CN112766092A - Method for quickly identifying background category based on brain-like neural network and application thereof - Google Patents
Method for quickly identifying background category based on brain-like neural network and application thereof Download PDFInfo
- Publication number
- CN112766092A CN112766092A CN202110007074.8A CN202110007074A CN112766092A CN 112766092 A CN112766092 A CN 112766092A CN 202110007074 A CN202110007074 A CN 202110007074A CN 112766092 A CN112766092 A CN 112766092A
- Authority
- CN
- China
- Prior art keywords
- target
- neural network
- background
- brain
- classification
- 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
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 22
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000012544 monitoring process Methods 0.000 claims abstract description 10
- 238000013527 convolutional neural network Methods 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 10
- 238000012216 screening Methods 0.000 claims description 6
- 230000000295 complement effect Effects 0.000 claims description 2
- 238000000605 extraction Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 abstract description 4
- 230000005540 biological transmission Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 11
- 238000001514 detection method Methods 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 6
- 238000002372 labelling Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the field of image processing, and discloses a method for quickly identifying background categories based on a brain-like neural network and application thereof. The method enhances the accuracy of the model in distinguishing the target from the background by introducing the brain-like neural network, omits the process of gradient back transmission and accelerates the processing speed of the model. The method is suitable for quickly identifying the target-background of the image, and is particularly suitable for identifying the target-background of the monitoring area of the unmanned aerial vehicle.
Description
Technical Field
The invention belongs to the field of image processing, and relates to a background category rapid identification method and application thereof, in particular to a background category rapid identification method based on a brain-like neural network and application thereof.
Background
Object detection is a direction of wide application and mature technology in the field of pattern recognition, and the object detection technology is a method for identifying the positions of objects in certain categories and the corresponding categories in the image range. The target detection technology extracts the characteristics and background information of a target from a data set consistent with a real scene through learning, and a model is constructed to identify the real scene. The target detection technology is widely applied to the fields of intelligent monitoring, man-machine interaction, intelligent transportation, unmanned aerial vehicles and the like.
In recent years, object detection technologies have been developed very rapidly, and many excellent object detection algorithms emerge, but these object detection technologies can only detect object classes of one closed set model, and cannot identify classes that are not seen in the closed set model, so that a background area cannot be accurately identified.
Disclosure of Invention
In order to solve the defects in the prior art, the invention aims to provide a method for quickly identifying the background category based on a brain-like neural network so as to accurately and quickly identify the background area aiming at an image;
the invention also provides application of the background category rapid identification method based on the brain-like neural network.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a background category rapid identification method based on a brain-like neural network comprises a training stage and an identification stage;
the training phase mainly comprises the following steps:
a1, inputting the training picture into a convolutional neural network for feature extraction, and establishing a target feature database;
a2, inputting live-action pictures to the convolutional neural network periodically, comparing the live-action pictures with a target characteristic database, and updating characteristic elements of the target characteristic database in the convolutional neural network by using a loss formula;
the identification phase mainly comprises the following steps:
b1, inputting the real-time picture shot by the unmanned aerial vehicle into the trained convolutional neural network for feature recognition;
b2, comparing the identified features with the established target feature database, and screening out a target part and a background part;
b3, performing secondary classification on the target part by using a classification function to screen out a primary target and a secondary target;
b4, labeling output primary target, secondary target and background;
as a limitation of the invention, the pictures for training comprise a main target element labeling picture and a generalized target labeling picture;
as another limitation of the present invention, the loss formula is a cross-entropy loss formula
Wherein, H is a calculated value of cross entropy, namely a calculated value of loss in the invention; p is the predicted probability value of the classification function; q is the complement of the prediction probability value of the classification function; i is a subscript for each category;
as a third limitation of the present invention, the classification function is a sigmoid function of
Wherein S represents a probability value predicted by a sigmoid classification function, and x is each element of an input one-dimensional classification characteristic vector;
the invention also provides an application of the background category rapid identification method based on the brain-like neural network, wherein the method is applied to target-background identification of the unmanned aerial vehicle on the monitoring area;
due to the adoption of the technical scheme, compared with the prior art, the invention has the following beneficial effects:
(1) according to the method for quickly identifying the background category based on the brain-like neural network, the brain-like neural network is introduced to learn the generalized target category, so that the model has the two-classification detection capability of positioning and judging whether the target object is the target object, the background area is divided, the interference of other category objects on the identification of the background area is avoided, and the prediction of the background category is more accurate;
(2) the background category rapid identification method based on the brain-like neural network provided by the invention replaces a back propagation neuron used in the traditional identification method by the pulse neuron, omits the step of gradient back propagation and improves the speed of target detection;
in summary, the method for rapidly identifying the background category based on the brain-like neural network provided by the invention enhances the accuracy of the model in distinguishing the target from the background by introducing the brain-like neural network, omits the process of gradient back transmission, and accelerates the processing speed of the model.
The method is suitable for quickly identifying the target-background of the image, and is particularly suitable for identifying the target-background of the monitoring area of the unmanned aerial vehicle.
Drawings
The invention is described in further detail below with reference to the figures and the embodiments.
FIG. 1 is a logic block diagram of a method for rapidly identifying a background category based on a neural network of the brain;
FIG. 2 is a labeled diagram of the main target elements for training in accordance with the present invention;
FIG. 3 is a full-scale plot of the monitored area of the present invention;
FIG. 4 is a live-action view taken by the unmanned aerial vehicle according to the embodiment of the present invention;
FIG. 5 is a target-background labeling diagram output by an embodiment of the present invention.
In fig. 5: 1. a primary target; 2. a secondary target; 3. a background region.
Detailed Description
Preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the description of the preferred embodiment is only for purposes of illustration and understanding, and is not intended to limit the invention.
Embodiment of the invention relates to a method for quickly identifying background categories based on a brain-like neural network
The present embodiment implements the provided method for quickly identifying the background category based on the brain-like neural network by using the unmanned aerial vehicle, the monitor mounted below the unmanned aerial vehicle, and the image processing module, and the logic block diagram of the method is shown in fig. 1.
Before the image processing module is carried on the unmanned aerial vehicle, firstly, the image processing module is trained according to the following steps:
a1, respectively inputting all visual information labeling pictures (figure 2) of the overall shape, outline and partial regional background for training and all essential element target labeling pictures (figure 3) of the monitoring region into a convolutional neural network in an image processing module for learning and extracting target category characteristics (the category characteristics refer to one-dimensional characteristics obtained by calculating a characteristic graph through classified full connection layers after the characteristic graph is obtained by model convolution of the input pictures through a model learned by the convolutional neural network), and establishing a target characteristic database;
a2, inputting the live-action pictures of the monitoring area in one week into a convolutional neural network in an image processing module, comparing the live-action pictures with the target class characteristics extracted in A1 through a cross entropy loss function, and updating characteristic elements in a database;
the cross entropy loss function calculation process is as follows:
wherein L is a loss value, y is a class predictor,
is a category true value, and log is a logarithmic function with 2 as a base;
before the unmanned aerial vehicle carries on the image processing module for the first time, the A1 is used for training the image processing module, and before the unmanned aerial vehicle is used for monitoring, only the A2 needs to be carried out, and the target element characteristics of the monitoring area can be updated.
Control unmanned aerial vehicle and under no highlight and haze interference, steadily fly in the air of 1 kilometer below apart from ground. In the flight process of the unmanned aerial vehicle, a monitor mounted below the unmanned aerial vehicle is controlled to shoot an area, an obtained image is uploaded to an image processing module, and the following steps are carried out to carry out target-background discrimination:
b1, inputting the uploaded real-time picture (figure 4) into a trained convolutional neural network, and extracting characteristic elements in the real-time picture;
b2, comparing the extracted characteristic elements with an established target element database in the convolutional neural network, and screening out a target part and a background part;
b3, performing secondary classification on the screened target part through a sigmoid function, and screening the target part into a main target and a secondary target;
the sigmoid function is calculated as follows:
wherein x is an input vector, y is an output vector, and e is a natural logarithm;
b4, selecting the background part, the primary target part and the secondary target part by using frame frames with different colors respectively, uploading the selected parts to a display screen of the monitoring terminal for display, and outputting the result as shown in fig. 5;
wherein, B3 is needed to be carried out for a plurality of times until the screening result is the same as the last screening result, and then the label output of B4 is carried out.
Although the present invention has been described in detail with reference to the above embodiments, it will be apparent to those skilled in the art that modifications may be made to the embodiments described above, or equivalents may be substituted for elements thereof. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (4)
1. A kind of background classification fast recognition method based on neural network of brain-like, said method is carried in the image processing module, through the neural network of convolution among them, through training stage and recognition stage to realize, characterized by that:
the training phase mainly comprises the following steps:
a1, inputting the training picture into a convolutional neural network for feature extraction, and establishing a target feature database;
a2, inputting live-action pictures to the convolutional neural network periodically, comparing the live-action pictures with a target characteristic database, and updating characteristic elements of the target characteristic database in the convolutional neural network by using a loss formula;
the identification phase mainly comprises the following steps:
b1, inputting the real-time picture shot by the unmanned aerial vehicle into the trained convolutional neural network for feature recognition;
b2, comparing the identified features with the established target feature database, and screening out a target part and a background part;
b3, performing secondary classification on the target part by using a classification function to screen out a primary target and a secondary target;
b4, label output primary target, secondary target, and background.
2. The method for rapidly identifying the background category based on the brain-like neural network as claimed in claim 1, wherein: the loss formula is a cross entropy loss formula
… … formula I
Wherein, H is a calculated value of cross entropy, namely a calculated value of loss in the invention; p is the predicted probability value of the classification function; q is the complement of the prediction probability value of the classification function; i is a subscript for each category.
3. The method for rapidly identifying the background category based on the brain-like neural network according to claim 1 or 2, wherein: the classification function is a sigmoid function as follows
Wherein S represents the probability value predicted by the sigmoid classification function, and x is each element of the input one-dimensional classification feature vector.
4. An application of the method for rapidly identifying the background category based on the cranial neural network according to any one of claims 1 to 3, wherein: the method is applied to target-background recognition of the unmanned aerial vehicle on the monitoring area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110007074.8A CN112766092A (en) | 2021-01-05 | 2021-01-05 | Method for quickly identifying background category based on brain-like neural network and application thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110007074.8A CN112766092A (en) | 2021-01-05 | 2021-01-05 | Method for quickly identifying background category based on brain-like neural network and application thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112766092A true CN112766092A (en) | 2021-05-07 |
Family
ID=75699194
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110007074.8A Pending CN112766092A (en) | 2021-01-05 | 2021-01-05 | Method for quickly identifying background category based on brain-like neural network and application thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112766092A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113569683A (en) * | 2021-07-20 | 2021-10-29 | 上海明略人工智能(集团)有限公司 | Scene classification method, system, device and medium combining salient region detection |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100272366A1 (en) * | 2009-04-24 | 2010-10-28 | Sony Corporation | Method and device of detecting object in image and system including the device |
| US20160342837A1 (en) * | 2015-05-19 | 2016-11-24 | Toyota Motor Engineering & Manufacturing North America, Inc. | Apparatus and method for object tracking |
| JP6407467B1 (en) * | 2018-05-21 | 2018-10-17 | 株式会社Ｇａｕｓｓ | Image processing apparatus, image processing method, and program |
| CN108875803A (en) * | 2018-05-30 | 2018-11-23 | 长安大学 | A kind of detection of harmful influence haulage vehicle and recognition methods based on video image |
| CN109993082A (en) * | 2019-03-20 | 2019-07-09 | 上海理工大学 | The classification of convolutional neural networks road scene and lane segmentation method |
| CN110826612A (en) * | 2019-10-31 | 2020-02-21 | 上海法路源医疗器械有限公司 | Training and identifying method for deep learning |
| CN110929774A (en) * | 2019-11-18 | 2020-03-27 | 腾讯科技（深圳）有限公司 | Method for classifying target objects in image, method and device for training model |
-
2021
- 2021-01-05 CN CN202110007074.8A patent/CN112766092A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100272366A1 (en) * | 2009-04-24 | 2010-10-28 | Sony Corporation | Method and device of detecting object in image and system including the device |
| US20160342837A1 (en) * | 2015-05-19 | 2016-11-24 | Toyota Motor Engineering & Manufacturing North America, Inc. | Apparatus and method for object tracking |
| JP6407467B1 (en) * | 2018-05-21 | 2018-10-17 | 株式会社Ｇａｕｓｓ | Image processing apparatus, image processing method, and program |
| CN108875803A (en) * | 2018-05-30 | 2018-11-23 | 长安大学 | A kind of detection of harmful influence haulage vehicle and recognition methods based on video image |
| CN109993082A (en) * | 2019-03-20 | 2019-07-09 | 上海理工大学 | The classification of convolutional neural networks road scene and lane segmentation method |
| CN110826612A (en) * | 2019-10-31 | 2020-02-21 | 上海法路源医疗器械有限公司 | Training and identifying method for deep learning |
| CN110929774A (en) * | 2019-11-18 | 2020-03-27 | 腾讯科技（深圳）有限公司 | Method for classifying target objects in image, method and device for training model |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113569683A (en) * | 2021-07-20 | 2021-10-29 | 上海明略人工智能(集团)有限公司 | Scene classification method, system, device and medium combining salient region detection |
| CN113569683B (en) * | 2021-07-20 | 2024-04-02 | 上海明略人工智能(集团)有限公司 | Scene classification method, system, equipment and medium combined with salient region detection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110070008B (en) | Bridge disease identification method adopting unmanned aerial vehicle image | |
| CN108388927B (en) | Small sample polarization SAR terrain classification method based on deep convolution twin network | |
| CN110070530B (en) | Transmission line icing detection method based on deep neural network | |
| CN111640101B (en) | Ghost convolution characteristic fusion neural network-based real-time traffic flow detection system and method | |
| CN109684922B (en) | Multi-model finished dish identification method based on convolutional neural network | |
| CN108039044B (en) | Vehicle intelligent queuing system and method based on multi-scale convolutional neural network | |
| CN109598268A (en) | A kind of RGB-D well-marked target detection method based on single flow depth degree network | |
| CN106096602A (en) | Chinese license plate recognition method based on convolutional neural network | |
| CN104517103A (en) | Traffic sign classification method based on deep neural network | |
| CN104598924A (en) | Target matching detection method | |
| He et al. | A robust method for wheatear detection using UAV in natural scenes | |
| CN110569779A (en) | Pedestrian attribute identification method based on pedestrian local and overall attribute joint learning | |
| CN109255286A (en) | A kind of quick detection recognition method of unmanned plane optics based on YOLO deep learning network frame | |
| US11449707B2 (en) | Method for processing automobile image data, apparatus, and readable storage medium | |
| CN108345900B (en) | Pedestrian re-identification method and system based on color texture distribution characteristics | |
| CN114627447A (en) | Road vehicle tracking method and system based on attention mechanism and multi-target tracking | |
| CN113435407B (en) | Small target identification method and device for power transmission system | |
| CN103336971A (en) | Target matching method among multiple cameras based on multi-feature fusion and incremental learning | |
| CN111680705A (en) | MB-SSD method and MB-SSD feature extraction network suitable for target detection | |
| CN113255634A (en) | Vehicle-mounted mobile terminal target detection method based on improved Yolov5 | |
| CN116563410A (en) | Electrical equipment electric spark image generation method based on two-stage generation countermeasure network | |
| CN114359167A (en) | Insulator defect detection method based on lightweight YOLOv4 in complex scene | |
| CN110618129A (en) | Automatic power grid wire clamp detection and defect identification method and device | |
| CN112766092A (en) | Method for quickly identifying background category based on brain-like neural network and application thereof | |
| CN110334703B (en) | Ship detection and identification method in day and night image |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210507 |