CN110781793A - Artificial intelligence real-time image recognition method based on quadtree algorithm - Google Patents
Artificial intelligence real-time image recognition method based on quadtree algorithm Download PDFInfo
- Publication number
- CN110781793A CN110781793A CN201910999616.7A CN201910999616A CN110781793A CN 110781793 A CN110781793 A CN 110781793A CN 201910999616 A CN201910999616 A CN 201910999616A CN 110781793 A CN110781793 A CN 110781793A
- Authority
- CN
- China
- Prior art keywords
- image
- quadtree
- current
- pictures
- 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.)
- Pending
Links
Images
Classifications
-
- 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
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- 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
-
- 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
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an artificial intelligence real-time image recognition method based on a quadtree algorithm, and relates to the technical field of image recognition. The method comprises the steps of separating an identification image or a divided picture according to a quadtree division method to obtain a plurality of current separated pictures; traversing all the current separating pictures, and matching the current separating pictures with the optimal matching blocks; if a match is satisfied, then the current separator tile is labeled R1 and the corresponding match tile is a domain tile and labeled D1. According to the invention, the image is cut by a quadtree method, so that the image quality is well ensured and the number of blocks is reduced; before segmenting the image, setting the maximum and minimum depths and allowable errors of the quadtree, and segmenting the image by utilizing a quartering method; finding the best matching block, if finding, not continuing to divide, if not finding, continuing to subdivide, and the process is continuously carried out until the set maximum depth; saving a lot of calculated amount, improving efficiency and accelerating image segmentation speed and accuracy.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to an artificial intelligence real-time image recognition method based on a quadtree algorithm.
Background
Image recognition technology is an important field of artificial intelligence. It refers to a technique of performing object recognition on an image to recognize various different modes of objects and objects. In general, image recognition technology, the first part is information acquisition, that is, information acquired by a sensor is converted into information that can be recognized by a machine through some method. The second part is to perform complex preprocessing on the information, and the third part is to extract features.
The feature extraction and selection means that feature extraction and selection are required in pattern recognition. It is simply understood that the images under study are various and if they are to be distinguished by some method, they are identified by their own features, and the process of obtaining these features is feature extraction. The features obtained in the feature extraction may not be all useful for this recognition, and in this case, useful features are extracted, which is the feature selection. Feature extraction and selection is one of the very critical techniques in image recognition, so understanding this step is the focus of image recognition.
The feature extraction and coding which are the most important steps of image recognition cannot be dynamically drawn on a common workstation and a computer in real time due to huge calculation amount, and the image recognition method based on the quad-tree algorithm, namely, the quad-tree algorithm is used for image segmentation, has higher speed than the traditional region segmentation technology, has less calculation energy consumption, keeps more excellent image details, and can obtain more information such as target size, edge, number and the like when the segmentation is finished. Therefore, the target characteristics of the image can be rapidly and accurately acquired, and real-time intelligent image recognition is realized.
Disclosure of Invention
The invention aims to provide an artificial intelligence real-time image recognition method based on a quadtree algorithm, wherein an image is cut by a quadtree method, so that the image quality is well ensured and the number of blocks is reduced; before segmenting the image, setting the maximum and minimum depths and allowable errors of the quadtree, and segmenting the image by utilizing a quartering method; finding the best matching block, if finding, not continuing to divide, if not finding, continuing to subdivide, and the process is continuously carried out until the set maximum depth; the problems of large calculated amount and low processing efficiency in the existing image recognition process are solved.
In order to solve the technical problems, the invention is realized by the following technical scheme:
the invention relates to an artificial intelligence real-time image recognition method based on a quadtree algorithm, which comprises the following steps:
s00: shooting three identification images by adopting a double-light simulation imaging sensor;
s01: respectively carrying out noise reduction, smoothing and transformation preprocessing on the three identification images;
s02: setting a corresponding minimum depth level range Min and an allowed maximum error value g according to the pixels of the identification image;
s03: separating the identification image or the divided picture according to a quadtree division method to obtain a plurality of current separated pictures;
s04: traversing all the current separating pictures, and matching the current separating pictures with the optimal matching blocks; if a match is satisfied, the current separator tile is labeled R1 and the corresponding matching tile is a domain tile and labeled D1; if not, go to S05;
s05: judging whether the depth level of the current partition picture is less than the minimum depth level rangeMin; if yes, go to S03; if not, keeping the current separation picture;
s06: training a neural network model in a deep learning mode, and carrying out classification and identification on the pictures; and matching the image characteristics obtained according to the quadtree method, and finally judging the image target.
Preferably, the three identification images include a normal image, a black-and-white mode image, and a pseudo color mode image; the double-light analog imaging sensor can not only acquire common optical image information, but also acquire image information of a black-and-white mode and a pseudo color mode in thermal imaging and convert the image information into a digital image which can be recognized by a computer; the black-and-white mode image is suitable for identifying the night image.
Preferably, the neural network model is specifically a convolutional neural network model; the convolution neural network model adopts PCA to estimate the surface normal to realize the colorization coding of the Depth mode image, and uses the Tensorflow model as the training parameter to finally realize the rapid identification of the ground weapon system and the operation unit.
The invention has the following beneficial effects:
1. according to the invention, the image is cut by a quadtree method, so that the image quality is well ensured and the number of blocks is reduced; before segmenting the image, setting the maximum and minimum depths and allowable errors of the quadtree, and segmenting the image by utilizing a quartering method; finding the best matching block, if finding, not continuing to divide, if not finding, continuing to subdivide, and the process is continuously carried out until the set maximum depth; a lot of calculated amount is saved, the efficiency is improved, and the image segmentation speed and accuracy are improved;
2. the invention adopts the double-light simulation imaging sensor to shoot three identification images, is practical for multi-scene image acquisition and has strong applicability.
Of course, it is not necessary for any product in which the invention is practiced to achieve all of the above-described advantages at the same time.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart of an artificial intelligence real-time image recognition method based on a quad-tree algorithm of the present invention;
FIG. 2 is a diagram illustrating a quadtree partition according to the present invention;
FIG. 3 is a representation of a quad-tree 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 only a part of the embodiments of the present invention, and not all of the embodiments. 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.
Referring to fig. 1, the present invention is an artificial intelligence real-time image recognition method based on a quadtree algorithm, including:
s00: shooting three identification images by adopting a double-light simulation imaging sensor; the three identification images comprise a common image, a black-and-white mode image and a pseudo-color mode image; the double-light analog imaging sensor can not only acquire common optical image information, but also acquire image information of a black-and-white mode and a pseudo color mode in thermal imaging and convert the image information into a digital image which can be recognized by a computer; the black-and-white mode image is suitable for identifying the night image.
S01: respectively carrying out noise reduction, smoothing and transformation preprocessing on the three identification images; the main features mainly used for enhancing the image;
s02: setting a corresponding minimum depth level range Min and an allowed maximum error value g according to the pixels of the identification image;
s03: separating the identification image or the divided picture according to a quadtree division method to obtain a plurality of current separated pictures;
s04: traversing all the current separating pictures, and matching the current separating pictures with the optimal matching blocks; if a match is satisfied, the current separator tile is labeled R1 and the corresponding matching tile is a domain tile and labeled D1; if not, go to S05;
s05: judging whether the depth level of the current partition picture is less than the minimum depth level rangeMin; if yes, go to S03; if not, keeping the current separation picture;
when in specific use: as shown in fig. 2, for example, the image is encoded into a 256 × 256 pixel image, we may set the minimum depth level to 4, partition the image into blocks by using a quadtree segmentation method until the block size is 16 × 16, then find a best matching block for each block, if the best matching block is found, mark the best matching block as R1, mark the corresponding matching block as domain block as D1, and do not segment the block any more, if no matching block is found, continue to perform quadtree segmentation on the changed block, and repeat the above steps, the image information recognized by the computer becomes clearer the more the number of layers is, until the maximum depth level is reached, and obtain a quadtree segmentation map;
s06: training a neural network model in a deep learning mode, and carrying out classification and identification on the pictures; matching according to the image characteristics obtained by the quadtree method, and finally judging the image target;
specifically, the method comprises the following steps: the core technology of deep learning image recognition is a multi-modal Deep Convolutional Neural Network (DCNN) improvement algorithm based on a supervised learning strategy. The algorithm adopts PCA to estimate the surface normal to realize the colorization coding of the Depth modal image, uses a Tensorflow model as a training parameter, and finally realizes the rapid identification of a ground weapon system and a combat unit. The network model scheme specifically adopting deep learning is as follows:
convolutional neural network
A Convolutional Neural Network (CNN) is similar to a conventional neural network structure, and is a multi-layer network structure, and each layer is composed of the largest neurons. Can be regarded as simple simulation of the human cranial nerve structure; the first part is an input layer, the second part is a combination of n convolutional layers and other layers, and the third part is a fully-connected multilayer perceptron classifier; the convolutional kernel obtains a reasonable weight in the training process of the network, and the shared weight can reduce the connection among all layers of the network and reduce the risk of overfitting. There are two forms of mean pooling and maximum pooling, both of which can be considered as special convolution processes. The arching and pooling greatly simplify the complexity of the model and reduce the parameters of the model; the classifier mainly comprises Softmax and SVN, and most convolutional neural networks adopt Softmax as the classifier.
Referring to FIG. 3, the quadtree partitioning algorithm of the present invention is illustrated as follows:
assuming that the upper left corner of the image is (0, 0) and the lower right corner is (m, n), the uppermost node in the image, i.e. the arbitrary father node p (i, j) in fig. 3, is the second layer of the father node, i.e. the third layer of the child node; generally, a father node is directly selected on a target image, so that the segmentation time is saved, all pixels which are difficult to identify or are not segmented in the target image can be used as father nodes to segment the whole image, the background is also used as a special target in the segmentation result, and the background is also easily distinguished by judging the size and the gray value of the target.
In the process of the quadtree image segmentation, the area blocks which do not meet the requirement of further tapping points can be directly removed, some sub-tapping points can be directly removed from the parent tapping points, and the tapping points can be directly removed in several situations under general conditions:
1. the child node does not satisfy the segmentation condition;
2. the child nodes comprise parent contacts;
3. the child node comprises the segmented sibling nodes;
removing judgment conditions:
(1) grayscale similarity decision
Let the average gray value of the divided region be
If the measured gray value is y, the similarity S between the measured region and the divided region can be expressed as:
and w is a weight value, and for S which is small enough, the pixel to be tested can be tasked and is similar to the segmented pixel, and the pixel is merged into the segmented target, otherwise, the pixel is eliminated.
(2) Mean square error decision
Let the number of pixels in the divided region R ben, average gray value of
The gray value of the pixel to be measured is y, and the mean value and the variance of R are defined as follows:
(3)
defining:
for the similarity criterion, if T is small enough, the pixel to be measured is considered to be similar to the segmented region and to be merged into the segmented object, otherwise the pixel is clipped. The similarity criterion can be edge similarity, gradient similarity, feature similarity and the like, and the similarity criterion can also be mutually combined and determined according to the target property.
It should be noted that, in the above system embodiment, each included unit is only divided according to functional logic, but is not limited to the above division as long as the corresponding function can be implemented; in addition, specific names of the functional units are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present invention.
In addition, it is understood by those skilled in the art that all or part of the steps in the method for implementing the embodiments described above may be implemented by a program instructing associated hardware, and the corresponding program may be stored in a computer-readable storage medium.
The preferred embodiments of the invention disclosed above are intended to be illustrative only. The preferred embodiments are not intended to be exhaustive or to limit the invention to the precise embodiments disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best utilize the invention. The invention is limited only by the claims and their full scope and equivalents.
Claims (3)
1. An artificial intelligence real-time image recognition method based on a quadtree algorithm is characterized by comprising the following steps:
s00: shooting three identification images by adopting a double-light simulation imaging sensor;
s01: respectively carrying out noise reduction, smoothing and transformation preprocessing on the three identification images;
s02: setting a corresponding minimum depth level range Min and an allowed maximum error value g according to the pixels of the identification image;
s03: separating the identification image or the divided picture according to a quadtree division method to obtain a plurality of current separated pictures;
s04: traversing all the current separating pictures, and matching the current separating pictures with the optimal matching blocks; if a match is satisfied, the current separator tile is labeled R1 and the corresponding matching tile is a domain tile and labeled D1; if not, go to S05;
s05: judging whether the depth level of the current partition picture is less than the minimum depth level rangeMin; if yes, go to S03; if not, keeping the current separation picture;
s06: training a neural network model in a deep learning mode, and carrying out classification and identification on the pictures; and matching the image characteristics obtained according to the quadtree method, and finally judging the image target.
2. The artificial intelligence real-time image recognition method based on the quadtree algorithm of claim 1, wherein the three recognition images comprise a normal image, a black and white mode image and a pseudo color mode image; the double-light analog imaging sensor can not only acquire common optical image information, but also acquire image information of a black-and-white mode and a pseudo color mode in thermal imaging and convert the image information into a digital image which can be recognized by a computer; the black-and-white mode image is suitable for identifying the night image.
3. The artificial intelligence real-time image recognition method based on the quadtree algorithm as claimed in claim 1, wherein the neural network model is specifically a convolutional neural network model; the convolution neural network model adopts PCA to estimate the surface normal to realize the colorization coding of the Depth mode image, and uses the Tensorflow model as the training parameter to finally realize the rapid identification of the ground weapon system and the operation unit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910999616.7A CN110781793A (en) | 2019-10-21 | 2019-10-21 | Artificial intelligence real-time image recognition method based on quadtree algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910999616.7A CN110781793A (en) | 2019-10-21 | 2019-10-21 | Artificial intelligence real-time image recognition method based on quadtree algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110781793A true CN110781793A (en) | 2020-02-11 |
Family
ID=69386025
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910999616.7A Pending CN110781793A (en) | 2019-10-21 | 2019-10-21 | Artificial intelligence real-time image recognition method based on quadtree algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110781793A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102595141A (en) * | 2012-03-13 | 2012-07-18 | 中国科学院上海应用物理研究所 | Fractal image compression method based on combination of quad tree and neighborhood searching |
| CN102833529A (en) * | 2012-09-19 | 2012-12-19 | 山东神戎电子股份有限公司 | Multispectral monitoring system with distance measurement and image stabilization functions |
| CN103714659A (en) * | 2013-12-26 | 2014-04-09 | 苏州清研微视电子科技有限公司 | Fatigue driving identification system based on double-spectrum fusion |
| CN107423760A (en) * | 2017-07-21 | 2017-12-01 | 西安电子科技大学 | Based on pre-segmentation and the deep learning object detection method returned |
| CN107451552A (en) * | 2017-07-25 | 2017-12-08 | 北京联合大学 | A kind of gesture identification method based on 3D CNN and convolution LSTM |
| CN109905712A (en) * | 2019-01-09 | 2019-06-18 | 西安邮电大学 | HEVC intraframe coding fast mode decision algorithm based on ResNet |
| CN110189308A (en) * | 2019-05-17 | 2019-08-30 | 山东财经大学 | A kind of lesion detection approach and device based on BM3D and the fusion of dense convolutional network |
-
2019
- 2019-10-21 CN CN201910999616.7A patent/CN110781793A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102595141A (en) * | 2012-03-13 | 2012-07-18 | 中国科学院上海应用物理研究所 | Fractal image compression method based on combination of quad tree and neighborhood searching |
| CN102833529A (en) * | 2012-09-19 | 2012-12-19 | 山东神戎电子股份有限公司 | Multispectral monitoring system with distance measurement and image stabilization functions |
| CN103714659A (en) * | 2013-12-26 | 2014-04-09 | 苏州清研微视电子科技有限公司 | Fatigue driving identification system based on double-spectrum fusion |
| CN107423760A (en) * | 2017-07-21 | 2017-12-01 | 西安电子科技大学 | Based on pre-segmentation and the deep learning object detection method returned |
| CN107451552A (en) * | 2017-07-25 | 2017-12-08 | 北京联合大学 | A kind of gesture identification method based on 3D CNN and convolution LSTM |
| CN109905712A (en) * | 2019-01-09 | 2019-06-18 | 西安邮电大学 | HEVC intraframe coding fast mode decision algorithm based on ResNet |
| CN110189308A (en) * | 2019-05-17 | 2019-08-30 | 山东财经大学 | A kind of lesion detection approach and device based on BM3D and the fusion of dense convolutional network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110956094B (en) | RGB-D multi-mode fusion personnel detection method based on asymmetric double-flow network | |
| CN104809443B (en) | Detection method of license plate and system based on convolutional neural networks | |
| CN108121991B (en) | Deep learning ship target detection method based on edge candidate region extraction | |
| CN109740460B (en) | Optical remote sensing image ship detection method based on depth residual error dense network | |
| CN110555465B (en) | Weather image identification method based on CNN and multi-feature fusion | |
| CN108537239B (en) | Method for detecting image saliency target | |
| CN108268859A (en) | A kind of facial expression recognizing method based on deep learning | |
| CN109886128B (en) | Face detection method under low resolution | |
| CN109635636B (en) | Pedestrian re-identification method based on fusion of attribute characteristics and weighted blocking characteristics | |
| JP4098021B2 (en) | Scene identification method, apparatus, and program | |
| CN112614136B (en) | Infrared small target real-time instance segmentation method and device | |
| CN107066916A (en) | Scene Semantics dividing method based on deconvolution neutral net | |
| CN112418087B (en) | Underwater video fish identification method based on neural network | |
| CN106407951B (en) | A kind of night front vehicles detection method based on monocular vision | |
| CN108320294B (en) | Intelligent full-automatic portrait background replacement method for second-generation identity card photos | |
| CN113111878B (en) | Infrared weak and small target detection method under complex background | |
| JP2005316888A (en) | Face recognition system | |
| CN111539320B (en) | Multi-view gait recognition method and system based on mutual learning network strategy | |
| CN115862066A (en) | Improved YOLOv5 lightweight community scene downlight detection method | |
| CN107122732B (en) | High-robustness rapid license plate positioning method in monitoring scene | |
| CN115019340A (en) | Night pedestrian detection algorithm based on deep learning | |
| CN113592893B (en) | Image foreground segmentation method for determining combination of main body and accurate edge | |
| CN113139431B (en) | Image saliency target detection method based on deep supervised learning | |
| CN110334628A (en) | A kind of outdoor monocular image depth estimation method based on structuring random forest | |
| CN110503049B (en) | Satellite video vehicle number estimation method based on generation countermeasure 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 |