CN105929845A - Unmanned aerial vehicle network-based river channel cruise system and cruise method - Google Patents

Unmanned aerial vehicle network-based river channel cruise system and cruise method Download PDF

Info

Publication number
CN105929845A
CN105929845A CN201610333740.6A CN201610333740A CN105929845A CN 105929845 A CN105929845 A CN 105929845A CN 201610333740 A CN201610333740 A CN 201610333740A CN 105929845 A CN105929845 A CN 105929845A
Authority
CN
China
Prior art keywords
unmanned aerial
aerial vehicle
river channel
hangar
layer
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
Application number
CN201610333740.6A
Other languages
Chinese (zh)
Other versions
CN105929845B (en
Inventor
郑恩辉
巫岳
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.)
China Jiliang University
Original Assignee
China Jiliang 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 China Jiliang University filed Critical China Jiliang University
Priority to CN201610333740.6A priority Critical patent/CN105929845B/en
Publication of CN105929845A publication Critical patent/CN105929845A/en
Application granted granted Critical
Publication of CN105929845B publication Critical patent/CN105929845B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/10—Simultaneous control of position or course in three dimensions
    • G05D1/101—Simultaneous control of position or course in three dimensions specially adapted for aircraft

Landscapes

  • Engineering & Computer Science (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses an unmanned aerial vehicle network-based river channel cruise system and an unmanned aerial vehicle network-based river channel cruise method. The unmanned aerial vehicle network-based river channel cruise system includes unmanned aerial vehicles, hangars and a ground control center; the hangars are adopted as the arrangement platforms of the unmanned aerial vehicles; the plurality of hangars are arranged at intervals along one side of a river channel; the pan-tilt mechanisms of the unmanned aerial vehicles are provided with cameras and microphones; communication and data transmission are carried between the ground control center and the unmanned aerial vehicles and between the ground control center and the hangars; each unmanned aerial vehicle flies to the center of the river channel where the corresponding hangar is located; when flying along the river channel and photographing the next hangar being located at the center of the river channel, the corresponding unmanned aerial vehicle turns back and photographs the other side of the river channel; and the ground control center carries out image processing according to returned video information. With the unmanned aerial vehicle network-based river channel cruise system and cruise method of the invention adopted, a lot of manpower and material resources can be saved, all-directional and zero-dead angle river channel cruise can be realized by using the unmanned aerial vehicles, and various kinds of defects of an existing river channel cruise method in the prior art can be eliminated.

Description

Unmanned aerial vehicle network-based river channel cruise system and cruise method
Technical Field
The invention relates to the field of river channel cruising, in particular to a river channel cruising system and a cruising method based on an unmanned aerial vehicle network, which are suitable for river channel cruising for checking illegal ships of river channels, illegal conditions of river banks and beaches and the like.
Background
In recent years, river channel traffic and river bank mudflat construction are receiving more and more attention, and meanwhile, the problems of river channel traffic safety and mudflat violation are more and more serious. The conventional river channel cruise mostly utilizes manual inspection or fixed-point cameras to acquire data. The method needs high cost of manpower and material resources, and has a blind area for collecting the visual field.
The multi-rotor unmanned aerial vehicle network has the advantages of reasonable distribution, good maneuverability, easy control, low cost and the like, can shoot the river channel site in an all-round and dead-angle-free manner by utilizing the camera in a short time, can effectively reduce the cost of manpower and material resources by combining with an image processing automatic identification method, and has a very favorable effect on river channel cruising.
Disclosure of Invention
In order to solve the problems, the invention provides a river channel cruising system and a cruising method based on an unmanned aerial vehicle network.
The technical scheme adopted by the invention is as follows:
a river channel cruising method based on an unmanned aerial vehicle network comprises the following steps:
1) the ground control center controls the unmanned aerial vehicle to take off from the hangar, the ground control center sends out a take-off signal, all hangar doors arranged beside a river section to be detected in a riverway are opened, the internal unmanned aerial vehicle takes off after receiving the take-off signal, the common unmanned aerial vehicle takes off under the normal condition, and if the common unmanned aerial vehicle fails, the standby unmanned aerial vehicle takes off;
a GPS positioning module, a barometer, a gyroscope, an accelerometer, a camera and a loudspeaker which are arranged on the unmanned aerial vehicle are all opened;
2) each unmanned aerial vehicle flies to the center of a river channel where a self-body hangar is located through a GPS positioning module, then flies along the river channel, controls the flying height through a barometer, controls the aircraft to keep flying in a horizontal straight line through a gyroscope and an accelerometer, a camera faces to the right side of the flying direction of the unmanned aerial vehicle, transmits a real-time video back to a ground control center through an RTP (real-time transport protocol), and the ground control center performs image processing according to transmitted video information;
if the situation of illegal watercourse ships is detected, the camera needs to be aligned to the river surface; if the condition of the river bank mudflat violation is checked, the camera needs to be aligned to the river bank mudflat.
3) When each unmanned aerial vehicle flies to the center of a river channel where the next hangar is located along the river channel, the airframe rotates 180 degrees and returns along the original path, the orientation of the camera is kept unchanged relative to the airframe, and the other half of the river bank which is not shot before cruising is at the moment;
4) after the unmanned aerial vehicle flies back to the center of the river channel where the self hangar is located, the ground control center controls the unmanned aerial vehicle to land in the hangar.
The method of the invention finds that the illegal condition exists on the river channel, no one can take surrounding flight shooting to further confirm the area, people in the control center can also take on-site law enforcement through voice playing of the loudspeaker, and the shot image is stored as the illegal evidence.
When the unmanned aerial vehicle takes off, all hangars receive the hangar door opening signal of the ground control center to open the hangar door, the ground control center sends a take-off signal to the unmanned aerial vehicle, and the unmanned aerial vehicle receives the take-off signal to take off to the center of a river channel.
When unmanned aerial vehicle descends, all hangars receive ground control center's hangar door opening signal and open the hangar door, and ground control center sends the descending signal to unmanned aerial vehicle, and unmanned aerial vehicle receives the descending signal and descends, and the river course cruises from this and finishes.
The transmitted video information is subjected to image processing and judgment by adopting a tidal flat scene image classifier trained by a convolutional neural network to obtain the tidal flat illegal condition: inputting the images into a tidal flat scene image classifier trained through a convolutional neural network, classifying the images by using training characteristics in a full connection layer of the convolutional neural network, and obtaining polluted or uncontaminated tidal flat illegal conditions through classification judgment.
The convolutional neural network simultaneously trains characteristics and classification in an end-to-end mode, the trained characteristics are full connection layers at the tail ends of the convolutional neural network structure, the full connection layers at the tail ends of the convolutional neural network structure are extracted and output to serve as depth characteristics of a polluted image, the obtained depth characteristics are input into a support vector machine classifier to be trained, and the mudflat scene image classifier is obtained.
The convolutional neural network training process specifically includes the following steps for all input sample images:
1) firstly, carrying out size normalization to normalize the image to 32 multiplied by 32 resolution;
2) and (3) convolutional layer calculation: parametric expression of convolutional layers: 4 × 4 × 20+1 (step size), representing a convolution kernel size of 4 × 4, a number of 20, and a step size of 1, the convolution layer calculation is performed using the following formula:
x j l = f ( z j l )
z j l = Σ i ∈ M j x i l - 1 * k i j l + b j l
wherein,respectively representing the jth characteristic diagram in the current convolution layer and the ith characteristic diagram in the previous layer;a convolution kernel between the jth characteristic diagram representing the current layer and the ith characteristic diagram representing the previous layer; mjA set of feature maps representing a previous layer requiring convolution,representing the bias corresponding to the jth convolution kernel in the current convolution layer; f is an activation function, and the weight and the threshold in the convolutional layer are obtained by a random gradient descent method; i and j both represent ordinal numbers of the feature map, and l represents step size;
3) and (3) pool layer calculation: the pooling layer adopts a maximum pooling mode, 2 multiplied by 2+2 (step length) indicates that the size of a pooling core is 2 multiplied by 2, and the step length is 2; the maximum pooling calculation uses the following formula:
h c d o = m a x { h c · s + m , d · s + n 0 ≤ m ≤ s , 0 ≤ n ≤ s o - 1 }
wherein the size of the pooling nucleus is s × s, s represents the side length of the pooling nucleus,a value representing the (c, d) th coordinate position in the current pooling feature map, c, d representing the abscissa and ordinate of the pooling feature map respectively,representing the set of values in the pooling kernel in the previous layer of feature map;representing the coordinate position in the previous layer of feature map as c.s + m, d.s + n, wherein m and n represent the coordinates in the pooling core;
4) convolution layer calculation is performed again: parametric expression of convolutional layers: 3 × 3 × 40+1 (step size), which means that the size of the convolution kernel is 3 × 3, the number is 40, the step size is 1, and the specific calculation of the convolution layer in this step is the same as that in step 2);
5) performing pooling layer calculation again by adopting the same process as the step 3):
6) calculating a full connection layer: the network comprises a full connection layer and two node units in total, and the calculation formula of the full connection layer is as follows:
R q = f ( Σ p = 1 N u p · v p q )
wherein f (-) represents an activation function, and a Rectified linear unit activation function is adopted, RqDenotes the q-th node unit, u, in the full connection layerpDenotes the p-th node element in the pooling layer, vpqRepresenting the connection weight of two nodes, N representing the number of node units of the pooling layer, and p and q being ordinal numbers of the node units of the pooling layer;
7) the calculation formula of the softmax layer is as follows:
r t = e r t , Σ t = 1 a e r t ,
r t ′ = Σ t = 1 a X t · W t , w + B t
in the formula, Wt,wRepresents the connection weight between the full connection layer and the softmax layer, BtTo be offset, XtThe t-th node element value, r, representing the full link layertRepresenting the probability value, r, over the t-th categoryt' represents the probability density of the t-th class; a represents the total number of categories, t is the ordinal number of the category, t is 1 or 2, a is 2, and w is the ordinal number of the connecting weight under the t-th category; r is1Representing the probability of an input mudflat illegal image being contaminated, r2Representing the probability that the input mudflat illegal image is uncontaminated;
8) the loss function formula of the convolutional neural network is as follows:
J ( θ ) = - 1 P [ Σ C = 1 P Σ D = 1 Q 1 { g ( P ) = Q } l o g g Q ( P ) ]
wherein 1{ } is an indicative function, and θ represents a network parameter;the probability value of the P-th sample on the class Q is shown, P represents the number of images in one iteration batch, Q represents the number of image classes, and C and D are ordinal numbers of the samples.
Secondly, a river course cruise system based on unmanned aerial vehicle network:
the method comprises the following steps that a multi-unmanned aerial vehicle cooperative river channel cruise system mainly comprising an unmanned aerial vehicle, a hangar and a ground control center is adopted, the hangar serves as a placing platform of the unmanned aerial vehicle, and the hangars are arranged at intervals along one side of a river channel; the unmanned aerial vehicle is loaded with a GPS positioning module, a barometer, a gyroscope, an accelerometer, a cradle head mechanism and a wireless two-way communication module, and the cradle head mechanism is provided with a camera and a loudspeaker; the ground control center communicates with the hangar and the unmanned aerial vehicle and transmits data, and sends the control signals of takeoff and landing signals, flight path information, the camera and the loudspeaker to the unmanned aerial vehicle to control takeoff and landing of the unmanned aerial vehicle and flight tracks and control opening and closing of the camera and the loudspeaker.
The hangar is provided with a common unmanned aerial vehicle and a standby unmanned aerial vehicle as well as a common power supply and a standby power supply which can charge the unmanned aerial vehicle, and the hangar receives signals sent by a ground control center in the take-off and landing processes of the unmanned aerial vehicle to open and close a hangar door.
The shape of the hangar is cuboid, and the unmanned aerial vehicle machine heads are parallelly arranged in the hangar side by side.
The ground control center and the unmanned aerial vehicle between carry out communication and data transmission, obtain the photo and the video of shooing from the unmanned aerial vehicle to erect the database that is equipped with storage unmanned aerial vehicle data collection and image, the communication adopts 4G honeycomb mobile network to communicate, data transmission uses the TCP agreement, video and audio data use the RTP agreement to transmit.
Unmanned aerial vehicle on GPS orientation module be used for receiving GPS satellite's real-time position data, the barometer is used for detecting unmanned aerial vehicle's flying height, the gyroscope is used for detecting unmanned aerial vehicle's flight angular velocity, the accelerometer is used for detecting unmanned aerial vehicle's angle of flight, the megaphone is used for carrying out pronunciation broadcast to the river course scene to carry out scene law enforcement.
The invention has the beneficial effects that:
1) the invention uses the multi-rotor unmanned aerial vehicle to cruise the river, can realize the all-around and zero-dead-angle inspection, and greatly saves manpower and material resources.
2) The invention can check the illegal ship condition of the river channel and the illegal beach condition of the river bank, and has multiple functions and multiple purposes.
3) The unmanned aerial vehicle in the invention flies autonomously without personnel operation, thereby solving the problem that the unmanned aerial vehicle needs to be in the visual field range of an operator to develop a flight area.
4) The invention adopts the image processing technology, and can automatically identify the illegal situation without manpower.
Drawings
FIG. 1 is a flow chart of the operation of the present invention;
fig. 2 is a schematic structural diagram of the present invention.
Detailed Description
Hereinafter, preferred examples of the present invention will be described in detail with reference to the accompanying drawings. It should be understood that the preferred examples are intended to illustrate the invention, and are not intended to limit the scope of the invention.
As shown in fig. 2, the river channel cruise system adopting cooperation of multiple unmanned aerial vehicles comprises the unmanned aerial vehicles, hangars and a ground control center, wherein the hangars are used as placing platforms of the unmanned aerial vehicles, the multiple hangars are arranged at intervals along one side of a river channel, and one hangar is arranged at intervals of X to form a river channel unmanned aerial vehicle network; the ground control center communicates with the hangar and the unmanned aerial vehicle and transmits data, and sends the control signals of takeoff and landing signals, flight path information, the camera and the loudspeaker to the unmanned aerial vehicle to control takeoff and landing of the unmanned aerial vehicle and flight tracks and control opening and closing of the camera and the loudspeaker.
Unmanned aerial vehicle carries equipment such as unmanned aerial vehicle cloud platform mechanism, lithium cell, PMU power management unit, barometer, gyroscope, accelerometer, GPS orientation module, camera, megaphone to have the remote controller. The holder mechanism is used for adjusting the shooting angle of the camera; the PMU power management unit is used for measuring real-time data of the electric quantity of the battery; the barometer is used for detecting the flight height data of the unmanned aerial vehicle; data measured by the gyroscope and the accelerometer are used for analyzing the attitude of the unmanned aerial vehicle, and the attitude angle of the unmanned aerial vehicle is calculated in real time; the GPS positioning module is used for receiving real-time position data of a GPS satellite; the megaphone is used for shouting to the on-the-spot pronunciation of river course, can reach the effect of on-the-spot law enforcement.
Referring to fig. 1, the embodiment of the present invention and its specific implementation are as follows:
in the embodiment, cruising is carried out to check the illegal condition of the mudflat of a certain river channel, a required river reach of the river channel is selected as an implementation object of the embodiment, the river reach is 50km long, 10 hangars are required to be arranged according to the flight speed and the cruising ability of the unmanned aerial vehicle, and the two adjacent hangars are 5km and are respectively numbered as 1-10. One end of the river section to be detected is a starting end, and the other end of the river section to be detected is a finishing end. And 10 hangars are sequentially arranged from the starting end to the end by 5km, and the hangars are arranged at the position 20 meters away from the river surface on the right side of the river channel. Place an unmanned aerial vehicle commonly used and reserve unmanned aerial vehicle in every hangar, inside has a power and an urgent stand-by power supply can charge for unmanned aerial vehicle.
The system needs to be started once a day, and the present embodiment will describe the process at one start only.
1) The ground control center sends a takeoff signal, all hangar doors arranged beside the river channel required to be detected are opened at the moment, and the unmanned aerial vehicle inside receives the takeoff signal to take off. If the common unmanned aerial vehicle breaks down, the standby unmanned aerial vehicle takes off. At this moment, the GPS positioning module, the barometer, the gyroscope, the accelerometer and the camera arranged on the unmanned aerial vehicle are all opened, and the hangar door is closed.
2) The unmanned aerial vehicle flies to the center of the river channel through the GPS positioning module, the flying height is controlled to be kept unchanged at 20 meters through the barometer, and the plane is controlled to keep flying in a horizontal line through the gyroscope and the accelerometer. The camera deflects to the right side of the flight direction of the unmanned aerial vehicle and is aligned with the river bank beach, and the aerial photography angle is 50 degrees.
3) The unmanned aerial vehicle carries out coding work while carrying out the task of cruising. And carrying out H.264 coding on the shot video, packaging the video information and the state information together according to an RTP protocol package, and sending the video information and the state information to a ground control center through a 4G communication module by a 4G mobile cellular network. And the ground control center inputs each frame of image in the field video into the tidal flat scene image classifier trained by the convolutional neural network, and whether the tidal flat illegal condition exists is obtained through classification and judgment. The common illegal conditions of the mudflat which can be detected by specific implementation are as follows: construction violations, dumping of waste cement violations, and the like.
The mudflat scene image classifier in the embodiment is obtained by adopting the following method: the number of the selected sample images is 10 ten thousand, 2 different scenes, namely mudflat pollution and non-pollution, are totally selected, and each scene is 5 ten thousand. The training mode of the convolutional neural network is a random gradient descent method, and weight attenuation is set to be 5e-4The potential energy is 0.9, the initial learning rate is set to be 0.01, the learning rate is fixed, and when the classification precision of the model on the test set does not rise any more, the learning rate is reduced, wherein the reduction mode is that the original learning rate is multiplied by 0.1.
The final classification performance of the classifier is as follows:
accuracy of classification
Contaminated image 97％
Non-contaminated image 82％
General of 89％
In the cruise, the beach image obtained by the unmanned aerial vehicle in the No. 3 hangar at the position of 2.1km of straight line flight is different from the conventional beach image. At this moment, unmanned aerial vehicle stops the straight line flight, flies to the river side that is close to this region and does the surrounding flight to this region, and the flight radius is 20 meters, and unmanned aerial vehicle flying speed sets up to 15 per second. And further confirming the area, continuously shooting the video during the period, packaging and sending the video to the ground control center. The related images are stored in a violation evidence storage place prepared in advance by the control center and related workers are informed. The staff finds out the illegal condition of the poured cement on the mudflat of the area through later verification.
Surround and shoot the completion back, ground control center signals, and this unmanned aerial vehicle gets back to on the original straight line air course again.
4) When all the unmanned aerial vehicles fly to the position of the next hangar in a straight line, the nose rotates 180 degrees, and the original path returns. And at the moment, the device can cruise the other half of the river bank which is not shot before, and the third step is repeated.
5) When the unmanned aerial vehicle flies back to the position of the original hangar, the ground control center sends a landing signal, all hangar doors are opened, and the unmanned aerial vehicle receives the landing signal to land. And ending the river channel cruising.
The above embodiments are merely examples of the present invention, and are not intended to limit the present invention. The invention is susceptible to various modifications and alternative forms. Any modification, equivalent replacement and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A river channel cruising method based on an unmanned aerial vehicle network is characterized in that the following method is adopted to cruise the river channel
1) The ground control center controls the unmanned aerial vehicle to take off from the hangar, all hangar doors arranged beside a river section to be detected in a river channel are opened, and a GPS positioning module, a barometer, a gyroscope, an accelerometer, a camera and a loudspeaker arranged on the unmanned aerial vehicle are all opened;
2) each unmanned aerial vehicle flies to the center of a river channel where a self-body hangar is located through a GPS positioning module, then flies along the river channel, controls the flying height through a barometer, controls the aircraft to keep flying in a horizontal straight line through a gyroscope and an accelerometer, a camera faces to the right side of the flying direction of the unmanned aerial vehicle, transmits a real-time video back to a ground control center through an RTP (real-time transport protocol), and the ground control center performs image processing according to transmitted video information;
3) when each unmanned aerial vehicle flies to the center of the river channel where the next hangar is located along the river channel, the airframe rotates 180 degrees and returns along the original path, and the orientation of the camera relative to the airframe is kept unchanged;
4) after the unmanned aerial vehicle flies back to the center of the river channel where the self hangar is located, the ground control center controls the unmanned aerial vehicle to land in the hangar.
2. The river channel cruising method based on the unmanned aerial vehicle network as claimed in claim 1, wherein: when the unmanned aerial vehicle takes off, all hangars receive the hangar door opening signal of the ground control center to open the hangar door, the ground control center sends a take-off signal to the unmanned aerial vehicle, and the unmanned aerial vehicle receives the take-off signal to take off to the center of a river channel.
3. The river channel cruising method based on the unmanned aerial vehicle network as claimed in claim 1, wherein: when unmanned aerial vehicle descends, all hangars receive ground control center's hangar door opening signal and open the hangar door, and ground control center sends the descending signal to unmanned aerial vehicle, and unmanned aerial vehicle receives the descending signal and descends.
4. The river channel cruising method based on the unmanned aerial vehicle network as claimed in claim 1, wherein: the transmitted video information is subjected to image processing and judgment by adopting a tidal flat scene image classifier trained by a convolutional neural network to obtain the tidal flat illegal condition: inputting the images into a tidal flat scene image classifier trained through a convolutional neural network, classifying the images by using training characteristics in a full connection layer of the convolutional neural network, and obtaining polluted or uncontaminated tidal flat illegal conditions through classification judgment.
5. The river channel cruising method based on the unmanned aerial vehicle network as claimed in claim 4, wherein: the convolutional neural network simultaneously trains characteristics and classification in an end-to-end mode, the trained characteristics are full connection layers at the tail ends of the convolutional neural network structure, the full connection layers at the tail ends of the convolutional neural network structure are extracted and output to serve as depth characteristics of a polluted image, the obtained depth characteristics are input into a support vector machine classifier to be trained, and the mudflat scene image classifier is obtained.
6. The river channel cruising method based on the unmanned aerial vehicle network as claimed in claim 4 or 5, wherein: the convolutional neural network training process specifically includes the following steps for all input sample images:
1) firstly, carrying out size normalization to normalize the image to 32 multiplied by 32 resolution;
2) and (3) convolutional layer calculation: parametric expression of convolutional layers: 4 × 4 × 20+1 (step size), representing a convolution kernel size of 4 × 4, a number of 20, and a step size of 1, the convolution layer calculation is performed using the following formula:
x j l = f ( z j l )
z j l = Σ i ∈ M j x i l - 1 * k i j l + b j l
wherein,respectively representing the jth characteristic diagram in the current convolution layer and the ith characteristic diagram in the previous layer;a convolution kernel between the jth characteristic diagram representing the current layer and the ith characteristic diagram representing the previous layer; mjA set of feature maps representing a previous layer requiring convolution,representing the bias corresponding to the jth convolution kernel in the current convolution layer; f is an activation function, and the weight and the threshold in the convolutional layer are obtained by a random gradient descent method; i and j both represent ordinal numbers of the feature map, and l represents step size;
3) and (3) pool layer calculation: the pooling layer adopts a maximum pooling mode, 2 multiplied by 2+2 (step length) indicates that the size of a pooling core is 2 multiplied by 2, and the step length is 2; the maximum pooling calculation uses the following formula:
h c d o = m a x { h c · s + m , d · s + n o - 1 } 0 ≤ m ≤ s , 0 ≤ n ≤ s
wherein the size of the pooling nucleus is s × s, s represents the side length of the pooling nucleus,a value representing the (c, d) th coordinate position in the current pooling feature map, c, d representing the abscissa and ordinate of the pooling feature map respectively,representing the set of values in the pooling kernel in the previous layer of feature map;representing the coordinate position in the previous layer of feature map as c.s + m, d.s + n, wherein m and n represent the coordinates in the pooling core;
4) convolution layer calculation is performed again: parametric expression of convolutional layers: 3 × 3 × 40+1 (step size), which means that the size of the convolution kernel is 3 × 3, the number is 40, the step size is 1, and the specific calculation of the convolution layer in this step is the same as that in step 2);
5) performing pooling layer calculation again by adopting the same process as the step 3):
6) calculating a full connection layer: the calculation formula of the full connection layer is as follows:
R q = f ( Σ p = 1 N u p · v p q )
wherein f (-) represents an activation function, and a Rectified linear unit activation function is adopted, RqDenotes the q-th node unit, u, in the full connection layerpDenotes the p-th node element in the pooling layer, vpqRepresenting the connection weight of two nodes, N representing the number of node units of the pooling layer, and p and q being ordinal numbers of the node units of the pooling layer;
7) the calculation formula of the softmax layer is as follows:
r t = e r t , Σ t = 1 a e r t ,
r t ′ = Σ t = 1 a X t · W t , w + B t
in the formula, Wt,wRepresents the connection weight between the full connection layer and the softmax layer, BtTo be offset, XtThe t-th node element value, r, representing the full link layertRepresenting a probability value, r 'over the t-th category'tRepresenting the probability density of the t-th class; a represents the total number of categories, t is the ordinal number of the category, t is 1 or 2, a is 2, and w is the ordinal number of the connecting weight under the t-th category; r is1Representing the probability of an input mudflat illegal image being contaminated, r2Representing the probability that the input mudflat illegal image is uncontaminated;
8) the loss function formula of the convolutional neural network is as follows:
J ( θ ) = - 1 P [ Σ C = 1 P Σ D = 1 Q 1 { g ( P ) = Q } l o g g Q ( P ) ]
wherein 1{ } is an indicative function, and θ represents a network parameter;the probability value of the P-th sample on the class Q is shown, P represents the number of images in one iteration batch, Q represents the number of image classes, and C and D are ordinal numbers of the samples.
7. The utility model provides a river course system of cruising based on unmanned aerial vehicle network which characterized in that: the system mainly comprises an unmanned aerial vehicle, a hangar and a ground control center, wherein the hangar is used as a placing platform of the unmanned aerial vehicle, a plurality of hangars are arranged at intervals along one side of a river channel, the unmanned aerial vehicle is loaded with a GPS positioning module, a barometer, a gyroscope, an accelerometer, a cradle head mechanism and a wireless two-way communication module, and a camera and a loudspeaker are arranged on the cradle head mechanism; the ground control center is communicated with the hangar and the unmanned aerial vehicle and transmits data, and the ground control center sends a take-off and landing signal, flight line information, a camera and a control signal of a loudspeaker to the unmanned aerial vehicle.
8. The river channel cruise system based on the unmanned aerial vehicle network as claimed in claim 7, wherein:
the hangar is provided with a common unmanned aerial vehicle and a standby unmanned aerial vehicle as well as a common power supply and a standby power supply which can charge the unmanned aerial vehicle, and the hangar receives a control signal sent by a ground control center in the take-off and landing processes of the unmanned aerial vehicle to open and close a hangar door.
9. The river channel cruise system based on the unmanned aerial vehicle network as claimed in claim 7, wherein: the shape of the hangar is cuboid, and the unmanned aerial vehicle machine heads are parallelly arranged in the hangar side by side.
10. The river channel cruise system based on the unmanned aerial vehicle network as claimed in claim 7, wherein: the ground control center and the unmanned aerial vehicle between carry out communication and data transmission, obtain the photo and the video of shooing from the unmanned aerial vehicle to erect the database that is equipped with storage unmanned aerial vehicle data collection and image, the communication adopts 4G honeycomb mobile network to communicate, data transmission uses the TCP agreement, video and audio data use the RTP agreement to transmit.
CN201610333740.6A 2016-05-18 2016-05-18 River cruise system and cruise method based on unmanned plane network Active CN105929845B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610333740.6A CN105929845B (en) 2016-05-18 2016-05-18 River cruise system and cruise method based on unmanned plane network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610333740.6A CN105929845B (en) 2016-05-18 2016-05-18 River cruise system and cruise method based on unmanned plane network

Publications (2)

Publication Number Publication Date
CN105929845A true CN105929845A (en) 2016-09-07
CN105929845B CN105929845B (en) 2018-10-19

Family

ID=56841687

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610333740.6A Active CN105929845B (en) 2016-05-18 2016-05-18 River cruise system and cruise method based on unmanned plane network

Country Status (1)

Country Link
CN (1) CN105929845B (en)

Cited By (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106598921A (en) * 2016-12-12 2017-04-26 清华大学 Method and device for converting to ancient poem from modern article based on long short term memory (LSTM) model
CN106682592A (en) * 2016-12-08 2017-05-17 北京泛化智能科技有限公司 Automatic image recognition system and method based on neural network method
CN106708090A (en) * 2016-12-23 2017-05-24 四川九洲电器集团有限责任公司 Unmanned aerial vehicle (UAV) cluster system
CN107085852A (en) * 2017-04-01 2017-08-22 南昌大学 A kind of river course surface flow field method of testing based on unmanned plane
CN107194396A (en) * 2017-05-08 2017-09-22 武汉大学 Method for early warning is recognized based on the specific architecture against regulations in land resources video monitoring system
CN107239803A (en) * 2017-07-21 2017-10-10 国家海洋局第海洋研究所 Utilize the sediment automatic classification method of deep learning neutral net
CN107657623A (en) * 2017-08-28 2018-02-02 北京工业大学 A kind of river course line detecting system and method for unmanned plane
CN107808425A (en) * 2017-11-28 2018-03-16 刘松林 Oil-gas pipeline cruising inspection system and its method for inspecting based on unmanned plane image
CN108256447A (en) * 2017-12-29 2018-07-06 广州海昇计算机科技有限公司 A kind of unmanned plane video analysis method based on deep neural network
CN108280395A (en) * 2017-12-22 2018-07-13 中国电子科技集团公司第三十研究所 A kind of efficient identification method flying control signal to low small slow unmanned plane
CN108376460A (en) * 2018-04-04 2018-08-07 武汉理工大学 System and method is monitored based on unmanned plane and the oil pollution at sea of BP neural network
CN109630905A (en) * 2019-01-25 2019-04-16 电子科技大学 A kind of full intelligent inspection system of oil-gas pipeline based on unmanned aerial vehicle remote sensing and deep learning
CN109885091A (en) * 2019-03-21 2019-06-14 华北电力大学（保定） A kind of unmanned plane autonomous flight control method and system
CN109934110A (en) * 2019-02-02 2019-06-25 广州中科云图智能科技有限公司 A kind of river squatter building house recognition methods nearby
CN110853174A (en) * 2019-10-30 2020-02-28 中设设计集团股份有限公司 Inland river patrol and comprehensive law enforcement method, device and system based on ship-shore cooperation
CN111096736A (en) * 2018-10-26 2020-05-05 深圳市理邦精密仪器股份有限公司 Electrocardiogram classification method, device and system based on active learning
CN111539362A (en) * 2020-04-28 2020-08-14 西北工业大学 Unmanned aerial vehicle image target detection device and method
CN112102369A (en) * 2020-09-11 2020-12-18 陕西欧卡电子智能科技有限公司 Autonomous inspection method, device and equipment for water surface floating target and storage medium
CN113344885A (en) * 2021-06-15 2021-09-03 温州大学 River floating object detection method based on cascade convolution neural network
CN114220044A (en) * 2021-11-23 2022-03-22 慧之安信息技术股份有限公司 River course floater detection method based on AI algorithm
CN114229024A (en) * 2021-12-08 2022-03-25 海宁量益智能装备有限公司 Multilayer unmanned aerial vehicle machine storehouse that charges

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100962615B1 (en) * 2008-01-17 2010-06-10 대한민국(관리부서:국립수산과학원) Observation system of measurement the sea circumstances and aerial vehicle with unmanned and methods thereof
CN103235600A (en) * 2013-04-10 2013-08-07 东南大学 Electric traffic observation device for fixed-wing model airplanes
CN105389988A (en) * 2015-12-07 2016-03-09 北京航空航天大学 Multi-unmanned aerial vehicle cooperation highway intelligent inspection system
CN105446351A (en) * 2015-11-16 2016-03-30 杭州码全信息科技有限公司 Robotic airship system capable of locking target area for observation based on autonomous navigation
CN105512666A (en) * 2015-12-16 2016-04-20 天津天地伟业数码科技有限公司 River garbage identification method based on videos

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100962615B1 (en) * 2008-01-17 2010-06-10 대한민국(관리부서:국립수산과학원) Observation system of measurement the sea circumstances and aerial vehicle with unmanned and methods thereof
CN103235600A (en) * 2013-04-10 2013-08-07 东南大学 Electric traffic observation device for fixed-wing model airplanes
CN105446351A (en) * 2015-11-16 2016-03-30 杭州码全信息科技有限公司 Robotic airship system capable of locking target area for observation based on autonomous navigation
CN105389988A (en) * 2015-12-07 2016-03-09 北京航空航天大学 Multi-unmanned aerial vehicle cooperation highway intelligent inspection system
CN105512666A (en) * 2015-12-16 2016-04-20 天津天地伟业数码科技有限公司 River garbage identification method based on videos

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
傅天驹 等: "复杂背景下基于深度卷积神经网络的森林火灾识别", 《复杂背景下基于深度卷积神经网络的森林火灾识别》 *

Cited By (26)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106682592A (en) * 2016-12-08 2017-05-17 北京泛化智能科技有限公司 Automatic image recognition system and method based on neural network method
CN106682592B (en) * 2016-12-08 2023-10-27 北京泛化智能科技有限公司 Image automatic identification system and method based on neural network method
CN106598921A (en) * 2016-12-12 2017-04-26 清华大学 Method and device for converting to ancient poem from modern article based on long short term memory (LSTM) model
CN106708090A (en) * 2016-12-23 2017-05-24 四川九洲电器集团有限责任公司 Unmanned aerial vehicle (UAV) cluster system
CN107085852A (en) * 2017-04-01 2017-08-22 南昌大学 A kind of river course surface flow field method of testing based on unmanned plane
CN107085852B (en) * 2017-04-01 2019-12-20 南昌大学 River channel surface flow field testing method based on unmanned aerial vehicle aerial photography
CN107194396A (en) * 2017-05-08 2017-09-22 武汉大学 Method for early warning is recognized based on the specific architecture against regulations in land resources video monitoring system
CN107239803A (en) * 2017-07-21 2017-10-10 国家海洋局第海洋研究所 Utilize the sediment automatic classification method of deep learning neutral net
CN107657623A (en) * 2017-08-28 2018-02-02 北京工业大学 A kind of river course line detecting system and method for unmanned plane
CN107808425A (en) * 2017-11-28 2018-03-16 刘松林 Oil-gas pipeline cruising inspection system and its method for inspecting based on unmanned plane image
CN108280395A (en) * 2017-12-22 2018-07-13 中国电子科技集团公司第三十研究所 A kind of efficient identification method flying control signal to low small slow unmanned plane
CN108280395B (en) * 2017-12-22 2021-12-17 中国电子科技集团公司第三十研究所 Efficient identification method for flight control signals of low-small-slow unmanned aerial vehicle
CN108256447A (en) * 2017-12-29 2018-07-06 广州海昇计算机科技有限公司 A kind of unmanned plane video analysis method based on deep neural network
CN108376460A (en) * 2018-04-04 2018-08-07 武汉理工大学 System and method is monitored based on unmanned plane and the oil pollution at sea of BP neural network
CN111096736A (en) * 2018-10-26 2020-05-05 深圳市理邦精密仪器股份有限公司 Electrocardiogram classification method, device and system based on active learning
CN109630905A (en) * 2019-01-25 2019-04-16 电子科技大学 A kind of full intelligent inspection system of oil-gas pipeline based on unmanned aerial vehicle remote sensing and deep learning
CN109934110A (en) * 2019-02-02 2019-06-25 广州中科云图智能科技有限公司 A kind of river squatter building house recognition methods nearby
CN109934110B (en) * 2019-02-02 2021-01-12 广州中科云图智能科技有限公司 Method for identifying illegal buildings near river channel
CN109885091A (en) * 2019-03-21 2019-06-14 华北电力大学（保定） A kind of unmanned plane autonomous flight control method and system
CN110853174A (en) * 2019-10-30 2020-02-28 中设设计集团股份有限公司 Inland river patrol and comprehensive law enforcement method, device and system based on ship-shore cooperation
CN111539362A (en) * 2020-04-28 2020-08-14 西北工业大学 Unmanned aerial vehicle image target detection device and method
CN112102369A (en) * 2020-09-11 2020-12-18 陕西欧卡电子智能科技有限公司 Autonomous inspection method, device and equipment for water surface floating target and storage medium
CN112102369B (en) * 2020-09-11 2024-04-09 陕西欧卡电子智能科技有限公司 Autonomous inspection method, device, equipment and storage medium for water surface floating target
CN113344885A (en) * 2021-06-15 2021-09-03 温州大学 River floating object detection method based on cascade convolution neural network
CN114220044A (en) * 2021-11-23 2022-03-22 慧之安信息技术股份有限公司 River course floater detection method based on AI algorithm
CN114229024A (en) * 2021-12-08 2022-03-25 海宁量益智能装备有限公司 Multilayer unmanned aerial vehicle machine storehouse that charges

Also Published As

Publication number Publication date
CN105929845B (en) 2018-10-19

Similar Documents

Publication Publication Date Title
CN105929845B (en) River cruise system and cruise method based on unmanned plane network
CN105913604B (en) Assay method and its device occur for the fire based on unmanned plane
CN112422783B (en) Unmanned aerial vehicle intelligent patrol system based on parking apron cluster
CN102654940B (en) Processing method of traffic information acquisition system based on unmanned aerial vehicle and
CN205750548U (en) A kind of inspection system based on unmanned plane
CN105389988B (en) A kind of express highway intelligent cruising inspection system of multiple no-manned plane collaboration
CN106054928B (en) A kind of full region fire generation measuring method based on unmanned plane network
CN105511495A (en) Control method and system for intelligent unmanned aerial vehicle patrol for power line
CN202050188U (en) Unmanned aerial vehicle
CN203219298U (en) Unmanned helicopter system special for inspecting electric grid in mountain area
JP7065477B2 (en) Disaster situation judgment system and disaster judgment flight system
CN110189411A (en) Emergency management and rescue Search Area method for visualizing after a kind of accident of aircraft
CN113077561A (en) Intelligent inspection system for unmanned aerial vehicle
CN112764427B (en) Relay unmanned aerial vehicle inspection system
CN103576691B (en) 3G police service unmanned plane managing and control system
CN111413999A (en) Safety inspection method based on unmanned aerial vehicle
CN112487894A (en) Automatic inspection method and device for rail transit protection area based on artificial intelligence
CN117558158A (en) Static traffic normalization inspection system and method based on unmanned airport networking
CN108845584B (en) Wind-resistant unmanned aerial vehicle tracking burst gas pollution source method based on LS-SVM control
CN103198666B (en) Highway traffic stream space average speed observation method based on fixed wing model airplane
Jin et al. Unmanned aerial vehicle (uav) based traffic monitoring and management
CN106998451A (en) The area condition panorama guide system and its method monitored based on unmanned vehicle
CN216647401U (en) Safety helmet recognition device
CN114655361B (en) Intelligent buoy and offshore cruise unmanned aerial vehicle cluster control management system and method
CN109559565A (en) Immediately the hovercar air control platform of online flight examination & approval is realized

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20160907

Assignee: Xinchang China Metrology University Enterprise Innovation Research Institute Co.,Ltd.

Assignor: China Jiliang University

Contract record no.: X2021330000071

Denomination of invention: River cruise system and method based on UAV network

Granted publication date: 20181019

License type: Common License

Record date: 20210816

EC01 Cancellation of recordation of patent licensing contract
EC01 Cancellation of recordation of patent licensing contract

Assignee: Xinchang China Metrology University Enterprise Innovation Research Institute Co.,Ltd.

Assignor: China Jiliang University

Contract record no.: X2021330000071

Date of cancellation: 20211231