CN110007666A - A kind of intelligent patrol detection fire early-warning system and its working method based on fish-eye camera trolley - Google Patents
A kind of intelligent patrol detection fire early-warning system and its working method based on fish-eye camera trolley Download PDFInfo
- Publication number
- CN110007666A CN110007666A CN201910303165.9A CN201910303165A CN110007666A CN 110007666 A CN110007666 A CN 110007666A CN 201910303165 A CN201910303165 A CN 201910303165A CN 110007666 A CN110007666 A CN 110007666A
- Authority
- CN
- China
- Prior art keywords
- speed
- trolley
- fish
- fire alarm
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 30
- 238000000034 method Methods 0.000 title claims abstract description 21
- 235000011034 Rubus glaucus Nutrition 0.000 claims abstract description 68
- 235000009122 Rubus idaeus Nutrition 0.000 claims abstract description 68
- 240000007651 Rubus glaucus Species 0.000 claims abstract description 65
- 230000004888 barrier function Effects 0.000 claims abstract description 14
- 230000004913 activation Effects 0.000 claims description 22
- 238000012545 processing Methods 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 10
- 238000010801 machine learning Methods 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 7
- 244000235659 Rubus idaeus Species 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 abstract description 3
- 230000007547 defect Effects 0.000 abstract description 2
- 239000000523 sample Substances 0.000 abstract description 2
- 210000002569 neuron Anatomy 0.000 description 8
- 238000010408 sweeping Methods 0.000 description 6
- 230000007423 decrease Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000006073 displacement reaction Methods 0.000 description 3
- 239000000779 smoke Substances 0.000 description 3
- 238000004804 winding Methods 0.000 description 3
- 229910000838 Al alloy Inorganic materials 0.000 description 2
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 229910052744 lithium Inorganic materials 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000004297 night vision Effects 0.000 description 2
- 238000012356 Product development Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 235000021028 berry Nutrition 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000008921 facial expression Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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
- G05D1/0088—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots characterized by the autonomous decision making process, e.g. artificial intelligence, predefined behaviours
-
- 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
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
-
- 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
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
- G05D1/0242—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using non-visible light signals, e.g. IR or UV signals
-
- 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
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
- G05D1/0246—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B17/00—Fire alarms; Alarms responsive to explosion
- G08B17/12—Actuation by presence of radiation or particles, e.g. of infrared radiation or of ions
- G08B17/125—Actuation by presence of radiation or particles, e.g. of infrared radiation or of ions by using a video camera to detect fire or smoke
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Electromagnetism (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Emergency Management (AREA)
- Medical Informatics (AREA)
- Game Theory and Decision Science (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
- Fire-Detection Mechanisms (AREA)
- Alarm Systems (AREA)
Abstract
The present invention relates to a kind of intelligent patrol detection fire early-warning systems and its working method based on fish-eye camera trolley, including image capture module, raspberry pie, server, control motor drive module；The image of acquisition is sent to the raspberry pie for acquiring image by image capture module；Raspberry pie is for realizing fire alarm or obtains the speed and angle information of trolley, and the case where fire alarm is sent to the server；Control motor drive module is used to receive the speed and angle information of trolley, realizes the intelligent barrier avoiding of trolley；Server is for the case where receiving fire alarm and is handled in time.The defect for compensating for common Conventional sensor probe technology, improves the accuracy rate of fire alarm.
Description
Technical field
The present invention relates to a kind of intelligent patrol detection fire early-warning systems and its working method based on fish-eye camera trolley, belong to
In artificial intelligence field of image detection and intelligent carriage technical field.
Background technique
With the industrialized development in China, the warehouse for all having built oneself of a large amount of factories, for placing industrial production material
Material, fire early-warning system are widely used in industrial indoor warehouse.
Currently, inside fire early warning technology is mainly sensor Detection Techniques, it is divided into temperature sensing, smoke detection, gas
Detection.Such device is often placed at the top of indoor room, detects temperature, smog or the gas when fire occurs by various kinds of sensors
Body is converted into the acceptable electric signal of computer after signal analysis, judgement, issues alarm of fire, and remind administrative staff
Make corresponding measure.It, the problem of traditional sensors are due to detection range, cannot however, being gradually increased with warehouse scale
Discovery fire condition in time, such as smoke alarm, in the case where warehouse is bigger, smoke propagation to sensor can be detected
Place, the intensity of a fire may be very big, be easy to cause weight huge economic loss and casualties.In addition, traditional sensors are visited
Examining system needs to arrange big quantity sensor and winding displacement, and cost is more expensive.
Chinese patent literature CN 208144974U discloses a kind of intelligent sweeping robot avoidance system based on Kinect
System, comprising: Kinect camera module, PC platform, raspberry pie, sweeping robot travelling control module, 802.11n wireless communication
Module.The sweeping robot obstacle avoidance system based on Kinect and raspberry pie includes the following course of work: in sweeping robot
Before operation and in operational process, the 3-D image for having depth of view information is obtained using Kinect camera module, and utilize PC
Platform models the 3-D image obtained from Kinect camera module, is carried out by gridding algorithm to indoor environment
Modeling, to control sweeping robot direction of travel using sweeping robot travelling control module by raspberry pie.But it should
Patent needs to carry out three-dimensional modeling using PC platform to need to model overlong time, when expending a large amount of if warehouse is bigger
Between, intermediate conveyor the problem of there is also loss of data, modeling inaccuracy is caused, avoidance accuracy rate is influenced.
Summary of the invention
In view of the deficiencies of the prior art, in order to make up the defect of common Conventional sensor probe technology, fire alarm is improved
Accuracy rate, the present invention provides a kind of intelligent patrol detection fire early-warning systems based on fish-eye camera trolley；
The present invention also provides the working methods of the above-mentioned intelligent patrol detection fire early-warning system based on fish-eye camera trolley.
Term is explained:
1, raspberry pie, Raspberry Pi are abbreviated as RPi or RasPi/RPI, are to program to educate for learning computer
And design, only the microcomputer of credit card-sized, system are based on Linux.
2, deep learning, it is a kind of based on the method for carrying out representative learning to data in machine learning.Observation (such as a width
Image) various ways can be used to indicate, such as vector of each pixel intensity value, or be more abstractively expressed as a series of
Side, region of specific shape etc..And use certain specific representation methods be easier from example learning tasks (recognition of face or
Human facial expression recognition).The benefit of deep learning is that the feature learning and layered characteristic with non-supervisory formula or Semi-supervised extract height
Effect algorithm obtains feature to substitute by hand.
3, tensorflow frame, TensorFlow are an open-source software libraries, are widely used in all kinds of machines
The programming that device learns (machine learning) algorithm realizes that predecessor is the neural network algorithm library of Google
DistBelief.Tensorflow possesses multi-level structure, can be deployed in all kinds of servers, PC terminal and webpage and support GPU
With TPU high performance numerical computing, the scientific research of the product development being widely used in inside Google and each field, be at present most
Welcome open source machine learning frame.
The technical solution of the present invention is as follows:
A kind of intelligent patrol detection fire early-warning system based on fish-eye camera trolley, including image capture module, raspberry pie,
Server, control motor drive module；
The image of acquisition is sent to the raspberry pie for acquiring image by described image acquisition module；The raspberry
The speed and angle information of trolley are sent for realizing fire alarm or obtained, and the case where fire alarm is sent to the service
Device；Control motor drive module is used to receive the speed and angle information of trolley, realizes the intelligent barrier avoiding of trolley；Server is used for
The case where receiving fire alarm is simultaneously handled in time.
Preferred according to the present invention, described image acquisition module includes fish-eye camera, infrared camera, the raspberry pie
Including fire alarm module, automatic obstacle-avoiding module, the fish-eye camera connects the fire alarm module, the infrared photography
Head, the automatic obstacle-avoiding module, the control motor drive module are sequentially connected；
The control motor drive module includes 4WD expansion board, four DC speed-reducings and tire；4WD expansion board is logical
It crosses winding displacement and connects the raspberry pie, 4WD expansion board is separately connected four DC speed-reducings by four motor interfaces, and four straight
Stream decelerating motor is separately connected four tires, and 4WD expansion board is used to receive the speed for the trolley that the raspberry pie transmits, and passes through
Motor interface controls the revolving speed of four DC speed-reducings respectively；
The fish-eye camera is used to acquire the image of scene to early warning fire, and described in the image of acquisition is sent to
Fire alarm module；
The fire alarm module refers to: defeated for being detected by machine learning tensorflow frame to image
Enter image to the fire alarm model, by operation, the fire alarm model exports occurrence index, and occurrence index is fire
The value of predetermined output represents fire if occurrence index is greater than given threshold when Early-warning Model training, and fire is pre-
Alert information passes to backstage manager, otherwise, without processing；
The frame road image that the infrared camera is 640*480 for acquisition resolution, and by the mileage chart of acquisition
As being sent to the automatic obstacle-avoiding module；
The automatic obstacle-avoiding module carries out image using automatic obstacle-avoiding model by machine learning tensorflow frame
Operation predicts the angle and speed of trolley advance, angle and speed preset output, tree when being the training of automatic obstacle-avoiding module
The certain kind of berries is sent according to angle and speed, the revolving speed of four DC speed-reducings is calculated by formula, raspberry pie slows down four direct currents
The revolving speed of motor passes to 4WD expansion board, the revolving speed of 4WD expansion board control four motor interfaces of connection, control by connecting line
The angle and speed that trolley advances, to realize intelligent barrier avoiding.
Control motor drive module receives the revolving speed for four DC speed-reducings that automatic obstacle avoidance module transmits, by expansion board
The revolving speed for controlling four DC speed-reducings controls small vehicle speed and angle, realizes automatic obstacle-avoiding.18650 lithium battery of Selection of Battery
Group may be repeated charging, charger 12.6V；Aluminum alloy chassis is selected on chassis；Expansion board selects 4WD expansion board, is used to
Control motor.
It is further preferred that the calculation formula of speed and revolving speed are as follows: V=v2 π R, V refer to speed, and v refers to that revolving speed, R refer to
The radius of gyration.
Preferred according to the present invention, the fire alarm model includes 5 layers of convolutional layer, 2 layers of pond layer, 3 layers of full articulamentum,
Activation primitive is ELU.
Gradient decline uses cross entropy, exports as fire alarm index, uses tensorflow in 1 1080ti server
Frame is trained, and data set is that the resolution ratio of 10000 tape labels is 640*480 picture, and training fit time is 6 hours.
Preferred according to the present invention, the automatic obstacle-avoiding model is End-to-End nvidia-cnn, including 5 layers of convolution
Layer, 3 layers of full articulamentum, activation primitive ELU.
Gradient decline uses cross entropy, exports the angle and speed advanced for trolley, uses in 1 1080ti server
Tensorflow frame is trained, and data set is the picture that 30000 tape label pixels are 640*480, training fit time
It is 16 hours.
Preferred according to the present invention, the fish-eye camera connects the fire alarm module by USB connecting line；It is described
Infrared camera connects the automatic obstacle-avoiding module by USB connecting line.
Preferred according to the present invention, the fire alarm module and the automatic obstacle-avoiding module include model 3B+
The raspberry pie plate of (Raspberry Pi3B+), SD card, SD card are inserted in raspberry pie.
It selects raspberry pie 3B+ (Raspberry Pi 3B+), is equipped with 64 four core Cortex-A53 processors of 1.4GHz,
1GB RAM, full-scale HDMI and 4 standard USB port, the gigabit Ethernet on USB2, CSI camera attaching device and DSI
It shows connector, a 40 needle GPIO connectors, carries double frequency 802.11ac wireless network card and bluetooth 4.2, be that a system is based on
The microcomputer of Linux.Raspbian operating system is carried, there is extremely strong scalability, while the offer of raspberry pie official is non-
Normal SDK packet abundant, system environments are python 3.5+tensorflow 1.7+opencv3.SD card selects 16G Jin Shidun,
It is used to store automatic obstacle-avoiding module and fire alarm module.
Preferred according to the present invention, the fish-eye camera is 5,000,000 pixel USB fish-eye cameras of riel prestige view；Extensively
Angle is 150 degree, and focusing is supported in infrared night vision；The infrared camera is the infrared camera of micro- snow.Included infrared light compensating lamp,
It supports focusing, 5,000,000 pixel pictures can be shot and records the video of 1080p.
Preferred according to the present invention, the control motor drive module includes expansion board, motor and tire, and control motor drives
The revolving speed for four DC speed-reducings that the automatic obstacle avoidance module that dynamic model block receives transmits controls four direct currents deceleration electricity by expansion board
The revolving speed of machine realizes automatic obstacle-avoiding.
The working method of the above-mentioned intelligent patrol detection fire early-warning system based on fish-eye camera trolley, comprises the following steps that
(1) described image acquisition module acquires image, and the image of acquisition is sent to the raspberry pie；
(2) raspberry pie realizes fire alarm or obtains the speed and angle information of trolley, and by the feelings of fire alarm
Condition is sent to the server, and the speed of trolley and angle letter are sent to the server；
(3) the control motor drive module receives the speed and angle information of trolley, realizes the intelligent barrier avoiding of trolley, institute
Server is stated for the case where receiving fire alarm and is handled in time.
Preferred according to the present invention, in the step (2), raspberry pie realizes fire alarm, comprises the following steps that
A, raspberry pie obtains the warehouse image of resolution ratio 640*480, warehouse chart with the speed of 1 frame per second from fish-eye camera
As being transmitted by USB connecting line；
B, raspberry pie receives the warehouse image that fish-eye camera transmits, and image input fire alarm model in warehouse is carried out pre-
It surveys, calculation processing is carried out to warehouse image using fire alarm model by tensorflow frame, obtains Warehouse Fire index；
C, raspberry pie judges result, judges whether storehouse occurrence index is more than the given threshold of setting, if being more than to set
Determine threshold value, then represent have fire, pass the result to server, if be no more than given threshold, abandon as a result, continue into
The prediction of row next frame；
D, server receives fire alarm, carries out fire extinguishing processing by storekeeper personnel.
It is preferred according to the present invention, it realizes the intelligent barrier avoiding of trolley, comprises the following steps that
E, raspberry pie obtains the road image of resolution ratio 640*480, road with the speed of 20 frame per second from infrared camera
Image is transmitted by USB connecting line；
F, raspberry pie receives the road image that infrared camera transmits, and road image input automatic obstacle-avoiding model is carried out pre-
Survey, by tensorflow frame using automatic obstacle-avoiding model to picture carry out calculation processing, obtain trolley advance angle with
Speed；
G, the respective revolving speed of each DC speed-reducing, raspberry are calculated by formula according to the angle and speed predicted
Group passes the result to control motor drive module by connecting line；
H, control obstacle avoidance module receives four respective revolving speeds of DC speed-reducing, four motor interface difference of expansion board
Four DC speed-reducing revolving speeds are controlled, the angle and speed that control trolley advances achieve the effect that automatic obstacle-avoiding.
The invention has the benefit that
1, a kind of intelligent patrol detection fire early-warning system and its working method based on fish-eye camera trolley of the present invention, automatically
Avoidance obtains road image information by infrared camera, gives automatic obstacle-avoiding model and carries out calculation processing, and according to prediction result,
Angle and speed that trolley advances are controlled, achievees the effect that automatic obstacle-avoiding, allows trolley is autonomous to move in warehouse；Fire
Early warning obtains warehouse image information by fish-eye camera, gives fire alarm model and carries out calculation processing, judges whether there is fire,
Achieve the effect that fire alarm.
2, a kind of intelligent patrol detection fire early-warning system and its working method based on fish-eye camera trolley of the present invention, can be with
The accuracy and speed for greatly improving fire alarm can find fire feelings compared to traditional sensors alarm mode faster
Condition informs storehouse management administrative staff, avoids great property loss.
3, a kind of intelligent patrol detection fire early-warning system and its working method based on fish-eye camera trolley of the present invention, has both
Practicability and economic benefit can upgrade according to later period program, extend function.
Detailed description of the invention
Fig. 1 is that the present invention is based on the structural block diagrams of the intelligent patrol detection fire early-warning system of fish-eye camera trolley；
Fig. 2 is the flow diagram that the present invention realizes automatic obstacle-avoiding；
Fig. 3 is the flow diagram that the present invention realizes fire alarm；
Fig. 4 is 4WD expansion board circuit connection diagram of the present invention；
Specific embodiment
The present invention is further qualified with embodiment with reference to the accompanying drawings of the specification, but not limited to this.
Embodiment 1
A kind of intelligent patrol detection fire early-warning system based on fish-eye camera trolley, as shown in Figure 1, including Image Acquisition mould
Block, raspberry pie, server, control motor drive module；
The image of acquisition is sent to raspberry pie for acquiring image by image capture module；Raspberry pie is for realizing fire
Calamity early warning or the speed and angle information for obtaining trolley, and the case where fire alarm, is sent to server；Control motor driven
Module is used to receive the speed and angle information of trolley, realizes the intelligent barrier avoiding of trolley；Server is for receiving fire alarm
Situation is simultaneously handled in time.
Embodiment 2
According to a kind of intelligent patrol detection fire early-warning system based on fish-eye camera trolley described in embodiment 1, difference
It is:
Image capture module includes fish-eye camera, infrared camera, and raspberry pie includes fire alarm module, automatic obstacle-avoiding
Module, fish-eye camera connect fire alarm module, and infrared camera, automatic obstacle-avoiding module, control motor drive module are successively
Connection；
Controlling motor drive module includes 4WD expansion board, four DC speed-reducings and tire；4WD expansion board such as Fig. 4 institute
Show, 4WD expansion board connects raspberry pie by winding displacement, and 4WD expansion board is separately connected four direct currents by four motor interfaces and slows down
Motor, four DC speed-reducings are separately connected four tires, and 4WD expansion board is used to receive the trolley that the raspberry pie transmits
Speed, and pass through the revolving speed that motor interface controls four DC speed-reducings respectively；
Fish-eye camera is used to acquire the image of the scene to early warning fire, and the image of acquisition is sent to fire alarm
Module；
Fire alarm module refers to: input figure for being detected by machine learning tensorflow frame to image
As arriving fire alarm model, by operation, fire alarm model exports occurrence index, and occurrence index is fire alarm model training
When predetermined output value represent fire if occurrence index is greater than given threshold, fire alarm information passed to
Backstage manager, otherwise, without processing；
The frame road image that infrared camera is 640*480 for acquisition resolution, and the road image of acquisition is sent out
It send to automatic obstacle-avoiding module；
Automatic obstacle-avoiding module carries out operation to image using automatic obstacle-avoiding model by machine learning tensorflow frame,
Predict the angle and speed of trolley advance, angle and speed preset output, raspberry pie when being the training of automatic obstacle-avoiding module
According to angle and speed, the revolving speed of four DC speed-reducings is calculated by formula, raspberry pie is by four DC speed-reducings
Revolving speed 4WD expansion board is passed to by connecting line, the revolving speed of 4WD expansion board control four motor interfaces of connection controls trolley
The angle and speed of advance, to realize intelligent barrier avoiding.
Control motor drive module receives the revolving speed for four DC speed-reducings that automatic obstacle avoidance module transmits, by expansion board
The revolving speed for controlling four DC speed-reducings controls small vehicle speed and angle, realizes automatic obstacle-avoiding.18650 lithium battery of Selection of Battery
Group may be repeated charging, charger 12.6V；Aluminum alloy chassis is selected on chassis；Expansion board selects 4WD expansion board, is used to
Control motor.
The calculation formula of speed and revolving speed are as follows: V=v2 π R, V refer to speed, and v refers to that revolving speed, R refer to the radius of gyration.
Fire alarm model includes 5 layers of convolutional layer, 2 layers of pond layer, 3 layers of full articulamentum, activation primitive ELU.
Image normalization
Convolution:5x5,filter:128,strides:2x2,activation:ELU
Convolution:5x5,filter:64,strides:2x2,activation:ELU
Convolution:5x5,filter:64,strides:2x2,activation:ELU
Convolution:3x3,filter:64,strides:1x1,activation:ELU
Convolution:3x3,filter:64,strides:1x1,activation:ELU
Polling:3x3,strides:2x2,activation:ELU
Polling:2x2,strides:1x1,activation:ELU
Fully connected:neurons:100,activation:ELU
Fully connected:neurons:50,activation:ELU
Fully connected:neurons:10,activation:ELU
Fully connected:neurons:1(output)
Gradient decline uses cross entropy, exports as fire alarm index, uses tensorflow in 1 1080ti server
Frame is trained, and data set is that the resolution ratio of 10000 tape labels is 640*480 picture, and training fit time is 6 hours.
Automatic obstacle-avoiding model is End-to-End nvidia-cnn, including 5 layers of convolutional layer, 3 layers of full articulamentum, activates letter
Number is ELU.
Convolution:5x5,filter:128,strides:2x2,activation:ELU
Convolution:5x5,filter:64,strides:2x2,activation:ELU
Convolution:5x5,filter:64,strides:2x2,activation:ELU
Convolution:3x3,filter:64,strides:1x1,activation:ELU
Convolution:3x3,filter:32,strides:1x1,activation:ELU
Drop out(0.5)
Fully connected:neurons:100,activation:ELU
Fully connected:neurons:50,activation:ELU
Fully connected:neurons:10,activation:ELU
Fully connected:neurons:2(output)
Gradient decline uses cross entropy, exports the angle and speed advanced for trolley, uses in 1 1080ti server
Tensorflow frame is trained, and data set is the picture that 30000 tape label pixels are 640*480, training fit time
It is 16 hours.
Fish-eye camera connects the fire alarm module by USB connecting line；Infrared camera is connected by USB connecting line
Connect the automatic obstacle-avoiding module.
Fire alarm module and automatic obstacle-avoiding module include the raspberry pie of model 3B+ (Raspberry Pi 3B+)
Plate, SD card, SD card are inserted in raspberry pie.
It selects raspberry pie 3B+ (Raspberry Pi 3B+), is equipped with 64 four core Cortex-A53 processors of 1.4GHz,
1GB RAM, full-scale HDMI and 4 standard USB port, the gigabit Ethernet on USB2, CSI camera attaching device and DSI
It shows connector, a 40 needle GPIO connectors, carries double frequency 802.11ac wireless network card and bluetooth 4.2, be that a system is based on
The microcomputer of Linux.Raspbian operating system is carried, there is extremely strong scalability, while the offer of raspberry pie official is non-
Normal SDK packet abundant, system environments are python 3.5+tensorflow 1.7+opencv3.SD card selects 16G Jin Shidun,
It is used to store automatic obstacle-avoiding module and fire alarm module.
Fish-eye camera is 5,000,000 pixel USB fish-eye cameras of riel prestige view；Wide-angle is 150 degree, infrared night vision, branch
Hold focusing；The infrared camera is the infrared camera of micro- snow.Included infrared light compensating lamp, supports focusing, can shoot 5,000,000
Pixel picture and the video for recording 1080p.
Control motor drive module includes expansion board, motor and tire, and control motor drive module receives automatic obstacle-avoiding mould
The revolving speed for four DC speed-reducings that block transmits is kept away automatically by the revolving speed realization that expansion board controls four DC speed-reducings
Barrier.
Embodiment 3
The working method of the above-mentioned intelligent patrol detection fire early-warning system based on fish-eye camera trolley, comprises the following steps that
(1) image capture module acquires image, and the image of acquisition is sent to raspberry pie；
(2) raspberry pie realizes fire alarm or obtains the speed and angle information of trolley, and the case where fire alarm is sent out
It send to server, the speed of trolley and angle letter is sent to the server；
(3) control motor drive module receives the speed and angle information of trolley, realizes the intelligent barrier avoiding of trolley, server
The case where for receiving fire alarm, is simultaneously handled in time.
In step (2), raspberry pie realizes fire alarm, as shown in figure 3, comprising the following steps that
A, raspberry pie obtains the warehouse image of resolution ratio 640*480, warehouse chart with the speed of 1 frame per second from fish-eye camera
As being transmitted by USB connecting line；
B, raspberry pie receives the warehouse image that fish-eye camera transmits, and image input fire alarm model in warehouse is carried out pre-
It surveys, calculation processing is carried out to warehouse image using fire alarm model by tensorflow frame, obtains Warehouse Fire index；
C, raspberry pie judges result, judges whether storehouse occurrence index is more than the given threshold 5 of setting, if being more than to set
Determine threshold value, then represent have fire, pass the result to server, if be no more than given threshold, abandon as a result, continue into
The prediction of row next frame；
D, server receives fire alarm, carries out fire extinguishing processing by storekeeper personnel.
The effect data of fire prediction method and existing method of the present invention compares as shown in table 1:
Table 1
| Fire accuracy rate | Fire finds speed | Cost | |
| Fire prediction method of the present invention | 90~95% | 5~10 seconds | 500~700 yuan |
| The prior art | 70~75% | 1 minute or more | 1500 yuan or more |
As shown in Table 1, compared with prior art, fire prediction method of the present invention can greatly improve the accurate of fire alarm
Degree and speed can find fire condition compared to traditional sensors alarm mode faster, inform storehouse management administrator
Member, avoids great property loss.
In step (3), the intelligent barrier avoiding of trolley is realized, as shown in Fig. 2, comprising the following steps that
E, raspberry pie obtains the road image of resolution ratio 640*480, road with the speed of 20 frame per second from infrared camera
Image is transmitted by USB connecting line；
F, raspberry pie receives the road image that infrared camera transmits, and road image input automatic obstacle-avoiding model is carried out pre-
Survey, by tensorflow frame using automatic obstacle-avoiding model to picture carry out calculation processing, obtain trolley advance angle with
Speed；
G, the respective revolving speed of each DC speed-reducing, raspberry are calculated by formula according to the angle and speed predicted
Group passes the result to control motor drive module by connecting line；
H, control obstacle avoidance module receives four respective revolving speeds of DC speed-reducing, four motor interface difference of expansion board
Four DC speed-reducing revolving speeds are controlled, the angle and speed that control trolley advances achieve the effect that automatic obstacle-avoiding.
Claims (10)
1. a kind of intelligent patrol detection fire early-warning system based on fish-eye camera trolley, which is characterized in that including Image Acquisition mould
Block, raspberry pie, server, control motor drive module；
The image of acquisition is sent to the raspberry pie for acquiring image by described image acquisition module；The raspberry pie is used
In the speed and angle information of realization fire alarm or acquisition trolley, and the case where fire alarm, is sent to the server；
Control motor drive module is used to receive the speed and angle information of trolley, realizes the intelligent barrier avoiding of trolley；Server is for connecing
The case where receiving fire alarm is simultaneously handled in time.
2. a kind of intelligent patrol detection fire early-warning system based on fish-eye camera trolley according to claim 1, feature
Be, described image acquisition module includes fish-eye camera, infrared camera, the raspberry pie include fire alarm module, from
Dynamic obstacle avoidance module, the fish-eye camera connect the fire alarm module, the infrared camera, the automatic obstacle-avoiding mould
Block, the control motor drive module are sequentially connected；
The control motor drive module includes 4WD expansion board, four DC speed-reducings and tire；4WD expansion board passes through row
Line connects the raspberry pie, and 4WD expansion board is separately connected four DC speed-reducings by four motor interfaces, and four direct currents subtract
Speed motor is separately connected four tires, and 4WD expansion board is used to receive the speed for the trolley that the raspberry pie transmits, and passes through motor
Interface controls the revolving speed of four DC speed-reducings respectively；
The fish-eye camera is used to acquire the image of the scene to early warning fire, and the image of acquisition is sent to the fire
Warning module；
The fire alarm module refers to: input figure for being detected by machine learning tensorflow frame to image
As arriving the fire alarm model, by operation, the fire alarm model exports occurrence index, and occurrence index is fire alarm
The value for making a reservation for output when model training represents fire, fire alarm is believed if occurrence index is greater than given threshold
Breath passes to backstage manager, otherwise, without processing；
The frame road image that the infrared camera is 640*480 for acquisition resolution, and the road image of acquisition is sent out
It send to the automatic obstacle-avoiding module；
The automatic obstacle-avoiding module carries out operation to image using automatic obstacle-avoiding model by machine learning tensorflow frame,
Predict the angle and speed of trolley advance, angle and speed preset output, raspberry pie when being the training of automatic obstacle-avoiding module
According to angle and speed, the revolving speed of four DC speed-reducings is calculated by formula, raspberry pie is by four DC speed-reducings
Revolving speed 4WD expansion board is passed to by connecting line, the revolving speed of 4WD expansion board control four motor interfaces of connection controls trolley
The angle and speed of advance, to realize intelligent barrier avoiding.
3. a kind of intelligent patrol detection fire early-warning system based on fish-eye camera trolley according to claim 2, feature
It is, the calculation formula of speed and revolving speed are as follows: V=v2 π R, V refer to speed, and v refers to that revolving speed, R refer to the radius of gyration.
4. a kind of intelligent patrol detection fire early-warning system based on fish-eye camera trolley according to claim 1, feature
It is, the fire alarm model includes 5 layers of convolutional layer, 2 layers of pond layer, 3 layers of full articulamentum, activation primitive ELU.
5. a kind of intelligent patrol detection fire early-warning system based on fish-eye camera trolley according to claim 1, feature
It is, the automatic obstacle-avoiding model is End-to-End nvidia-cnn, including 5 layers of convolutional layer, 3 layers of full articulamentum, activates letter
Number is ELU.
6. a kind of intelligent patrol detection fire early-warning system based on fish-eye camera trolley according to claim 2, feature
It is, the fish-eye camera connects the fire alarm module by USB connecting line；The infrared camera is connected by USB
Wiring connects the automatic obstacle-avoiding module.
7. a kind of intelligent patrol detection fire early-warning system based on fish-eye camera trolley according to claim 2, feature
It is, the fire alarm module and the automatic obstacle-avoiding module include the raspberry pie plate of model 3B+, SD card, and SD card is inserted
In raspberry pie；The fish-eye camera is 5,000,000 pixel USB fish-eye cameras of riel prestige view；The infrared camera is
The infrared camera of micro- snow；The control motor drive module includes expansion board, motor and tire, and control motor drive module connects
The revolving speed for receiving four DC speed-reducings that automatic obstacle-avoiding module transmits is controlled the revolving speed of four DC speed-reducings by expansion board
Realize automatic obstacle-avoiding.
8. the work side of any intelligent patrol detection fire early-warning system based on fish-eye camera trolley of claim 2,3,6,7
Method, which is characterized in that comprise the following steps that
(1) described image acquisition module acquires image, and the image of acquisition is sent to the raspberry pie；
(2) raspberry pie realizes fire alarm or obtains the speed and angle information of trolley, and the case where fire alarm is sent out
It send to the server, the speed of trolley and angle letter is sent to the server；
(3) the control motor drive module receives the speed and angle information of trolley, realizes the intelligent barrier avoiding of trolley, the clothes
Business device is for the case where receiving fire alarm and is handled in time.
9. the working method of the intelligent patrol detection fire early-warning system according to claim 8 based on fish-eye camera trolley,
It is characterized in that, raspberry pie realizes fire alarm in the step (2), comprise the following steps that
A, raspberry pie obtains the warehouse image of resolution ratio 640*480 with the speed of 1 frame per second from fish-eye camera, and warehouse image is logical
USB connecting line is crossed to be transmitted；
B, raspberry pie receives the warehouse image that fish-eye camera transmits, and image input fire alarm model in warehouse is predicted,
Calculation processing is carried out to warehouse image using fire alarm model by tensorflow frame, obtains Warehouse Fire index；
C, raspberry pie judges result, judges whether storehouse occurrence index is more than the given threshold of setting, if being more than setting threshold
Value, then representing has fire, passes the result to server, if being no more than given threshold, under abandoning as a result, continuing
The prediction of one frame；
D, server receives fire alarm, carries out fire extinguishing processing by storekeeper personnel.
10. the working method of the intelligent patrol detection fire early-warning system according to claim 8 based on fish-eye camera trolley,
It is characterized in that, realizing the intelligent barrier avoiding of trolley, comprise the following steps that
E, raspberry pie obtains the road image of resolution ratio 640*480, road image with the speed of 20 frame per second from infrared camera
It is transmitted by USB connecting line；
F, raspberry pie receives the road image that infrared camera transmits, and road image input automatic obstacle-avoiding model is predicted,
Calculation processing is carried out to picture using automatic obstacle-avoiding model by tensorflow frame, obtains angle and speed that trolley advances
Degree；
G, the respective revolving speed of each DC speed-reducing is calculated by formula according to the angle and speed predicted, raspberry pie is logical
It crosses connecting line and passes the result to control motor drive module；
H, control obstacle avoidance module receives four respective revolving speeds of DC speed-reducing, and four motor interfaces of expansion board control respectively
Four DC speed-reducing revolving speeds, the angle and speed that control trolley advances, achieve the effect that automatic obstacle-avoiding.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910303165.9A CN110007666A (en) | 2019-04-16 | 2019-04-16 | A kind of intelligent patrol detection fire early-warning system and its working method based on fish-eye camera trolley |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910303165.9A CN110007666A (en) | 2019-04-16 | 2019-04-16 | A kind of intelligent patrol detection fire early-warning system and its working method based on fish-eye camera trolley |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110007666A true CN110007666A (en) | 2019-07-12 |
Family
ID=67172215
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910303165.9A Pending CN110007666A (en) | 2019-04-16 | 2019-04-16 | A kind of intelligent patrol detection fire early-warning system and its working method based on fish-eye camera trolley |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110007666A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113885446A (en) * | 2020-04-16 | 2022-01-04 | 福州泰全工业有限公司 | Automobile production line quality control method for realizing quality trend prediction |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170039424A1 (en) * | 2012-08-06 | 2017-02-09 | Cloudparc, Inc. | Tracking a vehicle using an unmanned aerial vehicle |
| CN106683323A (en) * | 2017-01-24 | 2017-05-17 | 深圳双创科技发展有限公司 | Warning management method based on robot and system thereof |
| WO2017087583A1 (en) * | 2015-11-16 | 2017-05-26 | Campbell Robert Marshal | Camera sight device for a weapon |
| CN207352734U (en) * | 2017-11-01 | 2018-05-11 | 辛禹灯 | A kind of classroom fire alarm robot |
| CN108389359A (en) * | 2018-04-10 | 2018-08-10 | 中国矿业大学 | A kind of Urban Fires alarm method based on deep learning |
| CN108687742A (en) * | 2018-08-23 | 2018-10-23 | 合肥合安智为科技有限公司 | A kind of pole incipient fire early warning intelligent inspection robot |
| CN108932814A (en) * | 2018-07-18 | 2018-12-04 | 东华大学 | A kind of embedded image type cooking fire warning device |
| CN109276833A (en) * | 2018-08-01 | 2019-01-29 | 吉林大学珠海学院 | A kind of robot patrol fire-fighting system and its control method based on ROS |
| CN109298723A (en) * | 2018-11-30 | 2019-02-01 | 山东大学 | A kind of accurate landing method of vehicle-mounted unmanned aerial vehicle and system |
-
2019
- 2019-04-16 CN CN201910303165.9A patent/CN110007666A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170039424A1 (en) * | 2012-08-06 | 2017-02-09 | Cloudparc, Inc. | Tracking a vehicle using an unmanned aerial vehicle |
| WO2017087583A1 (en) * | 2015-11-16 | 2017-05-26 | Campbell Robert Marshal | Camera sight device for a weapon |
| CN106683323A (en) * | 2017-01-24 | 2017-05-17 | 深圳双创科技发展有限公司 | Warning management method based on robot and system thereof |
| CN207352734U (en) * | 2017-11-01 | 2018-05-11 | 辛禹灯 | A kind of classroom fire alarm robot |
| CN108389359A (en) * | 2018-04-10 | 2018-08-10 | 中国矿业大学 | A kind of Urban Fires alarm method based on deep learning |
| CN108932814A (en) * | 2018-07-18 | 2018-12-04 | 东华大学 | A kind of embedded image type cooking fire warning device |
| CN109276833A (en) * | 2018-08-01 | 2019-01-29 | 吉林大学珠海学院 | A kind of robot patrol fire-fighting system and its control method based on ROS |
| CN108687742A (en) * | 2018-08-23 | 2018-10-23 | 合肥合安智为科技有限公司 | A kind of pole incipient fire early warning intelligent inspection robot |
| CN109298723A (en) * | 2018-11-30 | 2019-02-01 | 山东大学 | A kind of accurate landing method of vehicle-mounted unmanned aerial vehicle and system |
Non-Patent Citations (2)
| Title |
|---|
| M.VANITHA，等: "Monitoring and Controlling of Mobile Robot Via Internet Through Raspberry PI Board", 《2016 SECOND INTERNATIONAL CONFERENCE ON SCIENCE TECHNOLOGY ENGINEERING AND MANAGEMENT》 * |
| 白辰甲: "基于计算机视觉和深度学习的自动驾驶方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113885446A (en) * | 2020-04-16 | 2022-01-04 | 福州泰全工业有限公司 | Automobile production line quality control method for realizing quality trend prediction |
| CN113885446B (en) * | 2020-04-16 | 2024-06-07 | 福州泰全工业有限公司 | Automobile production line quality management method for realizing quality trend prediction |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105563488B (en) | A kind of Ye Xun robots | |
| US11113841B1 (en) | Self-learning three-dimensional boundaries from images | |
| CN112487891B (en) | Visual intelligent dynamic identification model construction method applied to electric power operation site | |
| CN109558785A (en) | Safety defense monitoring system and the unmanned convenience store for applying it | |
| CN108229407A (en) | A kind of behavioral value method and system in video analysis | |
| US11763566B2 (en) | Target association using occlusion analysis, clustering, or both | |
| CN107273880A (en) | A kind of multi-storied garage safety-protection system and method based on intelligent video monitoring | |
| CN112911221B (en) | Remote live-action storage supervision system based on 5G and VR videos | |
| US20230007831A1 (en) | Method for warehouse storage-location monitoring, computer device, and non-volatile storage medium | |
| CN109409238A (en) | A kind of obstacle detection method, device and terminal device | |
| CN111274934A (en) | Implementation method and system for intelligently monitoring forklift operation track in warehousing management | |
| CN108734910A (en) | A kind of abnormal behaviour monitoring and warning system and method | |
| CN110008919A (en) | Four rotor unmanned aerial vehicle face identification system based on vision | |
| CN110647816A (en) | Target detection method for real-time monitoring of goods shelf medicines | |
| CN106331646A (en) | Artificial intelligence safety monitoring system | |
| US20220335725A1 (en) | Monitoring presence or absence of an object using local region matching | |
| CN113516102A (en) | Deep learning parabolic behavior detection method based on video | |
| CN103530988B (en) | A kind of lawn is trampleed on and the method for early warning of the extent of injury | |
| CN114020043A (en) | Unmanned aerial vehicle building project supervision system and method, electronic equipment and storage medium | |
| US20240242496A1 (en) | Adversarial masks for scene-customized false detection removal | |
| CN110007666A (en) | A kind of intelligent patrol detection fire early-warning system and its working method based on fish-eye camera trolley | |
| US11893714B2 (en) | Precipitation removal from video | |
| US11823452B2 (en) | Video analytics evaluation | |
| CN113569971A (en) | Image recognition-based catch target classification detection method and system | |
| CN116659518B (en) | Autonomous navigation method, device, terminal and medium for intelligent wheelchair |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190712 |