CN111460864B - Animal disease detection method based on image recognition - Google Patents
Animal disease detection method based on image recognition Download PDFInfo
- Publication number
- CN111460864B CN111460864B CN201910058338.5A CN201910058338A CN111460864B CN 111460864 B CN111460864 B CN 111460864B CN 201910058338 A CN201910058338 A CN 201910058338A CN 111460864 B CN111460864 B CN 111460864B
- Authority
- CN
- China
- Prior art keywords
- image
- animal
- similarity
- images
- comparison library
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
-
- 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
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
Abstract
An animal disease detection method based on image recognition is divided into three parts, wherein the first part is a construction comparison library, one animal construction comparison library is selected, and different comparison libraries can be constructed according to different animals; the second part is to process the target animal, including pretreatment, animal feature extraction, pixel conversion and image gray information statistics; the third part is to calculate the similarity of the images, and judge whether the disease exists by comparing the similarity of the extracted part and the corresponding part in the comparison library. The method can realize the monitoring of the health condition of animals; acquiring image information of the animal through a camera, and comparing a series of operations such as preprocessing, feature recognition, image segmentation, image registration and the like with images in a comparison library to judge whether the animal is ill; because the monitoring system can work uninterruptedly, the risk of animal epidemic situation is greatly reduced.
Description
Technical Field
The invention relates to the field of image recognition, in particular to an animal disease detection method based on image recognition.
Background
The main purpose of image preprocessing is to eliminate extraneous information in the image, recover useful real information, enhance the detectability of related information and maximally simplify data, thereby improving the reliability of feature extraction, image segmentation, matching and recognition. The preprocessing process generally includes the steps of digitizing, geometric transformation, normalization, smoothing, restoration, enhancement and the like.
The wavelet moment-based image recognition method is to use invariant moment of an input two-dimensional binary image as a recognition feature, use a BP network for recognition, normalize the input image, polar coordinate the input image, extract the rotation invariant wavelet moment feature, and send the input image into a BP network classifier for recognition to obtain a recognition result. The wavelet moment features have good resolving power on samples with translation, scaling and rotation, and can accurately resolve test samples under the condition of no noise, the recognition rate is superior to that of geometric moment, and the difference reaches 30 percent. With the addition of random noise, the recognition rate of the two moment features is reduced, but the wavelet moment has better capability of extracting the local features of the image, so that the recognition rate of the wavelet moment is reduced relatively slowly, and the highest correct recognition rate reaches 98%.
In terms of image registration, many scholars have made a lot of researches, and the currently mainstream image registration method mainly comprises two methods, namely a method based on gray scale or global features and a method based on local significant features. The method comprises the steps of measuring similarity between images by using gray level information of the whole or part of the images, and further realizing image registration. The latter first extracts salient features from the image and then matches these features to achieve registration of the images. And is widely applied at present. In particular, in point feature-based registration methods, the points that are features should have translational, rotational, and scale invariance, but how to guarantee their rotational and scale invariance characteristics is a difficulty and hotspot of current research; and these points should be able to accurately reflect the image features and should have a high degree of differentiation from each other, which makes the feature description method complicated.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention provides an animal disease detection method based on image recognition, thereby realizing the monitoring of animal health conditions; acquiring image information of the animal through a camera, and comparing a series of operations such as preprocessing, feature recognition, image segmentation, image registration and the like with images in a comparison library to judge whether the animal is ill; because the monitoring system can work uninterruptedly, the risk of animal epidemic situation is greatly reduced.
An animal disease detection method based on image recognition is divided into three parts, wherein the first part is a construction comparison library, one animal construction comparison library is selected, and different comparison libraries can be constructed according to different animals; the second part is to process the target animal, including pretreatment, animal feature extraction, pixel conversion and image gray information statistics; the third part is to calculate the similarity of the images, and judge whether the disease exists by comparing the similarity of the extracted part and the corresponding part in the comparison library. The whole scheme is shown in figure 1.
1 construction of a comparison library
According to the body surface characteristic parts shown by animal diseases, six obvious characteristic parts of the nose, eyes, tail, legs, feet and ears of the animal are selected, and a plurality of pictures in normal and diseased states are included under each characteristic part, wherein the pictures comprise factors affecting the gesture, illumination, body shape and recognition;
2 obtaining characteristic parts of animals
The acquired target image comes from a camera installed in the cultivation place, and the image needs to be subjected to white balance pretreatment before other operation treatment due to the limitation of shooting environment; considering that clear image information is required for extracting the characteristic parts, a high-definition camera with 200 ten thousand pixels, namely 1080P, is adopted in the selection of the monitoring camera; according to the construction of a comparison library, the characteristic parts are effectively identified by adopting an image identification technology based on wavelet moment; in order to realize that the extracted moment has translation, scaling and rotation invariance, firstly, carrying out normalization pretreatment on a sample, namely determining the centroid coordinates of a graph and normalizing pixel points according to the centroid; is processed by polar coordinatePerforming waveletExtracting, namely extracting the characteristics of the object by utilizing rotation invariance of wavelet moment and different scaling factors; inputting the extracted features into a BP neural network classifier for recognition to obtain feature parts of animals;
3 calculating the similarity of the images and determining whether it is ill
According to the registration method of the image gray level, the similarity of the image is measured by using all gray level information of the image; scanning the image by using a search window, and determining whether the animal is ill according to the similarity of the scanned image and the images in the comparison library; given two images I to be compared 1 (x, y) and I 2 (x, y), let CC denote the cross-correlation function between the two, the representation of the function being:
wherein, fatter x and fatter y are the offset in the horizontal direction and the vertical direction respectively; when the value of the function CC reaches the maximum, the similarity measure representing the two images reaches the maximum, and the determined image transformation model is the registration mapping of the images; in the method, each pixel point in the image is traversed in sequence directly from the upper left corner of the extracted feature image, the similarity of the image in the search window and the matching image in gray scale is calculated by using a cross-correlation function, and the matching is judged to be successful when the result is larger than a certain threshold value; meanwhile, in order to improve the robustness of the algorithm, the algorithm does not use individual judgment as a final result, but compares the final result with a plurality of images, and the final judgment result is obtained according to the majority of the occurrence results.
According to the image gray level registration method, the similarity between images is fully utilized, and identification of diseased and healthy animal individuals can be realized. The common gray level registration method usually uses all gray level information to measure the similarity of the images, the design extracts the characteristic parts in the images, and only measures the gray level information of the characteristic parts when the registration is carried out, so that the calculation amount of an algorithm is greatly reduced. In the aspect of feature extraction, a multilayer convolutional neural network is adopted, so that the acquisition of correct feature is fully ensured, and the increase of calculation amount of incorrect feature images is avoided.
Drawings
FIG. 1 is a schematic diagram of an overall design of an animal disease detection method based on image recognition;
FIG. 2 is a flow chart of animal feature acquisition;
fig. 3 is a flow chart of image gray scale registration.
Detailed Description
The present invention will be described in detail below with reference to the drawings and examples, but the scope of the present invention is not limited thereto.
The animal disease detection method based on image recognition is mainly gray level image registration, and the size of a comparison library cannot be too large because of the large calculation amount required by measuring all gray level information in the comparison image. And only one animal is bred in a common farm, so that different comparison libraries are established for different farms. On the establishment of a comparison library, 50 photos of each characteristic part in disease and health state are selected, and 600 comparison samples are taken, so that the identification requirement can be basically met. And a high-definition camera is arranged in the cultivation place, in particular to an animal rest place and an entrance and exit where animals frequently come in and go out, and is used for collecting image information of the animals. The acquired image information is input into an image processing module, as shown in fig. 2, the image is subjected to normalization preprocessing and polar coordination so as to extract wavelet moments, and then S is performed q (r) utilizing wavelet functionsExtracting features in radial region (r is more than or equal to 0 and less than or equal to 1), wherein the specific algorithm is +.>. The extracted features are then fed into a BP neural network, where the output to that node for the class to which the input pattern belongs is set to 1 and all output nodes are set to 0. The gradient descent iteration in the weight space is such that the whole systemThe output of all nodes in the system meets the desired output. And after the characteristic image is obtained, image registration is started, each pixel point in the image is traversed from the upper left corner of the extracted characteristic image in sequence, the similarity of the image in the search window and the matching image in gray scale is calculated by using a cross-correlation function, and the matching is judged to be successful when the result is larger than a certain threshold value. Analyzing the health condition of the animal according to the judging result, and if the acquired image is judged to be similar to the diseased animal part in the comparison library, marking the animal as diseased; otherwise, the animal judging result is normal. The flow chart of image registration is shown in fig. 3.
Claims (1)
1. An animal disease detection method based on image recognition is characterized in that: the method comprises the steps of dividing the method into three parts, wherein the first part is to construct a comparison library, selecting one animal to construct the comparison library, and constructing different comparison libraries according to different animals; the second part is to process the target animal, including pretreatment, animal feature extraction, pixel conversion and image gray information statistics; the third part is to calculate the similarity of the image, and judge whether the disease is caused by comparing the similarity of the extracted part and the corresponding part in the comparison library;
1 construction of a comparison library
According to the body surface characteristic parts shown by animal diseases, six obvious characteristic parts of the nose, eyes, tail, legs, feet and ears of the animal are selected, and a plurality of pictures in normal and diseased states are included under each characteristic part, wherein the pictures comprise factors affecting the gesture, illumination, body shape and recognition;
2 obtaining characteristic parts of animals
The acquired target image comes from a camera installed in the cultivation place, and the image needs to be subjected to white balance pretreatment before other operation treatment due to the limitation of shooting environment; considering that clear image information is required for extracting the characteristic parts, a high-definition camera with 200 ten thousand pixels, namely 1080P, is adopted in the selection of the monitoring camera; according to the construction of a comparison library, adopting a wavelet moment-based image recognition technologyThe sign part is effectively identified; in order to realize that the extracted moment has translation, scaling and rotation invariance, firstly, carrying out normalization pretreatment on a sample, namely determining the centroid coordinates of a graph and normalizing pixel points according to the centroid; is processed by polar coordinateWavelet extraction is carried out, and the feature extraction of the object can be realized by utilizing the rotation invariance of wavelet moment and different scaling factors; inputting the extracted features into a BP neural network classifier for recognition to obtain feature parts of animals;
3 calculating the similarity of the images and determining whether it is ill
According to the registration method of the image gray level, the similarity of the image is measured by using all gray level information of the image; scanning the image by using a search window, and determining whether the animal is ill according to the similarity of the scanned image and the images in the comparison library; given two images I to be compared 1 (x, y) and I 2 (x, y), let CC denote the cross-correlation function between the two, the representation of the function being:
wherein, fatter x and fatter y are the offset in the horizontal direction and the vertical direction respectively; when the value of the function CC reaches the maximum, the similarity measure representing the two images reaches the maximum, and the determined image transformation model is the registration mapping of the images; in the method, each pixel point in the image is traversed in sequence directly from the upper left corner of the extracted feature image, the similarity of the image in the search window and the matching image in gray scale is calculated by using a cross-correlation function, and the matching is judged to be successful when the result is larger than a certain threshold value; meanwhile, in order to improve the robustness of the algorithm, the algorithm does not use individual judgment as a final result, but compares the final result with a plurality of images, and the final judgment result is obtained according to the majority of the occurrence results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910058338.5A CN111460864B (en) | 2019-01-22 | 2019-01-22 | Animal disease detection method based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910058338.5A CN111460864B (en) | 2019-01-22 | 2019-01-22 | Animal disease detection method based on image recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111460864A CN111460864A (en) | 2020-07-28 |
| CN111460864B true CN111460864B (en) | 2023-10-17 |
Family
ID=71679877
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910058338.5A Active CN111460864B (en) | 2019-01-22 | 2019-01-22 | Animal disease detection method based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111460864B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112133391B (en) * | 2020-09-17 | 2024-01-26 | 吾征智能技术(北京)有限公司 | Disease cognitive system based on abnormal eyebrows of human body |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011188108A (en) * | 2010-03-05 | 2011-09-22 | Mitsubishi Electric Corp | Image processing apparatus, three-dimensional image display device, and image processing method |
| CN103886586A (en) * | 2014-02-18 | 2014-06-25 | 南京邮电大学 | Medical image registration method based on combination of mutual information and gradient information |
| CN104318548A (en) * | 2014-10-10 | 2015-01-28 | 西安电子科技大学 | Rapid image registration implementation method based on space sparsity and SIFT feature extraction |
| WO2016110005A1 (en) * | 2015-01-07 | 2016-07-14 | 深圳市唯特视科技有限公司 | Gray level and depth information based multi-layer fusion multi-modal face recognition device and method |
-
2019
- 2019-01-22 CN CN201910058338.5A patent/CN111460864B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011188108A (en) * | 2010-03-05 | 2011-09-22 | Mitsubishi Electric Corp | Image processing apparatus, three-dimensional image display device, and image processing method |
| CN103886586A (en) * | 2014-02-18 | 2014-06-25 | 南京邮电大学 | Medical image registration method based on combination of mutual information and gradient information |
| CN104318548A (en) * | 2014-10-10 | 2015-01-28 | 西安电子科技大学 | Rapid image registration implementation method based on space sparsity and SIFT feature extraction |
| WO2016110005A1 (en) * | 2015-01-07 | 2016-07-14 | 深圳市唯特视科技有限公司 | Gray level and depth information based multi-layer fusion multi-modal face recognition device and method |
Non-Patent Citations (1)
| Title |
|---|
| 一种非线性自适应图像光照影响去除方法;张钰;姚素英;徐江涛;;光电子.激光(第10期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111460864A (en) | 2020-07-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP6639123B2 (en) | Image processing apparatus, image processing method, and program | |
| CN105740780B (en) | Method and device for detecting living human face | |
| CN111611643B (en) | Household vectorization data acquisition method and device, electronic equipment and storage medium | |
| Zhao et al. | Individual identification of Holstein dairy cows based on detecting and matching feature points in body images | |
| WO2016150240A1 (en) | Identity authentication method and apparatus | |
| CN110909618B (en) | Method and device for identifying identity of pet | |
| CN111126240B (en) | Three-channel feature fusion face recognition method | |
| CN105335725A (en) | Gait identification identity authentication method based on feature fusion | |
| CN112883850B (en) | Multi-view space remote sensing image matching method based on convolutional neural network | |
| CN111754453A (en) | Pulmonary tuberculosis detection method and system based on chest radiography image and storage medium | |
| WO2021000832A1 (en) | Face matching method and apparatus, computer device, and storage medium | |
| WO2019033570A1 (en) | Lip movement analysis method, apparatus and storage medium | |
| WO2019184851A1 (en) | Image processing method and apparatus, and training method for neural network model | |
| JP2021503139A (en) | Image processing equipment, image processing method and image processing program | |
| CN109145704B (en) | Face portrait recognition method based on face attributes | |
| CN114429649B (en) | Target image identification method and device | |
| CN111460864B (en) | Animal disease detection method based on image recognition | |
| Bogo et al. | Automated detection of new or evolving melanocytic lesions using a 3D body model | |
| CN112257730A (en) | Plant pest image identification method, device, equipment and storage medium | |
| CN110008902B (en) | Finger vein recognition method and system fusing basic features and deformation features | |
| CN111666813A (en) | Subcutaneous sweat gland extraction method based on three-dimensional convolutional neural network of non-local information | |
| CN115909401A (en) | Cattle face identification method and device integrating deep learning, electronic equipment and medium | |
| JP6486084B2 (en) | Image processing method, image processing apparatus, and program | |
| CN112001896B (en) | Thyroid gland border irregularity detection device | |
| CN111783580B (en) | Pedestrian identification method based on human leg detection |
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 |