transmission, performing quality grade division on the original image to obtain a low-quality image and an invalid sample image, deleting the invalid sample image, and storing the invalid sample image to an invalid image database to be deleted in the image classification database;
low-quality image preprocessing: performing pixel expansion preprocessing on the low-quality image;
image enhancement processing: improving the definition of the preprocessed low-quality image in an image sharpening mode, and carrying out denoising treatment;
and (3) image recovery processing: and performing super-pixel segmentation on the low-quality image subjected to image enhancement processing to obtain a sparse depth map, and performing image recovery on the sparse depth map.
9. The method of claim 8, wherein the method further comprises:
if the identification result picture does not contain the wild animal image, marking the picture as a false trigger image and transferring the false trigger image into an invalid image database to be deleted in the image classification database;
if the wild animal image is identified according to the identification result and the result precision is more than or equal to the set threshold value, transferring the wild animal image into a sub-database of the label corresponding to the wild animal image database set in the image classification database according to the identification result;
and if the wild animal image is identified and the result precision is less than the set threshold, transferring the wild animal image into an image database to be processed in the image classification database.
10. An electronic device comprising a processor, an input device, an output device, and a memory, the processor, the input device, the output device, and the memory being interconnected, wherein the memory is configured to store a computer program comprising program instructions, the processor being configured to invoke the program instructions to perform the method of any of claims 7-9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011442143.XA CN112419202B (en) | 2020-12-11 | 2020-12-11 | Automatic wild animal image recognition system based on big data and deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011442143.XA CN112419202B (en) | 2020-12-11 | 2020-12-11 | Automatic wild animal image recognition system based on big data and deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112419202A true CN112419202A (en) | 2021-02-26 |
CN112419202B CN112419202B (en) | 2023-09-26 |
Family
ID=74775891
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011442143.XA Active CN112419202B (en) | 2020-12-11 | 2020-12-11 | Automatic wild animal image recognition system based on big data and deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112419202B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113283306A (en) * | 2021-04-30 | 2021-08-20 | 青岛云智环境数据管理有限公司 | Rodent identification and analysis method based on deep learning and transfer learning |
CN113573140A (en) * | 2021-07-09 | 2021-10-29 | 西安交通大学 | Code rate self-adaptive decision-making method supporting face detection and real-time super-resolution |
CN113887460A (en) * | 2021-10-12 | 2022-01-04 | 泰州市雷信农机电制造有限公司 | Block chain application platform for pass verification |
CN114429309A (en) * | 2022-02-08 | 2022-05-03 | 闫雪 | Regional data detection platform based on artificial intelligence |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101783961A (en) * | 2010-03-05 | 2010-07-21 | 厦门大学 | Underwater video image transmission control method based on perception quality |
CN104796718A (en) * | 2015-04-23 | 2015-07-22 | 南京工程学院 | CS (compressed sensing) distributed type based video coding transmission method in WMSN (wireless multimedia sensor network) |
CN105074789A (en) * | 2013-04-09 | 2015-11-18 | 热成像雷达有限责任公司 | Fire detection system |
CN110598461A (en) * | 2019-09-27 | 2019-12-20 | 腾讯科技(深圳)有限公司 | Wild animal information management method, device, terminal, system and storage medium |
-
2020
- 2020-12-11 CN CN202011442143.XA patent/CN112419202B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101783961A (en) * | 2010-03-05 | 2010-07-21 | 厦门大学 | Underwater video image transmission control method based on perception quality |
CN105074789A (en) * | 2013-04-09 | 2015-11-18 | 热成像雷达有限责任公司 | Fire detection system |
CN104796718A (en) * | 2015-04-23 | 2015-07-22 | 南京工程学院 | CS (compressed sensing) distributed type based video coding transmission method in WMSN (wireless multimedia sensor network) |
CN110598461A (en) * | 2019-09-27 | 2019-12-20 | 腾讯科技(深圳)有限公司 | Wild animal information management method, device, terminal, system and storage medium |
Non-Patent Citations (2)
Title |
---|
刘文定 等: ""基于ROI-CNN的赛罕乌拉国家级自然保护区陆生野生动物自动识别"", 《北京林业大学学报》 * |
谢琳 等: ""自然保护区海量视频数据快速分析与检索"", 《计算机系统应用》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113283306A (en) * | 2021-04-30 | 2021-08-20 | 青岛云智环境数据管理有限公司 | Rodent identification and analysis method based on deep learning and transfer learning |
CN113573140A (en) * | 2021-07-09 | 2021-10-29 | 西安交通大学 | Code rate self-adaptive decision-making method supporting face detection and real-time super-resolution |
CN113573140B (en) * | 2021-07-09 | 2022-05-03 | 西安交通大学 | Code rate self-adaptive decision-making method supporting face detection and real-time super-resolution |
CN113887460A (en) * | 2021-10-12 | 2022-01-04 | 泰州市雷信农机电制造有限公司 | Block chain application platform for pass verification |
CN114429309A (en) * | 2022-02-08 | 2022-05-03 | 闫雪 | Regional data detection platform based on artificial intelligence |
CN114429309B (en) * | 2022-02-08 | 2022-11-08 | 深圳市润信数据技术有限公司 | Regional data detection platform based on artificial intelligence |
Also Published As
Publication number | Publication date |
---|---|
CN112419202B (en) | 2023-09-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110059694B (en) | Intelligent identification method for character data in complex scene of power industry | |
US11429818B2 (en) | Method, system and device for multi-label object detection based on an object detection network | |
CN112419202B (en) | Automatic wild animal image recognition system based on big data and deep learning | |
CN111260055B (en) | Model training method based on three-dimensional image recognition, storage medium and device | |
US7983486B2 (en) | Method and apparatus for automatic image categorization using image texture | |
CN107784288B (en) | Iterative positioning type face detection method based on deep neural network | |
CN111291637A (en) | Face detection method, device and equipment based on convolutional neural network | |
CN110781756A (en) | Urban road extraction method and device based on remote sensing image | |
CN111126359A (en) | High-definition image small target detection method based on self-encoder and YOLO algorithm | |
CN112232371B (en) | American license plate recognition method based on YOLOv3 and text recognition | |
CN111652869B (en) | Slab void identification method, system, medium and terminal based on deep learning | |
CN111931931A (en) | Deep neural network training method and device for pathology full-field image | |
CN111091101A (en) | High-precision pedestrian detection method, system and device based on one-step method | |
CN112861970A (en) | Fine-grained image classification method based on feature fusion | |
CN111738164B (en) | Pedestrian detection method based on deep learning | |
CN114037640A (en) | Image generation method and device | |
CN116645592A (en) | Crack detection method based on image processing and storage medium | |
CN117422936B (en) | Remote sensing image classification method and system | |
CN114548606A (en) | Construction method of cyclone strength prediction model and cyclone strength prediction method | |
CN114494786A (en) | Fine-grained image classification method based on multilayer coordination convolutional neural network | |
CN112528058B (en) | Fine-grained image classification method based on image attribute active learning | |
WO2024078112A1 (en) | Method for intelligent recognition of ship outfitting items, and computer device | |
CN114494441B (en) | Grape and picking point synchronous identification and positioning method and device based on deep learning | |
CN116612280A (en) | Vehicle segmentation method, device, computer equipment and computer readable storage medium | |
CN116310832A (en) | Remote sensing image processing method, device, equipment, medium and product |
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 |