CN110436294A - A kind of battery truck enters elevator detection method - Google Patents
A kind of battery truck enters elevator detection method Download PDFInfo
- Publication number
- CN110436294A CN110436294A CN201910729623.5A CN201910729623A CN110436294A CN 110436294 A CN110436294 A CN 110436294A CN 201910729623 A CN201910729623 A CN 201910729623A CN 110436294 A CN110436294 A CN 110436294A
- Authority
- CN
- China
- Prior art keywords
- elevator
- battery truck
- layer
- data
- target object
- 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
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B66—HOISTING; LIFTING; HAULING
- B66B—ELEVATORS; ESCALATORS OR MOVING WALKWAYS
- B66B5/00—Applications of checking, fault-correcting, or safety devices in elevators
- B66B5/0006—Monitoring devices or performance analysers
Landscapes
- Image Analysis (AREA)
Abstract
Present invention mainly discloses a kind of battery trucks to enter elevator detection method, comprising the following steps: S1, elevator car roof installation have the camera of voice, shoot to the target object entered in lift car, camera is acquired data to target object；S2, data are decoded by Video Decoder, then data are transmitted to algorithm plate and carry out battery truck detection by deep learning algorithm, carry out picture analyzing using NPU and export analysis target object with the presence or absence of there is battery truck；S3, if there is battery truck, generate elevator control signals and be sent to background server, play alarm voice, and elevator door does not turn off, elevator temporarily ceases operation；S4, when battery truck remove lift car, disconnect elevator control signals, stop playing voice, the present invention uses deep learning algorithm, and picture analyzing and processing, after finding battery truck, timely voice prompting are carried out in special NPU, message is sent and elevator door control, finds simultaneously early warning battery truck into elevator in time.
Description
Technical field
The present invention relates to elevator detection technique field, especially a kind of battery truck enters elevator detection method.
Background technique
The accident for entering elevator initiation by battery truck at present is more and more, and especially battery truck family's charging upstairs generates fire, meeting
It causes a serious accident, and the usually not early warning and the precautionary measures of elevator on the market needs to reduce the generation of safety accident
Research and development battery truck enters elevator detection system.Market is used for the scheme of target detection at present, needs mostly using gpu, not using gpu
It is only expensive, it is bulky, and also power consumption is very high, is unfavorable for the integrated and application of product.
Summary of the invention
In view of the deficienciess of the prior art, the present invention, which provides a kind of battery truck, enters elevator detection method, pass through acquisition electricity
Terraced monitoring camera data carry out picture analyzing using deep learning algorithm and NPU, find simultaneously early warning battery truck into electricity in time
The situation of ladder.
In order to achieve the above object, the present invention is achieved through the following technical solutions: a kind of battery truck enters elevator detection side
Method, it is characterised in that the following steps are included:
S1, elevator car roof installation have the camera of voice, clap the target object entered in lift car
It takes the photograph, camera is acquired data to target object；
S2, data are decoded by Video Decoder, then data be transmitted to algorithm plate by deep learning algorithm into
Row battery truck detection carries out picture analyzing using NPU and exports analysis target object with the presence or absence of there is battery truck；
S3, if there is battery truck, generate elevator control signals and be sent to background server, play alarm voice, and elevator
Door does not turn off, and elevator temporarily ceases operation；
S4, when battery truck remove lift car, disconnect elevator control signals, stop play voice.
The present invention is further, deep learning frame is Caffe.
The present invention further, using NPU carry out picture analyzing include that picture is cut into image block, be input to neural network
First layer carries the data to the second layer in each neuron of first layer, and the neuron of the second layer carries the data to
Three layers, and so on, to the last one layer, then generate target object testing result.
Deep learning frame used in the present invention is Caffe, full name Convolutional Architecture for
Fast Feature Embedding is the deep learning frame for having both expressivity, speed and Thinking module.Caffe
It increases income under BSD license, is write using C++, have Python interface.Caffe is applied to academic research project, establishment prototype very
To the large-scale industrial application of vision, voice and MultiMedia Field.Caffe increases income completely, and multiple enlivens community's ditch having
It is logical to answer a question, while providing one and being used to the complete tools packet such as train, test, it may help to user's quickly upper hand.This
Outer Caffe is realized with the design of modularization principle to new data format, network layer and loss function easy expansion.Caffe is
Through the Protocl Buffer Definition Model file with Google.Network structure is indicated using special text file prototxt,
Network struction in the form of directed acyclic graph.GPU accelerates: MKL, Open BLAS, cu BLAS etc. is utilized and calculates library, utilizes
GPU, which realizes to calculate, to be accelerated.Data structure in Caffe is existed in the form of Blobs-layers-Net.Wherein, Blobs is logical
Cross all weights, activation value and Direct/Reverse in 4 dimensional vector forms (num, channel, height, width) storage network
Data.As the standard data format of Caffe, Blob provides unified memory interface.What Layers was indicated is neural network
In specific layer, such as convolutional layer etc., be Caffe model essential content and execute calculate basic unit.Layer layers of reception bottom
The Blobs of layer input exports Blobs to high level.Propagated forward, back-propagating can be realized at every layer.Net is connected by multiple layers
Together, the directed acyclic graph of composition.Initial data data Layer load data are started the layer of loss to the end by one network
Group is combined into entirety.
Use VGG-16 as trunk model in terms of neural network, the modification full articulamentum fc6 of VGG is 3*3 convolutional layer
Conv6, modification fc7 are 1*1 convolutional layer conv7, while pond layer pool5 being become by the 2*2 of original stride=2
The 3*3 of stride=1.Separately increase by 4 convolutional layers.Conv4_3 layers are first characteristic pattern detection module.Increase convolutional layer newly below
Successively are as follows: Conv8_2, Conv9_2, Conv10_2, Conv11_2.Check box is arranged on characteristic pattern mainly to consider: scale is (big
It is small) and length-width ratio.Check box can inswept characteristic pattern each point, generate corresponding check box.Confidence can be retained during prediction
Top-k high BOX is spent, is then overlapped biggish BOX using the filtering of nms algorithm, last remaining BOX is the knot predicted
Fruit.
The present invention has the beneficial effect that
The present invention, which uses, uses deep learning algorithm, and carries out picture analyzing and processing in special NPU, in discovery storage battery
Che Hou, timely voice prompting, message is sent and elevator door control, discovery in time and the case where early warning battery truck is into elevator.It adopts
It is not only small in size with dedicated NPU, it is at low cost, and also power consumption is very low, especially suitable for edge calculations occasion.
Detailed description of the invention
Fig. 1 is battery truck detection system flow chart of the present invention；
Specific embodiment
In conjunction with attached drawing, present pre-ferred embodiments are described in further details.
A kind of battery truck as described in Figure 1 enters elevator detection method, it is characterised in that the following steps are included:
S1, elevator car roof installation have the camera of voice, clap the target object entered in lift car
It takes the photograph, camera is acquired data to target object；
S2, data are decoded by Video Decoder, then data be transmitted to algorithm plate by deep learning algorithm into
Row battery truck detection carries out picture analyzing using NPU and exports analysis target object with the presence or absence of there is battery truck；Using NPU into
Row picture analyzing includes that picture is cut into image block, is input to neural network first layer, in each neuron of first layer
The second layer is carried the data to, the neuron of the second layer carries the data to third layer, and so on, to the last one layer, so
Target object testing result is generated afterwards；
S3, if there is battery truck, generate elevator control signals and be sent to background server, play alarm voice, and elevator
Door does not turn off, and elevator temporarily ceases operation；
S4, when battery truck remove lift car, disconnect elevator control signals, stop play voice.
Learning algorithm intralamellar part program has an automatic upgrade function, supports local, network two ways upgrading, facilitates the later period to be
System upgrading and maintenance；
Deep learning frame used in the present invention is Caffe, full name Convolutional Architecture for
Fast Feature Embedding is the deep learning frame for having both expressivity, speed and Thinking module.Caffe
It increases income under BSD license, is write using C++, have Python interface.Caffe is applied to academic research project, establishment prototype very
To the large-scale industrial application of vision, voice and MultiMedia Field.Caffe increases income completely, and multiple enlivens community's ditch having
It is logical to answer a question, while providing one and being used to the complete tools packet such as train, test, it may help to user's quickly upper hand.This
Outer Caffe is realized with the design of modularization principle to new data format, network layer and loss function easy expansion.Caffe is
Through the Protocl Buffer Definition Model file with Google.Network structure is indicated using special text file prototxt,
Network struction in the form of directed acyclic graph.GPU accelerates: MKL, Open BLAS, cu BLAS etc. is utilized and calculates library, utilizes
GPU, which realizes to calculate, to be accelerated.Data structure in Caffe is existed in the form of Blobs-layers-Net.Wherein, Blobs is logical
Cross all weights, activation value and Direct/Reverse in 4 dimensional vector forms (num, channel, height, width) storage network
Data.As the standard data format of Caffe, Blob provides unified memory interface.What Layers was indicated is neural network
In specific layer, such as convolutional layer etc., be Caffe model essential content and execute calculate basic unit.Layer layers of reception bottom
The Blobs of layer input exports Blobs to high level.Propagated forward, back-propagating can be realized at every layer.Net is connected by multiple layers
Together, the directed acyclic graph of composition.Initial data data Layer load data are started the layer of loss to the end by one network
Group is combined into entirety.
Use VGG-16 as trunk model in terms of neural network, the modification full articulamentum fc6 of VGG is 3*3 convolutional layer
Conv6, modification fc7 are 1*1 convolutional layer conv7, while pond layer pool5 being become by the 2*2 of original stride=2
The 3*3 of stride=1.Separately increase by 4 convolutional layers.Conv4_3 layers are first characteristic pattern detection module.Increase convolutional layer newly below
Successively are as follows: Conv8_2, Conv9_2, Conv10_2, Conv11_2.Check box is arranged on characteristic pattern mainly to consider: scale is (big
It is small) and length-width ratio.Check box can inswept characteristic pattern each point, generate corresponding check box.Confidence can be retained during prediction
Top-k high BOX is spent, is then overlapped biggish BOX using the filtering of nms algorithm, last remaining BOX is the knot predicted
Fruit.
The present invention, which uses, uses deep learning algorithm, and carries out picture analyzing and processing in special NPU, in discovery storage battery
Che Hou, timely voice prompting, message is sent and elevator door control, discovery in time and the case where early warning battery truck is into elevator.It adopts
It is not only small in size with dedicated NPU, it is at low cost, and also power consumption is very low, especially suitable for edge calculations occasion.
Above-described embodiment is only used for illustrating inventive concept of the invention, rather than the restriction to rights protection of the present invention,
It is all to be made a non-material change to the present invention using this design, protection scope of the present invention should all be fallen into.
Claims (3)
1. a kind of battery truck enters elevator detection method, it is characterised in that the following steps are included:
S1, elevator car roof installation have the camera of voice, shoot, take the photograph to the target object entered in lift car
As head is acquired data to target object；
S2, data are decoded by Video Decoder, then data are transmitted to algorithm plate and carry out electricity by deep learning algorithm
Bottle car test is surveyed, and is carried out picture analyzing using NPU and is exported analysis target object with the presence or absence of there is battery truck；
S3, if there is battery truck, generate elevator control signals and be sent to background server, play alarm voice, and elevator door is not
It can close, elevator temporarily ceases operation；
S4, when battery truck remove lift car, disconnect elevator control signals, stop play voice.
2. a kind of battery truck according to claim 1 enters elevator detection method, it is characterised in that: deep learning frame is
Caffe。
3. a kind of battery truck according to claim 1 enters elevator detection method, it is characterised in that: carry out picture using NPU
Analysis includes that picture is cut into image block, is input to neural network first layer, first layer each neuron data
It is transmitted to the second layer, the neuron of the second layer carries the data to third layer, and so on, it to the last one layer, then generates
Target object testing result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910729623.5A CN110436294A (en) | 2019-08-08 | 2019-08-08 | A kind of battery truck enters elevator detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910729623.5A CN110436294A (en) | 2019-08-08 | 2019-08-08 | A kind of battery truck enters elevator detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110436294A true CN110436294A (en) | 2019-11-12 |
Family
ID=68433886
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910729623.5A Pending CN110436294A (en) | 2019-08-08 | 2019-08-08 | A kind of battery truck enters elevator detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110436294A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111160321A (en) * | 2020-02-10 | 2020-05-15 | 杭州大数云智科技有限公司 | Storage battery car goes up terraced detection and early warning system |
| CN111353451A (en) * | 2020-03-06 | 2020-06-30 | 深圳市赛为智能股份有限公司 | Battery car detection method and device, computer equipment and storage medium |
| CN111439644A (en) * | 2020-02-28 | 2020-07-24 | 浙江大华技术股份有限公司 | Alarming method of storage battery car in elevator and related device |
| CN111776905A (en) * | 2020-08-13 | 2020-10-16 | 浙江新再灵科技股份有限公司 | Battery car elevator entering warning method and system combining re-identification |
| CN112079213A (en) * | 2020-08-24 | 2020-12-15 | 浙江新再灵科技股份有限公司 | Elevator entry control method and elevator entry control system |
| CN112347873A (en) * | 2020-10-26 | 2021-02-09 | 浙江新再灵科技股份有限公司 | Ladder control method |
| CN112712048A (en) * | 2021-01-11 | 2021-04-27 | 武汉爱科森网络科技有限公司 | Method for monitoring and early warning of entering of electric vehicle into building |
| CN113291943A (en) * | 2021-06-10 | 2021-08-24 | 通力电梯有限公司 | Detection control system and detection control method suitable for elevator |
| CN113903143A (en) * | 2021-09-26 | 2022-01-07 | 深圳市爱深盈通信息技术有限公司 | Electric vehicle monitoring method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107609597A (en) * | 2017-09-26 | 2018-01-19 | 嘉世达电梯有限公司 | A kind of number of people in lift car detecting system and its detection method |
| CN107601236A (en) * | 2017-11-06 | 2018-01-19 | 孙复庆 | Lift car vehicle prevents system |
| CN207566690U (en) * | 2017-12-14 | 2018-07-03 | 杭州容创电梯有限公司 | A kind of intelligent anticollision car |
| CN108290704A (en) * | 2015-11-16 | 2018-07-17 | 通力股份公司 | Method and apparatus for determining Decision of Allocation at least one elevator |
| CN109179133A (en) * | 2018-11-05 | 2019-01-11 | 常熟理工学院 | For prejudging the elevator intelligent maintenance prediction technique and system of failure |
| CN110002315A (en) * | 2018-11-30 | 2019-07-12 | 浙江新再灵科技股份有限公司 | Vertical ladder electric vehicle detection method and warning system based on deep learning |
-
2019
- 2019-08-08 CN CN201910729623.5A patent/CN110436294A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108290704A (en) * | 2015-11-16 | 2018-07-17 | 通力股份公司 | Method and apparatus for determining Decision of Allocation at least one elevator |
| CN107609597A (en) * | 2017-09-26 | 2018-01-19 | 嘉世达电梯有限公司 | A kind of number of people in lift car detecting system and its detection method |
| CN107601236A (en) * | 2017-11-06 | 2018-01-19 | 孙复庆 | Lift car vehicle prevents system |
| CN207566690U (en) * | 2017-12-14 | 2018-07-03 | 杭州容创电梯有限公司 | A kind of intelligent anticollision car |
| CN109179133A (en) * | 2018-11-05 | 2019-01-11 | 常熟理工学院 | For prejudging the elevator intelligent maintenance prediction technique and system of failure |
| CN110002315A (en) * | 2018-11-30 | 2019-07-12 | 浙江新再灵科技股份有限公司 | Vertical ladder electric vehicle detection method and warning system based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 杨露菁等: "《智能图像处理及应用》", 31 March 2019 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111160321A (en) * | 2020-02-10 | 2020-05-15 | 杭州大数云智科技有限公司 | Storage battery car goes up terraced detection and early warning system |
| CN111439644A (en) * | 2020-02-28 | 2020-07-24 | 浙江大华技术股份有限公司 | Alarming method of storage battery car in elevator and related device |
| CN111353451A (en) * | 2020-03-06 | 2020-06-30 | 深圳市赛为智能股份有限公司 | Battery car detection method and device, computer equipment and storage medium |
| CN111776905A (en) * | 2020-08-13 | 2020-10-16 | 浙江新再灵科技股份有限公司 | Battery car elevator entering warning method and system combining re-identification |
| CN112079213A (en) * | 2020-08-24 | 2020-12-15 | 浙江新再灵科技股份有限公司 | Elevator entry control method and elevator entry control system |
| CN112079213B (en) * | 2020-08-24 | 2022-08-23 | 浙江新再灵科技股份有限公司 | Elevator entry control method and elevator entry control system |
| CN112347873A (en) * | 2020-10-26 | 2021-02-09 | 浙江新再灵科技股份有限公司 | Ladder control method |
| CN112712048A (en) * | 2021-01-11 | 2021-04-27 | 武汉爱科森网络科技有限公司 | Method for monitoring and early warning of entering of electric vehicle into building |
| CN113291943A (en) * | 2021-06-10 | 2021-08-24 | 通力电梯有限公司 | Detection control system and detection control method suitable for elevator |
| CN113903143A (en) * | 2021-09-26 | 2022-01-07 | 深圳市爱深盈通信息技术有限公司 | Electric vehicle monitoring method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110436294A (en) | A kind of battery truck enters elevator detection method | |
| Dong et al. | A lightweight vehicles detection network model based on YOLOv5 | |
| CN109409242B (en) | Black smoke vehicle detection method based on cyclic convolution neural network | |
| CN115082855B (en) | Pedestrian shielding detection method based on improved YOLOX algorithm | |
| CN106611162A (en) | Method for real-time detection of road vehicle based on deep learning SSD frame | |
| CN113420607A (en) | Multi-scale target detection and identification method for unmanned aerial vehicle | |
| CN104516497A (en) | Method for controlling electronic device, handheld electronic device and monitoring system | |
| CN106934378A (en) | A kind of dazzle light identifying system and method based on video depth study | |
| CN107679495A (en) | A kind of detection method of transmission line of electricity periphery activity engineering truck | |
| CN203849870U (en) | Wireless network-based traffic flow monitoring and signal lamp intelligent control system | |
| Zhang et al. | Transmission line abnormal target detection based on machine learning yolo v3 | |
| CN103530640A (en) | Unlicensed vehicle detection method based on AdaBoost and SVM (support vector machine) | |
| CN104809470A (en) | Vehicle converse running detection device and method based on SVM | |
| CN103281518A (en) | Multifunctional networking all-weather intelligent video monitoring system | |
| Ding | LENet: Lightweight and efficient LiDAR semantic segmentation using multi-scale convolution attention | |
| CN110065867B (en) | Method and system for evaluating elevator comfort level based on audio and video | |
| CN206209759U (en) | A kind of online fugitive personnel's automatic identification synthesis police service booth | |
| CN204613981U (en) | A kind of vehicle based on SVM drives in the wrong direction pick-up unit | |
| CN113705427B (en) | Fatigue driving monitoring and early warning method and system based on vehicle-mounted chip SoC | |
| CN114462578A (en) | Method for improving forecast precision of short rainfall | |
| CN114359578A (en) | Application method and system of pest and disease damage identification intelligent terminal | |
| Zhang et al. | Electric bicycle detection based on improved YOLOv5 | |
| CN117593623A (en) | Lightweight vehicle detection method based on improved YOLOv8n model | |
| CN116934627A (en) | Target detection method in low-light environment | |
| CN103402079A (en) | VW (Video Wall)-based intelligent tracking, storing and displaying method for abnormal events |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191112 |
|
| RJ01 | Rejection of invention patent application after publication |