CN110837768A - Rare animal protection oriented online detection and identification method - Google Patents
Rare animal protection oriented online detection and identification method Download PDFInfo
- Publication number
- CN110837768A CN110837768A CN201810959642.2A CN201810959642A CN110837768A CN 110837768 A CN110837768 A CN 110837768A CN 201810959642 A CN201810959642 A CN 201810959642A CN 110837768 A CN110837768 A CN 110837768A
- Authority
- CN
- China
- Prior art keywords
- image
- animal
- features
- edge
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06V10/267—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 by performing operations on regions, e.g. growing, shrinking or watersheds
-
- 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
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
Abstract
The invention relates to an on-line detection and identification method for rare animal protection, which firstly uses a traditional image processing method to carry out first fuzzy classification on animal images, reduces the identification range, carries out second accurate classification in a smaller range according to the characteristics of specific animal species, adopts a deep learning method for aquatic animals and birds, adopts a traditional image processing method for non-birds, simplifies the algorithm through two classifications and improves the identification speed. The invention can simplify the identification process as much as possible, obtains the identification result by simple classification for many times, overcomes the defects that the prior method can not extract abstract features and needs a large number of training samples, reduces the computational force requirement as much as possible under the condition of meeting the requirements of identification speed and accuracy, thereby reducing the hardware configuration, being used for realizing image processing on front-end equipment, and being unnecessary to transmit the acquired image information to a back-end computer for processing, and having simpler and quicker realization mode.
Description
Technical Field
The invention belongs to the field of image recognition, and particularly relates to a method for recognizing rare animal species by front-end embedded intelligent equipment.
Background
Wild animals are important components of a natural ecosystem, but due to excessive development and utilization of the natural ecosystem by human beings, the areas of forests, grasslands and wetlands are reduced, the living space of the wild animals is continuously reduced, the number of part of animals is sharply reduced, and the types of the animals are less and less. In the context of this type of background,
people begin to take measures to protect rare animals, wherein an effective means is to establish a rare animal field monitoring station, the monitoring station generally uses an infrared camera to collect data such as front environment data, rare object images and the like, then transmits the information to a server, and the server processes the image data, so that the types of the rare animals can be identified, and the protection is convenient. In the method, the collected real-time image data needs to be transmitted and processed in the identification process, and transmission delay is generated when the image data is transmitted to the server, so that the information of related animal species cannot be captured and identified in real time; in addition, the image data has huge information amount, and has higher requirements on a network for transmitting data, so the image data is more easily influenced by network environment fluctuation, and offline identification cannot be realized. However, if the rare animal species is identified on the front-end embedded intelligent device, the image data can be collected and simultaneously directly processed and identified by the front-end device without being uploaded to a server for processing, so that the transmission delay is reduced, the off-line identification can be realized, and only the identified result needs to be transmitted to the server, the requirement on a transmission network is low, so that the rare animal species is identified on the front-end embedded intelligent device, which is a better scheme.
In the aspect of identifying rare animal species, image-based identification methods are increasingly used for identifying animal species, and generally traditional image classification techniques are adopted in the animal species identification methods, and the animal species identification methods are classified and identified by extracting characteristics such as shapes, colors, textures and the like, but the method is difficult to extract more abstract characteristics, and is difficult to accurately distinguish rare animals with unobvious color distinction, similar appearances and topologically changeable animal body surface textures. The method needs less computing resources and can be operated on the front-end embedded intelligent equipment, but the accuracy rate of identifying the animal species is low, and the requirement of real-time identification cannot be met. In recent years, deep learning methods are also used for animal species identification, which can automatically extract abstract features of animal images, but such methods require a large number of high-quality samples for training, require more computing resources, and some rare animals have fewer samples, so that neural networks cannot be effectively trained, and deep learning is not necessary for species whose shape and color features are obviously different from those of other species of animals. The method can realize the requirement of accurately identifying the animal types, but the training database is established for all the animal types, the required computing resources are huge, and the method cannot be realized on the front-end embedded intelligent equipment.
In summary, in order to achieve the purpose of accurately and rapidly identifying the rare animal species and to be implemented on the front-end embedded smart device, a new method, that is, an animal species identification method combining the conventional image processing method and the deep learning technology, is required. The method has the advantages of less required computing resources and high identification accuracy, and can be realized on front-end embedded intelligent equipment, so that the rare animal species can be accurately and quickly identified.
Disclosure of Invention
The technical scheme of the invention comprises the following steps:
an on-line detection and identification method for rare animal protection is characterized by comprising
Step 1: acquiring data, specifically acquiring an image sequence acquired by a monitoring station in real time, inputting image information of each frame, acquiring an image when no animal appears in most of time, calculating a gray average value of a background environment, and when the gray average value of the image changes, indicating that an animal appears, and acquiring image data of the animal;
step 2: performing image preprocessing, specifically, performing brightness correction and geometric transformation on image data of an animal, removing the influence of illumination in the image acquisition process, and then filtering and denoising to filter interference and noise; using an OSTU algorithm method to improve and finish the selection of a threshold value of a binary image, then using a discriminant analysis method to divide a set of gray levels into two groups by using the threshold value, and preprocessing the image;
and step 3: establishing an edge detection model, specifically, solving edge points by using a Canny operator according to the characteristics of image gray distribution, carrying out image segmentation on an image by using a self-adaptive adjacent domain, removing background information, extracting the edge of the image to obtain an interested region and an animal contour, smoothing the image by using a Gaussian filter, calculating the gradient amplitude and direction by using a first-order partial derivative finite difference, carrying out non-maximum suppression on the gradient amplitude, detecting and connecting the edge by using a dual-threshold algorithm, and finally finishing the edge detection on the image;
and 4, step 4: establishing a classification model based on the edge detection model obtained in the step 3, specifically, roughly classifying images from animal contours by using shape feature detection according to contour boundary features obtained in the step 3, wherein the images are terrestrial animals with the features of legs and the like, and the images are aquatic animals without the features of legs and the like, and the terrestrial animals are divided into birds and non-birds through the appearance features of wings and the like;
and 5: training an AlexNet network by using image samples of aquatic animals and birds, constructing a label index of the number of labels transmitted by a data iterator, constructing the number of transmission steps of the training iterator, and constructing a deep convolution network of the aquatic animals and the birds by updating and continuously adjusting network parameters through multiple iterations; inputting image data into a neural network, accelerating the identification process of the image data through a neural calculation acceleration rod, and outputting animal species information;
step 6: for non-bird image data, extracting appearance features by using an HOG algorithm, extracting color features by using a color histogram method, extracting texture features by using a model method, obtaining feature vectors including the appearance features, the color features and the texture features and used for representing animal categories, classifying the feature vectors by using a mathematical method including clustering and the like, and outputting animal species information.
In the above method for detecting and identifying rare animal protection on line, step 3 specifically includes
Step 3.1, changing the color image into a Gray image by using a formula Gray of 0.299R +0.587G +0.114B, and performing convolution on the image data through a two-dimensional Gaussian kernel to complete Gaussian filtering and realize smooth image;
step 3.2, calculating the difference Gx and Gy in the horizontal direction and the vertical direction by using an edge difference operator to obtain the gradient amplitude and the gradient direction, comparing the gradient values in front of and behind the pixel points along the gradient direction, searching the local maximum value of the pixel points, and carrying out non-maximum value inhibition;
3.3, distinguishing edge pixels by using a high threshold and a low threshold, and marking the edge pixels as strong edge points if the gradient value of the edge pixels is greater than the high threshold; if the edge gradient value is smaller than the high threshold value and larger than the low threshold value, marking as a weak edge point; and inhibiting the points smaller than the low threshold value, marking the strong edge points as image edges, and finishing the edge detection of the image.
In the above method for on-line detection and identification of rare animal protection, step 4, fourier transform is performed on the boundary of an object by using a fourier shape descriptor method, three shape expressions of a curvature function, a centroid distance and a complex coordinate function are derived from boundary points, so as to obtain feature vectors of non-birds, aquatic animals and birds, and a classifier is established, which specifically includes:
step 4.1, assigning a region position function to each pixel node of the image of the non-bird animal, minimizing a cost function, and simultaneously enabling the path of each pixel node to be shortest, so that an image segmentation result is obtained, and performing morphological processing such as expansion corrosion and the like;
step 4.2, extracting color features of the image by using an HOG algorithm, measuring distances among the color histogram features by adopting a calculation simple histogram intersection algorithm, extracting the color features, describing texture formation by using an MRF model distribution model, and estimating parameters of the distribution model from the realization of the texture image to obtain the texture features;
and 4.3, finding out the optimal classification hyperplane of each class of feature sample and other feature samples in the sample feature space by using the SVM, obtaining a support vector set representing the features of each sample and the corresponding VC (VC) reliability thereof, forming a discriminant function for judging each feature class, establishing an SVM (support vector machine) classifier, and determining classifier parameters for identifying the type of animal.
In the above method for detecting and identifying rare animal protection on line, the step 5 specifically includes
Step 5.1, training an AlexNet network by using image samples of animals such as aquatic animals and birds, generating an array passing through shuffle from each type of animal image, determining the number and arrangement mode of convolution layers and downsampling layers of a neural network, adjusting the size of a sliding window in a pooling layer, setting a data iterator and a training iterator, and performing data training;
and 5.2, constructing a single-layer neural network layer by layer, training a single-layer network each time, adjusting weight coefficients by using a wake-sleep algorithm after training of all the single-layer networks is finished, setting corresponding deep neural network structures, training all the deep neural network structures to identify animal species, comparing the performances of the deep neural network structures, including identification speed and accuracy, selecting optimal network adjustment parameters and improving the network structures, and obtaining an animal species identification deep neural network model.
The invention can simplify the identification process as much as possible, obtains the identification result by simple classification for many times, overcomes the defects that the prior method can not extract abstract features and needs a large number of training samples, reduces the computational force requirement as much as possible under the condition of meeting the requirements of identification speed and accuracy, thereby reducing the hardware configuration, being used for realizing image processing on front-end equipment, and being unnecessary to transmit the acquired image information to a back-end computer for processing, and having simpler and quicker realization mode.
The invention fully absorbs and utilizes the advantages of the traditional image processing method and the deep learning method, firstly uses the traditional image processing method to carry out the first fuzzy classification on the animal image, reduces the identification range, carries out the second accurate classification in a smaller range according to the characteristics of specific animal species, adopts the deep learning method for aquatic animals and birds, adopts the traditional image processing method for non-birds, simplifies the algorithm through two classifications and improves the identification speed.
Drawings
Fig. 1 is a flowchart of a method for identifying an animal species according to the present invention.
Detailed Description
The invention is used for identifying animal species, combines the traditional image processing method with the deep learning technology, can realize the flow on the front-end equipment, and carries out accurate and rapid identification, and the invention is further explained by combining the attached drawings and the embodiment.
Step 1: and (6) data acquisition. The method comprises the steps of collecting image data of each type of animal by using a camera module, wherein the quantity of collected images of each type of animal is at least 100, the outline of the animal in the images is required to be obvious, the images of the type of animal taken from various angles are included as much as possible, and a sample training library of each type of animal is established.
Step 2: and (5) image preprocessing. The method comprises the steps of carrying out brightness correction and geometric transformation on image data of animals, removing the influence of illumination in the image acquisition process, then using a vector median filtering algorithm to obtain an image gray median, and using median filtering to remove noise so as to eliminate isolated noise points. And (3) improving by using an OSTU algorithm to complete the selection of a threshold value of a binary image, dividing the image into a background part and a target part, dividing a set of gray levels into two groups by using a threshold value by using a discriminant analysis method, and preprocessing the image.
And step 3: and establishing an edge detection and classification module. For images of non-birds, aquatic animals and birds, a formula Gray is 0.299R +0.587G +0.114B to change a color image into a Gray image, a two-dimensional Gaussian kernel is convoluted with image data to complete Gaussian filtering to realize smooth image, an edge difference operator is used for calculating difference Gx and Gy in horizontal and vertical directions to obtain gradient amplitude and direction, gradient values in front of and behind the gradient operator are compared along the gradient direction to find local maximum values of pixel points and inhibit non-maximum values, a high threshold value and a low threshold value are used for distinguishing edge pixels, and if the gradient value of the edge pixel points is larger than the high threshold value, the edge pixels are marked as strong edge points. If the edge gradient value is less than the high threshold and greater than the low threshold, then it is marked as a weak edge point. And inhibiting points smaller than the low threshold value, marking strong edge points as image edges, finally completing the edge detection of the image, performing Fourier transform on the object boundary by using a Fourier shape descriptor method, deriving three shape expressions of a curvature function, a centroid distance and a complex coordinate function from the boundary points, obtaining the feature vectors of the non-birds, the aquatic animals and the birds, and establishing a classifier.
And 4, step 4: assigning a region location function to each pixel node of the image of the non-avian animal, minimizing a cost function, at the same time, the path of each pixel node is shortest, so as to obtain the image segmentation result, and make morphological treatments of expansion corrosion, etc., then, the HOG algorithm is used for extracting the color features of the image, the calculation simple histogram intersection algorithm is used for measuring the distance between the color histogram features, the color features are extracted, the MRF model distribution model is used for describing texture formation, the parameters of the distribution model are estimated from the realization of the texture image to obtain the texture features, an SVM is used for finding out the optimal classification hyperplane of each class of feature samples and other feature samples in the sample feature space to obtain a support vector set representing each sample feature and the corresponding VC credibility thereof, a discriminant function for judging each feature class is formed, an SVM classifier is established, and the classifier parameters for identifying the type of animal are determined.
And 5: the AlexNet network is trained by using image samples of animals such as aquatic animals and birds, an array passing through shuffle is generated from images of each type of animals, the number and arrangement mode of convolution layers and down-sampling layers of a neural network are determined, the size of a sliding window in a pooling layer is adjusted, a data iterator and a training iterator are arranged for data training, a single-layer neural network is constructed layer by layer, a single-layer network is trained each time, after all single-layer network training is finished, a wake-sleep algorithm is used for adjusting weight coefficients, a corresponding deep neural network structure is designed, the animal type is trained and recognized, the performance of the deep neural network model is compared, including recognition speed, accuracy and the like, the optimal network adjusting parameters are selected, the network structure is improved, and a more applicable animal type recognition deep neural network model is obtained.
Step 6: and (5) training a model. The method comprises the steps of testing a designed deep neural network model for 500 times by using an image acquired in real time, outputting a recognition result to obtain the recognition accuracy, increasing the training amount of a sample for a model which cannot meet the requirement that the recognition accuracy is more than 98%, adjusting the learning rate, increasing the iteration times to enable each step of training to iterate more times, adjusting a loss function and providing the recognition accuracy.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.
Claims (4)
1. An on-line detection and identification method for rare animal protection is characterized by comprising
Step 1: acquiring data, specifically acquiring an image sequence acquired by a monitoring station in real time, inputting image information of each frame, acquiring an image when no animal appears in most of time, calculating a gray average value of a background environment, and when the gray average value of the image changes, indicating that an animal appears, and acquiring image data of the animal;
step 2: performing image preprocessing, specifically, performing brightness correction and geometric transformation on image data of an animal, removing the influence of illumination in the image acquisition process, and then filtering and denoising to filter interference and noise; using an OSTU algorithm method to improve and finish the selection of a threshold value of a binary image, then using a discriminant analysis method to divide a set of gray levels into two groups by using the threshold value, and preprocessing the image;
and step 3: establishing an edge detection model, specifically, solving edge points by using a Canny operator according to the characteristics of image gray distribution, carrying out image segmentation on an image by using a self-adaptive adjacent domain, removing background information, extracting the edge of the image to obtain an interested region and an animal contour, smoothing the image by using a Gaussian filter, calculating the gradient amplitude and direction by using a first-order partial derivative finite difference, carrying out non-maximum suppression on the gradient amplitude, detecting and connecting the edge by using a dual-threshold algorithm, and finally finishing the edge detection on the image;
and 4, step 4: establishing a classification model based on the edge detection model obtained in the step 3, specifically, roughly classifying images from animal contours by using shape feature detection according to contour boundary features obtained in the step 3, wherein the images are terrestrial animals with the features of legs and the like, and the images are aquatic animals without the features of legs and the like, and the terrestrial animals are divided into birds and non-birds through the appearance features of wings and the like;
and 5: training an AlexNet network by using image samples of aquatic animals and birds, constructing a label index of the number of labels transmitted by a data iterator, constructing the number of transmission steps of the training iterator, and constructing a deep convolution network of the aquatic animals and the birds by updating and continuously adjusting network parameters through multiple iterations; inputting image data into a neural network, accelerating the identification process of the image data through a neural calculation acceleration rod, and outputting animal species information;
step 6: for non-bird image data, extracting appearance features by using an HOG algorithm, extracting color features by using a color histogram method, extracting texture features by using a model method, obtaining feature vectors including the appearance features, the color features and the texture features and used for representing animal categories, classifying the feature vectors by using a mathematical method including clustering and the like, and outputting animal species information.
2. The method for detecting and identifying rare animal protection on line as claimed in claim 1, wherein the step 3 specifically comprises
Step 3.1, changing the color image into a Gray image by using a formula Gray of 0.299R +0.587G +0.114B, and performing convolution on the image data through a two-dimensional Gaussian kernel to complete Gaussian filtering and realize smooth image;
step 3.2, calculating the difference Gx and Gy in the horizontal direction and the vertical direction by using an edge difference operator to obtain the gradient amplitude and the gradient direction, comparing the gradient values in front of and behind the pixel points along the gradient direction, searching the local maximum value of the pixel points, and carrying out non-maximum value inhibition;
3.3, distinguishing edge pixels by using a high threshold and a low threshold, and marking the edge pixels as strong edge points if the gradient value of the edge pixels is greater than the high threshold; if the edge gradient value is smaller than the high threshold value and larger than the low threshold value, marking as a weak edge point; and inhibiting the points smaller than the low threshold value, marking the strong edge points as image edges, and finishing the edge detection of the image.
3. The rare animal protection-oriented online detection and identification method according to claim 1, wherein the step 4 is to perform fourier transform on the object boundary by using a fourier shape descriptor method, derive three shape expressions of a curvature function, a centroid distance and a complex coordinate function from the boundary points, obtain feature vectors of non-birds, aquatic animals and birds, and establish a classifier, and specifically comprises:
step 4.1, assigning a region position function to each pixel node of the image of the non-bird animal, minimizing a cost function, and simultaneously enabling the path of each pixel node to be shortest, so that an image segmentation result is obtained, and performing morphological processing such as expansion corrosion and the like;
step 4.2, extracting color features of the image by using an HOG algorithm, measuring distances among the color histogram features by adopting a calculation simple histogram intersection algorithm, extracting the color features, describing texture formation by using an MRF model distribution model, and estimating parameters of the distribution model from the realization of the texture image to obtain the texture features;
and 4.3, finding out the optimal classification hyperplane of each class of feature sample and other feature samples in the sample feature space by using the SVM, obtaining a support vector set representing the features of each sample and the corresponding VC (VC) reliability thereof, forming a discriminant function for judging each feature class, establishing an SVM (support vector machine) classifier, and determining classifier parameters for identifying the type of animal.
4. The method for detecting and identifying rare animal protection on line as claimed in claim 1, wherein the step 5 specifically comprises
Step 5.1, training an AlexNet network by using image samples of animals such as aquatic animals and birds, generating an array passing through shuffle from each type of animal image, determining the number and arrangement mode of convolution layers and downsampling layers of a neural network, adjusting the size of a sliding window in a pooling layer, setting a data iterator and a training iterator, and performing data training;
and 5.2, constructing a single-layer neural network layer by layer, training a single-layer network each time, adjusting weight coefficients by using a wake-sleep algorithm after training of all the single-layer networks is finished, setting corresponding deep neural network structures, training all the deep neural network structures to identify animal species, comparing the performances of the deep neural network structures, including identification speed and accuracy, selecting optimal network adjustment parameters and improving the network structures, and obtaining an animal species identification deep neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810959642.2A CN110837768B (en) | 2018-08-16 | 2018-08-16 | Online detection and identification method for rare animal protection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810959642.2A CN110837768B (en) | 2018-08-16 | 2018-08-16 | Online detection and identification method for rare animal protection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110837768A true CN110837768A (en) | 2020-02-25 |
| CN110837768B CN110837768B (en) | 2023-06-20 |
Family
ID=69574511
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810959642.2A Active CN110837768B (en) | 2018-08-16 | 2018-08-16 | Online detection and identification method for rare animal protection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110837768B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112036280A (en) * | 2020-08-24 | 2020-12-04 | 方海涛 | Waterfowl population dynamic monitoring method, device and equipment |
| CN112101301A (en) * | 2020-11-03 | 2020-12-18 | 武汉工程大学 | Good sound stability early warning method and device for screw water cooling unit and storage medium |
| CN112241466A (en) * | 2020-09-22 | 2021-01-19 | 天津永兴泰科技股份有限公司 | Wild animal protection law recommendation system based on animal identification map |
| CN112288793A (en) * | 2020-11-06 | 2021-01-29 | 洛阳语音云创新研究院 | Livestock individual backfat detection method and device, electronic equipment and storage medium |
| CN112612914A (en) * | 2020-12-29 | 2021-04-06 | 浙江金实乐环境工程有限公司 | Image garbage recognition method based on deep learning |
| CN112630729A (en) * | 2020-12-11 | 2021-04-09 | 杭州博镨科技有限公司 | Method for positioning and tracking indoor human target based on thermopile sensor |
| CN112668646A (en) * | 2020-12-28 | 2021-04-16 | 扬州市玉器产品质量监督检验中心 | Jade and jewelry identification quality tracing management method and system |
| 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 |
| CN113627501A (en) * | 2021-07-30 | 2021-11-09 | 武汉大学 | Animal image type identification method based on transfer learning |
| CN115982534A (en) * | 2023-03-18 | 2023-04-18 | 湖北一方科技发展有限责任公司 | Processing method of river hydrological monitoring data |
| CN117611885A (en) * | 2023-11-17 | 2024-02-27 | 贵州省生物研究所 | Waiting bird ecological regulation and control method based on Canny edge detection |
| CN112630729B (en) * | 2020-12-11 | 2024-05-07 | 杭州博镨科技有限公司 | Method for positioning and tracking indoor human target based on thermopile sensor |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104036235A (en) * | 2014-05-27 | 2014-09-10 | 同济大学 | Plant species identification method based on leaf HOG features and intelligent terminal platform |
| CN106096531A (en) * | 2016-05-31 | 2016-11-09 | 安徽省云力信息技术有限公司 | A kind of traffic image polymorphic type vehicle checking method based on degree of depth study |
| CN106290359A (en) * | 2016-07-22 | 2017-01-04 | 南京农业大学 | A kind of method of the lossless classification of apple crisp slices quality |
| CN107180241A (en) * | 2017-04-20 | 2017-09-19 | 华南理工大学 | A kind of animal classification method of the profound neutral net based on Gabor characteristic with fractal structure |
| CN107292891A (en) * | 2017-06-20 | 2017-10-24 | 华南农业大学 | A kind of detection method of counting of the southern vegetables Severe pests based on machine vision |
| US20180032844A1 (en) * | 2015-03-20 | 2018-02-01 | Intel Corporation | Object recognition based on boosting binary convolutional neural network features |
| CN108182423A (en) * | 2018-01-26 | 2018-06-19 | 山东科技大学 | A kind of poultry Activity recognition method based on depth convolutional neural networks |
-
2018
- 2018-08-16 CN CN201810959642.2A patent/CN110837768B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104036235A (en) * | 2014-05-27 | 2014-09-10 | 同济大学 | Plant species identification method based on leaf HOG features and intelligent terminal platform |
| US20180032844A1 (en) * | 2015-03-20 | 2018-02-01 | Intel Corporation | Object recognition based on boosting binary convolutional neural network features |
| CN106096531A (en) * | 2016-05-31 | 2016-11-09 | 安徽省云力信息技术有限公司 | A kind of traffic image polymorphic type vehicle checking method based on degree of depth study |
| CN106290359A (en) * | 2016-07-22 | 2017-01-04 | 南京农业大学 | A kind of method of the lossless classification of apple crisp slices quality |
| CN107180241A (en) * | 2017-04-20 | 2017-09-19 | 华南理工大学 | A kind of animal classification method of the profound neutral net based on Gabor characteristic with fractal structure |
| CN107292891A (en) * | 2017-06-20 | 2017-10-24 | 华南农业大学 | A kind of detection method of counting of the southern vegetables Severe pests based on machine vision |
| CN108182423A (en) * | 2018-01-26 | 2018-06-19 | 山东科技大学 | A kind of poultry Activity recognition method based on depth convolutional neural networks |
Non-Patent Citations (4)
| Title |
|---|
| HUNG NY ET AL: "《Animal Recognition and Identification with Deep Convolutional Neural Networks for Automated Wildlife Monitoring》" * |
| 乔小燕: "《基于生物形态学的赤潮藻显微图像分割与特征提取研究》", 《中国博士学位论文全文数据库 信息科技辑》 * |
| 姚青等: "《基于图像的昆虫自动识别与计数研究进展》", 《中国农业科学》 * |
| 徐文韬: "《运动目标跟踪检测与识别关键算法的研究与实现》", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112036280A (en) * | 2020-08-24 | 2020-12-04 | 方海涛 | Waterfowl population dynamic monitoring method, device and equipment |
| CN112241466A (en) * | 2020-09-22 | 2021-01-19 | 天津永兴泰科技股份有限公司 | Wild animal protection law recommendation system based on animal identification map |
| CN112101301A (en) * | 2020-11-03 | 2020-12-18 | 武汉工程大学 | Good sound stability early warning method and device for screw water cooling unit and storage medium |
| CN112288793A (en) * | 2020-11-06 | 2021-01-29 | 洛阳语音云创新研究院 | Livestock individual backfat detection method and device, electronic equipment and storage medium |
| CN112630729A (en) * | 2020-12-11 | 2021-04-09 | 杭州博镨科技有限公司 | Method for positioning and tracking indoor human target based on thermopile sensor |
| CN112630729B (en) * | 2020-12-11 | 2024-05-07 | 杭州博镨科技有限公司 | Method for positioning and tracking indoor human target based on thermopile sensor |
| CN112668646A (en) * | 2020-12-28 | 2021-04-16 | 扬州市玉器产品质量监督检验中心 | Jade and jewelry identification quality tracing management method and system |
| CN112612914A (en) * | 2020-12-29 | 2021-04-06 | 浙江金实乐环境工程有限公司 | Image garbage recognition method based on deep learning |
| 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 |
| CN113627501A (en) * | 2021-07-30 | 2021-11-09 | 武汉大学 | Animal image type identification method based on transfer learning |
| CN115982534A (en) * | 2023-03-18 | 2023-04-18 | 湖北一方科技发展有限责任公司 | Processing method of river hydrological monitoring data |
| CN117611885A (en) * | 2023-11-17 | 2024-02-27 | 贵州省生物研究所 | Waiting bird ecological regulation and control method based on Canny edge detection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110837768B (en) | 2023-06-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110837768B (en) | Online detection and identification method for rare animal protection | |
| CN107657279B (en) | Remote sensing target detection method based on small amount of samples | |
| CN111325764B (en) | Fruit image contour recognition method | |
| CN104834922B (en) | Gesture identification method based on hybrid neural networks | |
| CN106709950B (en) | Binocular vision-based inspection robot obstacle crossing wire positioning method | |
| CN107292252B (en) | Identity recognition method for autonomous learning | |
| CN109919960B (en) | Image continuous edge detection method based on multi-scale Gabor filter | |
| CN105913081B (en) | SAR image classification method based on improved PCAnet | |
| CN103942577A (en) | Identity identification method based on self-established sample library and composite characters in video monitoring | |
| CN106778474A (en) | 3D human body recognition methods and equipment | |
| CN111340824A (en) | Image feature segmentation method based on data mining | |
| CN111460953B (en) | Electrocardiosignal classification method based on self-adaptive learning of countermeasure domain | |
| CN110728302A (en) | Method for identifying color textile fabric tissue based on HSV (hue, saturation, value) and Lab (Lab) color spaces | |
| CN106529441B (en) | Depth motion figure Human bodys' response method based on smeared out boundary fragment | |
| CN106611158A (en) | Method and equipment for obtaining human body 3D characteristic information | |
| CN108154176B (en) | 3D human body posture estimation algorithm aiming at single depth image | |
| CN109858438A (en) | A kind of method for detecting lane lines based on models fitting | |
| CN106778491B (en) | The acquisition methods and equipment of face 3D characteristic information | |
| CN109448024B (en) | Visual tracking method and system for constructing constraint correlation filter by using depth data | |
| CN108898621B (en) | Related filtering tracking method based on instance perception target suggestion window | |
| CN109508674A (en) | Airborne lower view isomery image matching method based on region division | |
| CN116386118B (en) | Drama matching cosmetic system and method based on human image recognition | |
| Zhao et al. | Automatic sweet pepper detection based on point cloud images using subtractive clustering | |
| CN116342653A (en) | Target tracking method, system, equipment and medium based on correlation filter | |
| Quraishi et al. | A novel approach for face detection using artificial neural network |
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 |