CN109255286A - A kind of quick detection recognition method of unmanned plane optics based on YOLO deep learning network frame - Google Patents
A kind of quick detection recognition method of unmanned plane optics based on YOLO deep learning network frame Download PDFInfo
- Publication number
- CN109255286A CN109255286A CN201810807503.8A CN201810807503A CN109255286A CN 109255286 A CN109255286 A CN 109255286A CN 201810807503 A CN201810807503 A CN 201810807503A CN 109255286 A CN109255286 A CN 109255286A
- Authority
- CN
- China
- Prior art keywords
- yolo
- layer
- unmanned plane
- network frame
- deep learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000013135 deep learning Methods 0.000 title claims abstract description 19
- 238000002474 experimental method Methods 0.000 claims abstract description 12
- 238000012634 optical imaging Methods 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims abstract description 11
- 230000006870 function Effects 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 5
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 230000003287 optical effect Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 5
- 238000004880 explosion Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000003796 beauty Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000003475 lamination Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
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
- G06F18/24—Classification techniques
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
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)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of quick detection recognition methods of unmanned plane optics based on YOLO deep learning network frame, the method comprises the following steps: step 1: carrying out flight test acquisition optical imaging experiments data for five kinds of unmanned planes of existing market mainstream, and handled according to optical imaging experiments data of the standard VOC data format to acquisition；Step 2: building YOLO network frame, is improved using residual error network module to YOLO network frame, is trained to improved YOLO network frame, obtains detection identification model；Step 3: the real scene shooting flight optical imaging experiment data comprising five kinds of unmanned planes are chosen, carry out detection identification using the resulting detection identification model of step 2.The invention avoids the problems such as complicated, applicability is not strong is manually modeled to unmanned plane and complex background feature, the speed and accuracy rate of moving-target detection identification under complex background can be greatly improved.
Description
Technical field
The invention belongs to technical field of image processing, are related to a kind of unmanned plane detection recognition method based on optical imagery,
More particularly to a kind of unmanned plane optics for being based on YOLO (You Only Look Once, only have a look at) deep learning network frame
Quick detection recognition method.
Background technique
In recent years, unmanned air vehicle technique is grown rapidly, and is widely used to take photo by plane, agricultural plant protection, traffic monitoring, calamity
Area's detection, power-line patrolling, quickly numerous Military and civil fields such as investigation, surprise attack.But spreading unchecked for unmanned plane market also brings various peaces
Full hidden danger, the unmanned plane that especially do not abide by supervision regulation, surmount the flight of regulation airspace, gives airport airline carriers of passengers safety, emphasis
The security protection of area peripheral edge causes great threat.Therefore, need to develop China's critical facility periphery, crowd massing place, important
The monitoring of aerial unmanned plane and the Prevention-Security measure of the key areas such as military district periphery.
Currently, the air surveillance of unmanned plane and defence be mainly based on radar, GPS signal, as ThaleSA (method),
DroneShield (beauty), JAMMER (ink) company are monitored unmanned plane using radar, big vast electronics technology invention behind Shanghai
Portable unmanned machine interference unit, which then passes through no-manned machine distant control signal and the compacting of GPS positioning signal, makes unmanned plane fall or make a return voyage.
But the method based on radar or GPS signal is mainly used for remote quick discovery, capture and the tracking of unmanned plane, still difficult at present
To realize that efficient, high-accuracy the detection to unmanned plane type identifies.And detection knowledge is carried out to unmanned plane using optical instrument
It does not have the advantage that 1) compared to radar, it is thin that optical imagery can obtain the targets such as gray scale more abundant, texture, structure
Information is saved, the high-accuracy identification of unmanned plane type is more suitable for；2) compared to GPS, optical instrument belongs to passive mode, is not required to
Target is wanted actively to receive signal.Therefore the unmanned machine testing identification based on optical imagery has become important trend.Traditional
Optical detection recognition methods is the first artificial suitable classifier of construction feature reselection, background is simple, target invariant features
It is showed in metastable situation good.And unmanned plane, in the sky in flight course, relative position, posture become constantly
Change, is difficult to find the characteristics of image for meeting Scale invariant, angle invariability, and the background during unmanned plane during flying is equally
Movement, and it is likely to complicated and changeable.
Summary of the invention
The present invention is difficult to realize using means such as radar, GPS to the high-accuracy identification of unmanned plane classification, biography for current
The problems such as artificial constructed process of the optical signature of system is cumbersome, generalization ability is not strong proposes a kind of based on YOLO deep learning net
The quick detection recognition method of unmanned plane optics of network frame.This method can utilize actual test flight data, learning training
Network, directly output recognition result are avoided compared to traditional detection identification process to unmanned plane and complex background feature people
The problems such as work modeling is complicated, applicability is not strong can greatly improve under complex background the speed of moving-target detection identification and accurate
Rate.
The purpose of the present invention is what is be achieved through the following technical solutions:
A kind of quick detection recognition method of unmanned plane optics based on YOLO deep learning network frame, including walk as follows
It is rapid:
Step 1: carrying out flight test acquisition optical imaging experiments data for five kinds of unmanned planes of existing market mainstream,
And it is handled according to optical imaging experiments data of the standard VOC data format to acquisition；
Step 2: building YOLO network frame, is improved using residual error network module to YOLO network frame, to improvement
YOLO network frame afterwards is trained, and obtains detection identification model；
Step 3: choosing the real scene shooting flight optical imaging experiment data comprising five kinds of unmanned planes, resulting using step 2
Detection identification model carries out detection identification.
Compared with the prior art, the present invention has the advantage that
(1) present invention in view of the target detection recognition methods based on deep learning be by feature construction and fusion for classification at
One entirety, i.e. input are initial data, direct output category result, do not need artificial constructed feature, are more suitable for solving multiple
The automatic detection identification problem of moving-target under miscellaneous dynamic background, proposes a kind of unmanned plane light based on YOLO deep learning network frame
Learn quick, autonomous detection recognition method.
(2) method of the invention is answered suitable for various complexity such as airport, stadium, concert, important military district peripheries
Quick discovery, high-accuracy detection and the identification invaded with unmanned plane under scene are the effective supervision and defense in the air of unmanned plane
It provides and supports.
(3) present invention is in unmanned plane in the sky flight course, and relative position, posture are constantly changing, very
Difficulty finds the characteristics of image for meeting scale invariability, angle invariability, and the background during unmanned plane during flying is usually complicated
It is the problems such as changeable, theoretical based on deep learning, using heterogeneous networks layer (convolutional layer, pond layer, recurrence layer etc.) function, in conjunction with spy
Different network configuration design, realizes automation, the abstract expression of target signature under complex situations, and is automatically performed target point
Class.Compared to traditional target detection identification process, this method can greatly improve moving-target detection identification under complex background
Speed and accuracy rate.
(4) present invention selects YOLO network frame, which, which will test, appoints to realize that the quick detection of unmanned plane identifies
Business is handled as regression problem, directly obtains bounding box coordinates and setting comprising object by all pixels of entire image
Reliability and class probability are significantly better than R-CNN, Fast R-CNN even deep learning network frame in detection speed.And it utilizes
Residual error network module improves YOLO network, reduces a possibility that gradient explosion or gradient disappear in training process,
To effectively improve the usable probability of training pattern, identification is quickly detected to the unmanned plane based on YOLO deep learning network frame
The practical application of method is of great significance.
Detailed description of the invention
Fig. 1 is that the present invention is based on the quick detection recognition method processes of the unmanned plane optics of YOLO deep learning network frame
Figure；
Fig. 2 is VOC standard data format example；
Fig. 3 is network structure；
Fig. 4 is the sample image in database；
Fig. 5 is real image test result.
Specific embodiment
Further description of the technical solution of the present invention with reference to the accompanying drawing, and however, it is not limited to this, all to this
Inventive technique scheme is modified or replaced equivalently, and without departing from the spirit and scope of the technical solution of the present invention, should all be covered
Within the protection scope of the present invention.
The present invention provides a kind of quick detection recognition method of unmanned plane optics based on YOLO deep learning network frame,
As shown in Figure 1, step specific as follows:
Step 1: carrying out flight test acquisition optical imaging experiments data for five kinds of unmanned planes of existing market mainstream,
And it is handled according to optical imaging experiments data of the standard VOC data format to acquisition.Specific step is as follows:
Big boundary M100, Phantom-3, Inspire-1, agricultural plant protection unmanned plane, police unmanned plane is selected to carry out flight examination
It tests acquisition experimental data, and operations is labeled etc. to experimental data and comply with VOC data format, treated specific format
As shown in Figure 2.
Step 2: building YOLO network frame, is improved using residual error network module to YOLO network frame, to improvement
YOLO network frame afterwards is trained, and obtains detection identification model.Specific step is as follows:
The detection method of YOLO is single order detection method, input picture is divided into S × S grid, each grid is responsible for inspection
Survey the object of " falling into " grid.If the coordinate of the center of some object drops into some grid, then this grid is just
It is responsible for detecting this object.The output information of each grid includes two large divisions, is comprising object rectangular area information respectively
B boundingbox (bounding box) information and C object belong to the probabilistic information of certain classification.
Bounding box information includes 5 values, is x, y, w, h and confidence (confidence level) respectively.Wherein x, y
Refer to deviant of the center of the bounding box for the object that current grid is predicted relative to current grid position,
And it is normalized to the coordinate of [0,1].W, h are the width of bounding box with height and using the width and height of image
It is normalized in [0,1].
Confidence reflect current bounding box whether include object and object space accuracy, calculating side
Formula is as follows:
Confidence=P (object) IOU.
Wherein, it indicates that bounding box includes object (target object) as P (object)=1, does not include then P
(object)=0；IOU (hand over and compare) is the intersection area for predicting bounding box and object real estate, and area is with pixel
As a result areal calculation is normalized in [0,1] section with real estate area elemental area.
It builds YOLO network (shown in network structure such as Fig. 3 (a)), network includes 24 convolutional layers and 2 full articulamentums, volume
Lamination is used for future position and class probability for extracting characteristics of image, full articulamentum.The loss function definition of network is such as
Under:
Each meaning of parameters of above-mentioned formula is as follows: S2Indicate that image segmentation is the number of grid, value is 13 × 13；B is indicated
The number of anchor point frame in each grid, value 5；X, y, w, h are the centre coordinate and wide height of predicted boundary frame；Indicate the centre coordinate and wide height of actual boundary frame；C indicates the confidence level in predicted boundary frame comprising target；Indicate the friendship of realistic objective and realistic objective bounding box and ratio；P (c) indicates the prediction probability for belonging to a certain classification；Table
Show the actual probabilities for belonging to a certain classification, belonging to then value is 1, and otherwise value is 0；IobjIndicate anchor point frame in whether inclusion
Otherwise body is 0 comprising being then 1；λcoordIndicate that position prediction loses weight, value 3；λnoobjMesh when indicating without target
Mark confidence weight, value 0.7.
First two of function are coordinate prediction loss, and the confidence that Section 3 is the box comprising object predicts damage
It loses, Section 4 is that the confidence not comprising object predicts loss, and Section 5 is class prediction loss.
Activation primitive is defined as follows in network:
Gradient explosion is generated when to avoid training or gradient disappears, and the present invention changes this network using residual error module
Into after improvement shown in network structure such as Fig. 3 (b), specific location is as follows: the 3rd layer of pond layer output steering and the 9th layer of convolutional layer are defeated
It merges out, the 12nd layer of convolutional layer output steering is merged with the 15th layer of convolutional layer output, the 4th layer of pond layer output steering
It is merged with the 18th layer of convolutional layer output, the 19th layer of convolutional layer output steering is merged with the 22nd layer of convolutional layer output.?
It is added at 4 after short circuit connects and composes residual unit, the gradient explosion or gradient when effectively alleviating trained disappear.Residual error at 4
Unit part gradient circulation way is as follows:
Loss=F (xi,Wi)
In formula, loss indicates loss function, xi、WiFor i-th layer of network input and i-th layer of weight, loss function is expressed as defeated
Enter the function F (x with weighti,Wi), xLIndicate the output of residual error module shunting layer, xlIt indicates to export at the merging of residual error module, first
A factorIndicate that loss function reaches L layers of gradient, 1 in round bracket shows that short-circuit mechanism can be propagated nondestructively
Gradient, and an other residual error gradient then needs the layer by having weight.
Hyper parameter setting is carried out to network, sets 0.0003 for initial learning rate, it is (i.e. every using stochastic gradient descent method
Secondary update using sample size is 1, and update times are 50000) to train network in total, obtain detection identification model.
Step 3: choosing the real scene shooting flight optical imaging experiment data comprising five class unmanned planes, resulting using step 2
Identification model is detected, testing image is inputted, feature, pond layer downscaled images size, finally by connecting entirely are extracted by convolutional layer
Layer output target prodiction value and target category probabilistic forecasting value are connect, it is identification that target category maximum probability value, which corresponds to classification,
As a result.As shown in Figure 5 known to each classification detection recognition correct rate statistical result: M100 91.96%, Inspire-1 are
91.74%, Phantom-3 89.78%, agriculture unmanned plane are 94.13%, and police unmanned plane is 89.84%, have reached higher
Detection recognition correct rate, and in the case where being accelerated using GPU, every frame processing speed can reach a millisecond rank, realize quick
Detection identification.
Claims (6)
1. a kind of quick detection recognition method of unmanned plane optics based on YOLO deep learning network frame, it is characterised in that described
Method and step is as follows:
Step 1: carry out flight test acquisition optical imaging experiments data for five kinds of unmanned planes of existing market mainstream, and press
The quasi- VOC data format of sighting target handles the optical imaging experiments data of acquisition；
Step 2: building YOLO network frame, is improved using residual error network module to YOLO network frame, to improved
YOLO network frame is trained, and obtains detection identification model；
Step 3: the real scene shooting flight optical imaging experiment data comprising five kinds of unmanned planes are chosen, the resulting detection of step 2 is utilized
Identification model carries out detection identification.
2. the unmanned plane optics according to claim 1 based on YOLO deep learning network frame quickly detects identification side
Method, it is characterised in that five kinds of unmanned planes are big boundary M100, Phantom-3, Inspire-1, agricultural plant protection unmanned plane, police
Unmanned plane.
3. the unmanned plane optics according to claim 1 based on YOLO deep learning network frame quickly detects identification side
Method, it is characterised in that specific step is as follows for the step 2:
(1) input picture is divided into 13 × 13 grid by YOLO, and each grid is responsible for detecting the object of " falling into " grid, each
The output information of grid include two large divisions, respectively be comprising object rectangular area information 5 boundingbox information and
C object belongs to the probabilistic information of certain classification；
(2) YOLO network is built, network includes 24 convolutional layers and 2 full articulamentums, and convolutional layer is for extracting characteristics of image, entirely
Articulamentum is used for future position and class probability.
(3) YOLO network is improved using residual error module, short circuit at 4 is added and connects and composes residual unit, residual unit at 4
Part gradient circulation way is as follows:
Loss=F (xi,Wi)
In formula, loss indicates loss function, xi、WiFor i-th layer of network input and i-th layer of weight, loss function be expressed as input with
Function F (the x of weighti,Wi), xLIndicate the output of residual error module shunting layer, xlIndicate the merging of residual error module at export, first because
SonIndicate that loss function reaches L layers of gradient；
(4) hyper parameter setting is carried out to the improved YOLO network of step (3), sets 0.0003 for initial learning rate, uses
Stochastic gradient descent method trains network, obtains detection identification model.
4. the unmanned plane optics according to claim 3 based on YOLO deep learning network frame quickly detects identification side
Method, it is characterised in that in the step (3), the specific location that short circuit connects at 4 is as follows: the 3rd layer of pond layer output steering and the 9th
Layer convolutional layer output merges, and the 12nd layer of convolutional layer output steering is merged with the 15th layer of convolutional layer output, the 4th layer of pond
Layer output steering is exported with the 18th layer of convolutional layer and is merged, the 19th layer of convolutional layer output steering and the 22nd layer of convolutional layer export into
Row merges.
5. the unmanned plane optics according to claim 3 based on YOLO deep learning network frame quickly detects identification side
Method, it is characterised in that in the step (4), the update times in total of stochastic gradient descent method training network are 50000.
6. the unmanned plane optics according to claim 1 based on YOLO deep learning network frame quickly detects identification side
Method, it is characterised in that in the step 3, it is as follows to carry out detection knowledge method for distinguishing using the resulting detection identification model of step 2:
Testing image is inputted, feature is extracted by convolutional layer, pond layer downscaled images size exports target position finally by full articulamentum
Predicted value and target category probabilistic forecasting value are set, target category maximum probability value corresponds to the result that classification is identification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810807503.8A CN109255286B (en) | 2018-07-21 | 2018-07-21 | Optical fast detection and recognition method of UAV based on deep learning network framework |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810807503.8A CN109255286B (en) | 2018-07-21 | 2018-07-21 | Optical fast detection and recognition method of UAV based on deep learning network framework |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109255286A true CN109255286A (en) | 2019-01-22 |
| CN109255286B CN109255286B (en) | 2021-08-24 |
Family
ID=65049063
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810807503.8A Active CN109255286B (en) | 2018-07-21 | 2018-07-21 | Optical fast detection and recognition method of UAV based on deep learning network framework |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109255286B (en) |
Cited By (12)
| 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 |
| CN110490155A (en) * | 2019-08-23 | 2019-11-22 | 电子科技大学 | A detection method for unmanned aerial vehicles in no-fly airspace |
| CN110850897A (en) * | 2019-11-13 | 2020-02-28 | 中国人民解放军空军工程大学 | Pose and attitude data collection method for small unmanned aerial vehicles based on deep neural network |
| CN111611918A (en) * | 2020-05-20 | 2020-09-01 | 重庆大学 | Traffic flow dataset collection and construction method based on aerial photography data and deep learning |
| CN111723690A (en) * | 2020-06-03 | 2020-09-29 | 北京全路通信信号研究设计院集团有限公司 | Circuit equipment state monitoring method and system |
| CN111797940A (en) * | 2020-07-20 | 2020-10-20 | 中国科学院长春光学精密机械与物理研究所 | An image recognition method and related device based on marine search and rescue |
| CN112329768A (en) * | 2020-10-23 | 2021-02-05 | 上善智城(苏州)信息科技有限公司 | Improved YOLO-based method for identifying fuel-discharging stop sign of gas station |
| CN112668445A (en) * | 2020-12-24 | 2021-04-16 | 南京泓图人工智能技术研究院有限公司 | Vegetable type detection and identification method based on yolov5 |
| CN112699810A (en) * | 2020-12-31 | 2021-04-23 | 中国电子科技集团公司信息科学研究院 | Method and device for improving figure identification precision of indoor monitoring system |
| CN113822375A (en) * | 2021-11-08 | 2021-12-21 | 北京工业大学 | Improved traffic image target detection method |
| CN113822372A (en) * | 2021-10-20 | 2021-12-21 | 中国民航大学 | Unmanned aerial vehicle detection method based on YOLOv5 neural network |
| CN118942037A (en) * | 2024-07-26 | 2024-11-12 | 北京昊普康科技股份有限公司 | A method for intelligent identification of power safety violations |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017077348A1 (en) * | 2015-11-06 | 2017-05-11 | Squarehead Technology As | Uav detection |
| CN107862705A (en) * | 2017-11-21 | 2018-03-30 | 重庆邮电大学 | A kind of unmanned plane small target detecting method based on motion feature and deep learning feature |
-
2018
- 2018-07-21 CN CN201810807503.8A patent/CN109255286B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017077348A1 (en) * | 2015-11-06 | 2017-05-11 | Squarehead Technology As | Uav detection |
| CN107862705A (en) * | 2017-11-21 | 2018-03-30 | 重庆邮电大学 | A kind of unmanned plane small target detecting method based on motion feature and deep learning feature |
Non-Patent Citations (3)
| Title |
|---|
| XIE, B等: "Research on UAV target recognition algorithm based on transfer learning SAE", 《INFRARED AND LASER ENGINEERING》 * |
| 王靖宇等: "基于深度神经网络的低空弱小无人机目标检测研究", 《西北工业大学学报》 * |
| 蒋兆军等: "基于深度学习的无人机识别算法研究", 《电子技术应用》 * |
Cited By (18)
| 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 |
| CN110490155B (en) * | 2019-08-23 | 2022-05-17 | 电子科技大学 | Method for detecting unmanned aerial vehicle in no-fly airspace |
| CN110490155A (en) * | 2019-08-23 | 2019-11-22 | 电子科技大学 | A detection method for unmanned aerial vehicles in no-fly airspace |
| CN110850897A (en) * | 2019-11-13 | 2020-02-28 | 中国人民解放军空军工程大学 | Pose and attitude data collection method for small unmanned aerial vehicles based on deep neural network |
| CN110850897B (en) * | 2019-11-13 | 2023-06-13 | 中国人民解放军空军工程大学 | A small UAV pose data acquisition method based on deep neural network |
| CN111611918A (en) * | 2020-05-20 | 2020-09-01 | 重庆大学 | Traffic flow dataset collection and construction method based on aerial photography data and deep learning |
| CN111611918B (en) * | 2020-05-20 | 2023-07-21 | 重庆大学 | Acquisition and Construction Method of Traffic Flow Dataset Based on Aerial Photography Data and Deep Learning |
| CN111723690A (en) * | 2020-06-03 | 2020-09-29 | 北京全路通信信号研究设计院集团有限公司 | Circuit equipment state monitoring method and system |
| CN111723690B (en) * | 2020-06-03 | 2023-10-20 | 北京全路通信信号研究设计院集团有限公司 | Method and system for monitoring state of circuit equipment |
| CN111797940A (en) * | 2020-07-20 | 2020-10-20 | 中国科学院长春光学精密机械与物理研究所 | An image recognition method and related device based on marine search and rescue |
| CN112329768A (en) * | 2020-10-23 | 2021-02-05 | 上善智城(苏州)信息科技有限公司 | Improved YOLO-based method for identifying fuel-discharging stop sign of gas station |
| CN112668445A (en) * | 2020-12-24 | 2021-04-16 | 南京泓图人工智能技术研究院有限公司 | Vegetable type detection and identification method based on yolov5 |
| CN112699810A (en) * | 2020-12-31 | 2021-04-23 | 中国电子科技集团公司信息科学研究院 | Method and device for improving figure identification precision of indoor monitoring system |
| CN112699810B (en) * | 2020-12-31 | 2024-04-09 | 中国电子科技集团公司信息科学研究院 | Method and device for improving character recognition precision of indoor monitoring system |
| CN113822372A (en) * | 2021-10-20 | 2021-12-21 | 中国民航大学 | Unmanned aerial vehicle detection method based on YOLOv5 neural network |
| CN113822375A (en) * | 2021-11-08 | 2021-12-21 | 北京工业大学 | Improved traffic image target detection method |
| CN113822375B (en) * | 2021-11-08 | 2024-04-26 | 北京工业大学 | Improved traffic image target detection method |
| CN118942037A (en) * | 2024-07-26 | 2024-11-12 | 北京昊普康科技股份有限公司 | A method for intelligent identification of power safety violations |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109255286B (en) | 2021-08-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109255286A (en) | A kind of quick detection recognition method of unmanned plane optics based on YOLO deep learning network frame | |
| CN103679674B (en) | Method and system for splicing images of unmanned aircrafts in real time | |
| CN112068111A (en) | A UAV target detection method based on multi-sensor information fusion | |
| Shi et al. | Objects detection of UAV for anti-UAV based on YOLOv4 | |
| CN106529538A (en) | Method and device for positioning aircraft | |
| Ye et al. | CT-Net: An efficient network for low-altitude object detection based on convolution and transformer | |
| CN108875754B (en) | A vehicle re-identification method based on multi-depth feature fusion network | |
| CN110619276A (en) | Anomaly and violence detection system and method based on unmanned aerial vehicle mobile monitoring | |
| CN116846059A (en) | Edge detection system for power grid inspection and monitoring | |
| CN107742276A (en) | One kind is based on the quick processing system of the airborne integration of unmanned aerial vehicle remote sensing image and method | |
| CN111444801A (en) | A real-time detection method for infrared target of unmanned aerial vehicle | |
| Shangzheng | A traffic sign image recognition and classification approach based on convolutional neural network | |
| CN116109950A (en) | Low-airspace anti-unmanned aerial vehicle visual detection, identification and tracking method | |
| Ghosh et al. | AirTrack: Onboard deep learning framework for long-range aircraft detection and tracking | |
| Tong et al. | UAV target detection based on RetinaNet | |
| Ding et al. | Research on UAV detection technology of Gm-APD Lidar based on YOLO model | |
| Cheng et al. | Moving target detection technology based on UAV Vision | |
| Zhai et al. | Target Detection of Low‐Altitude UAV Based on Improved YOLOv3 Network | |
| Niu et al. | UAV detection based on improved YOLOv4 object detection model | |
| Wu et al. | Research on asphalt pavement disease detection based on improved YOLOv5s | |
| Wu et al. | Multimodal collaboration networks for geospatial vehicle detection in dense, occluded, and large-scale events | |
| Tsapparellas et al. | Vision-based runway detection and landing for unmanned aerial vehicle enhanced autonomy | |
| Delleji et al. | An Improved YOLOv5 for Real-time Mini-UAV Detection in No Fly Zones. | |
| CN115937700B (en) | A multi-source collaborative moving target online detection and recognition method | |
| CN107045805A (en) | A kind of monitoring method and system for small-sized aerial craft and thing drifted by wind |
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 | ||
| CB03 | Change of inventor or designer information |
Inventor after: Zhi Xiyang Inventor after: Yu Lijian Inventor after: Hu Jianming Inventor after: Gong Jinnan Inventor after: Jiang Shikai Inventor after: Chen Wenbin Inventor before: Zhi Xiyang Inventor before: Yu Lijian Inventor before: Gong Jinnan Inventor before: Jiang Shikai Inventor before: Chen Wenbin Inventor before: Hu Jianming |
|
| CB03 | Change of inventor or designer information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |