CN108090442A - A kind of airport scene monitoring method based on convolutional neural networks - Google Patents

A kind of airport scene monitoring method based on convolutional neural networks Download PDF

Info

Publication number
CN108090442A
CN108090442A CN201711344003.7A CN201711344003A CN108090442A CN 108090442 A CN108090442 A CN 108090442A CN 201711344003 A CN201711344003 A CN 201711344003A CN 108090442 A CN108090442 A CN 108090442A
Authority
CN
China
Prior art keywords
scene
neural networks
convolutional neural
monitoring
airport
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
Application number
CN201711344003.7A
Other languages
Chinese (zh)
Inventor
韩松臣
黄国新
黄畅昕
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sichuan University
Original Assignee
Sichuan University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Sichuan University filed Critical Sichuan University
Priority to CN201711344003.7A priority Critical patent/CN108090442A/en
Publication of CN108090442A publication Critical patent/CN108090442A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
    • G06K9/00Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
    • G06K9/00624Recognising scenes, i.e. recognition of a whole field of perception; recognising scene-specific objects
    • G06K9/00711Recognising video content, e.g. extracting audiovisual features from movies, extracting representative key-frames, discriminating news vs. sport content
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computer systems based on biological models
    • G06N3/02Computer systems based on biological models using neural network models
    • G06N3/08Learning methods
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N7/00Television systems
    • H04N7/18Closed circuit television systems, i.e. systems in which the signal is not broadcast

Abstract

Scene surveillance radar (SMR) expense of great number can not be undertaken for medium and small airport, and the scene monitoring mode for relying on controller visual can not meet modern Aviation security needs.Video image monitoring is a kind of versatile, lower-cost monitoring means, thus proposes a kind of video detection system based on Single Shot MultiBox Detector convolutional neural networks detection methods.Real airdrome scene video is acquired, and aircraft therein, automobile, people are demarcated, one is created and includes the airdrome scene data set of a large amount of pictures, and SSD models are trained on this data set.The effect of real-time detection has been basically reached, a new method is provided for middle-size and small-size airport scene monitoring system.

Description

A kind of airport scene monitoring method based on convolutional neural networks
Technical field
The invention belongs to airport scene monitoring field, the airport scene monitoring method of video flowing is based especially on.
Background technology
In recent years, China Air Transport scale constantly expands, and has developed into world's second largest air transportation state, general boat Empty development also achieves no small achievement simultaneously, and the quantity on medium and small airport increases therewith.Field in existing airport monitoring means Face surveillance radar, multipoint positioning cost are higher, and medium and small airport is difficult to realize, and can there are blind areas.Again in this case, based on volume The scene monitoring of product neutral net comes into being.
At present, common airport scene monitoring means include scene surveillance radar, multipoint positioning(Multilateration, Abbreviation MLAT)And Automatic dependent surveillance broadcast(Automatic Dependent Surveillance-Broadcast, referred to as ADS-B)Etc. technological means.But since scene surveillance radar is sufficiently expensive, it is more than dynamic then up to ten million members, not all airport Such big expense, middle-size and small-size airport especially less to flight sortie of taking off and landing can be afforded.Regardless of be MLAT or ADS-B, due to it, to need the R-T unit by cordless communication network and in monitored target that could realize more high-precision The positioning and monitoring of degree.
For being fitted without the noncooperative target of R-T unit, such as most vehicles in scene operation and flight crew, MLAT and ADS-B can not realize effective position and monitoring.
The content of the invention
In view of this, the present invention provides a kind of airport scene monitoring method based on convolutional neural networks, this method is adopted With deep learning network SSD (Single Shot MultiBox Detector), by real time to target in airdrome scene image It is identified, realizes scene monitoring.
Technical scheme is implemented as follows:
A kind of method of trained SSD convolutional neural networks, detailed process are:
(1) by scene monitoring camera, substantial amounts of airdrome scene picture is shot, wherein the target for needing to identify should be included;
(2) in each pictures of above-mentioned acquisition the target identified will be needed to demarcate, obtains including the four of target rectangle frame The xml document of the coordinate of a point;
(3) deep learning frame caffe is built, compiles SSD source codes, calibration picture is learnt.Generation identification spotting Specific convolutional neural networks.
A kind of scene monitoring method based on SSD concretely comprises the following steps:
A. airdrome scene monitoring video flow is obtained, and picture frame is converted into using OpenCV;
B. the picture frame that the needs of acquisition detect is inputted into previously trained SSD convolutional neural networks, network is to figure Target in piece is identified;
C. a threshold value is setT(0<T<1) confidence level generated with SSD Network Recognitions targetZ(0<Z<1) it is compared, whenTZWhen, enter step D;Otherwise E is entered step;
D. in picture is inputted output is used as plus the rectangle frame of identified target;
E. confidence level does not reach requirement, it is believed that is the target of identification mistake, not plus rectangle frame.
The present invention is based on SSD (Single Shot MultiBox Detector) convolutional neural networks to carry out target identification, The target identification image with high-accuracy is obtained, solves the problems, such as airport scene monitoring somewhat expensive.
Description of the drawings
Fig. 1 is characteristic pattern object matching process in present example.
Fig. 2 is SSD network structures in present example.
Fig. 3 is SSD training figure in present example.
Fig. 4 is training picture calibration in present example.
Fig. 5 is detection result figure in present example.

Claims (2)

1. a kind of airport scene monitoring method based on convolutional neural networks, wherein, the convolutional neural networks include but unlimited In SSD (Single Shot MultiBox Detector), which is characterized in that comprise the following steps:
Step 1: the convolutional neural networks of structure detection;
Step 2: it shoots and demarcates the substantial amounts of airdrome scene image to be monitored;
Step 3: the convolutional neural networks are carried out to learn the demarcating file obtained in the step 2, target identification is obtained Model.
2. it according to the method described in claim 1, it is characterized in that, is further included after the step 3:
Step 4: the airdrome scene monitoring video flow is inputted into the Model of Target Recognition.
CN201711344003.7A 2017-12-15 2017-12-15 A kind of airport scene monitoring method based on convolutional neural networks Pending CN108090442A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711344003.7A CN108090442A (en) 2017-12-15 2017-12-15 A kind of airport scene monitoring method based on convolutional neural networks

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711344003.7A CN108090442A (en) 2017-12-15 2017-12-15 A kind of airport scene monitoring method based on convolutional neural networks

Publications (1)

Publication Number Publication Date
CN108090442A true CN108090442A (en) 2018-05-29

Family

ID=62176183

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711344003.7A Pending CN108090442A (en) 2017-12-15 2017-12-15 A kind of airport scene monitoring method based on convolutional neural networks

Country Status (1)

Country Link
CN (1) CN108090442A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109977840A (en) * 2019-03-20 2019-07-05 四川川大智胜软件股份有限公司 A kind of airport scene monitoring method based on deep learning

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102509091A (en) * 2011-11-29 2012-06-20 北京航空航天大学 Airplane tail number recognition method
CN103455982A (en) * 2013-08-31 2013-12-18 四川川大智胜软件股份有限公司 Airport scene monitoring vision enhancing method based on virtual-real fusion
CN104243935A (en) * 2014-10-10 2014-12-24 南京莱斯信息技术股份有限公司 Target monitoring method for airport field monitoring system on basis of video recognition
CN106709568A (en) * 2016-12-16 2017-05-24 北京工业大学 RGB-D image object detection and semantic segmentation method based on deep convolution network
CN106846813A (en) * 2017-03-17 2017-06-13 西安电子科技大学 The method for building urban road vehicle image data base
CN107066953A (en) * 2017-03-22 2017-08-18 北京邮电大学 It is a kind of towards the vehicle cab recognition of monitor video, tracking and antidote and device
CN107122439A (en) * 2017-04-21 2017-09-01 图麟信息科技(深圳)有限公司 A kind of video segment querying method and device

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102509091A (en) * 2011-11-29 2012-06-20 北京航空航天大学 Airplane tail number recognition method
CN103455982A (en) * 2013-08-31 2013-12-18 四川川大智胜软件股份有限公司 Airport scene monitoring vision enhancing method based on virtual-real fusion
CN104243935A (en) * 2014-10-10 2014-12-24 南京莱斯信息技术股份有限公司 Target monitoring method for airport field monitoring system on basis of video recognition
CN106709568A (en) * 2016-12-16 2017-05-24 北京工业大学 RGB-D image object detection and semantic segmentation method based on deep convolution network
CN106846813A (en) * 2017-03-17 2017-06-13 西安电子科技大学 The method for building urban road vehicle image data base
CN107066953A (en) * 2017-03-22 2017-08-18 北京邮电大学 It is a kind of towards the vehicle cab recognition of monitor video, tracking and antidote and device
CN107122439A (en) * 2017-04-21 2017-09-01 图麟信息科技(深圳)有限公司 A kind of video segment querying method and device

Non-Patent Citations (7)

* Cited by examiner, † Cited by third party
Title
MATTHEW COOMBES 等: "Machine Vision for UAS Ground Operations", 《JOURNAL OF INTELLIGENT & ROBOTIC SYSTEMS》 *
TINGHE ZHU 等: "Integrating saliency and ResNet for airport detection in large-size remote sensing images", 《INTERNATIONAL CONFERENCE ON IMAGE, VISION AND COMPUTING》 *
XIAOGUANG CAO 等: "Detecting of foreign object debris on airfield pavement using convolution neural network", 《PROCEEDINGS OF SPIE》 *
俞汝劼 等: "基于深度卷积神经网络的航空器检测与识别", 《计算机应用》 *
孙升林 等: "《煤炭地质勘查与资源评价管理》", 31 December 2007 *
牛新 等: "基于深度学习的光学遥感机场与飞行器目标识别技术", 《大数据》 *
顾月清: "《全国普通高等院校生物医学工程规划教材 生物医学工程技术》", 31 July 2017 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109977840A (en) * 2019-03-20 2019-07-05 四川川大智胜软件股份有限公司 A kind of airport scene monitoring method based on deep learning

Similar Documents

Publication Publication Date Title
Sun et al. A camera-based target detection and positioning UAV system for search and rescue (SAR) purposes
CN105389988B (en) A kind of express highway intelligent cruising inspection system of multiple no-manned plane collaboration
US10565458B2 (en) Simulation system, simulation program and simulation method
CN104809443B (en) Detection method of license plate and system based on convolutional neural networks
US20190220650A1 (en) Systems and methods for depth map sampling
CN105447459B (en) A kind of unmanned plane detects target and tracking automatically
CN106845547B (en) A kind of intelligent automobile positioning and road markings identifying system and method based on camera
US9196054B2 (en) Method and system for recovery of 3D scene structure and camera motion from a video sequence
Kochenderfer et al. Airspace encounter models for estimating collision risk
US20160171896A1 (en) Commercial and General Aircraft Avoidance using Light Pattern Detection
CN102253381B (en) System and method for automatically detecting foreign object debris (FOD) on airfield runways
CN107742093B (en) Real-time detection method, server and system for infrared image power equipment components
US8509965B2 (en) Integrated collision avoidance system for air vehicle
US9685089B2 (en) Commercial and general aircraft avoidance using acoustic pattern recognition
CN102654940B (en) Processing method of traffic information acquisition system based on unmanned aerial vehicle and
Zhang et al. Deep convolutional neural networks for forest fire detection
CN103150559B (en) Head recognition and tracking method based on Kinect three-dimensional depth image
US7043355B2 (en) Multisource target correlation
US6810322B2 (en) Multisource target correlation
Guido et al. Evaluating the accuracy of vehicle tracking data obtained from Unmanned Aerial Vehicles
CN107862705B (en) Unmanned aerial vehicle small target detection method based on motion characteristics and deep learning characteristics
FR3037429A1 (en) System and method for automatic surface inspection
CN104320617B (en) A kind of round-the-clock video frequency monitoring method based on deep learning
CN105303748B (en) Based on the fire alarm system taken photo by plane in the air
CN105913604A (en) Fire occurrence determining method and device based on unmanned aerial vehicle

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