CN113435296A - Method, system, storage medium and elevator for detecting foreign matters based on rotated-yolov5 - Google Patents
Method, system, storage medium and elevator for detecting foreign matters based on rotated-yolov5 Download PDFInfo
- Publication number
- CN113435296A CN113435296A CN202110696320.5A CN202110696320A CN113435296A CN 113435296 A CN113435296 A CN 113435296A CN 202110696320 A CN202110696320 A CN 202110696320A CN 113435296 A CN113435296 A CN 113435296A
- Authority
- CN
- China
- Prior art keywords
- rotated
- elevator
- yolov5
- foreign matter
- neural network
- 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
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000001514 detection method Methods 0.000 claims abstract description 40
- 238000013528 artificial neural network Methods 0.000 claims abstract description 36
- 238000004364 calculation method Methods 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 238000012805 post-processing Methods 0.000 claims abstract description 5
- 238000001914 filtration Methods 0.000 claims abstract description 4
- 230000001629 suppression Effects 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 29
- 230000006870 function Effects 0.000 claims description 13
- 238000011478 gradient descent method Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000008014 freezing Effects 0.000 description 1
- 238000007710 freezing Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/046—Forward inferencing; Production systems
Abstract
The invention discloses a method and a system for detecting foreign matters in an elevator, a storage medium and the elevator based on rotated-yolov5, and relates to the technical field of elevator safety, wherein each frame of image is decoded by acquiring a video stream of a camera in the elevator; preprocessing the image, feeding the obtained dimensional data into a rotated-yolov5 neural network, performing forward reasoning calculation, and outputting a foreign matter prediction box with an angle; inputting the prediction frame into a post-processing and non-maximum value suppression module, and filtering to obtain an effective foreign matter detection frame; and judging whether the foreign matter detection frame falls into the safety area. The invention can not only detect the position of the foreign body, but also detect the rotation angle of the foreign body, thereby greatly improving the precision and robustness of the detection algorithm of the invention.
Description
Technical Field
The invention relates to the technical field of elevator safety, in particular to an elevator foreign matter detection method and system based on a rotated-yolov5 neural network, a computer storage medium and an elevator.
Background
Early traditional elevator foreign matter detection system's automation was relatively weak, and the personnel of generally adopting on duty looked over the camera and monitored elevator safety in real time, or adopted and set up infrared ray in the lift-cabin door and monitored the object that gets into in the elevator. With the rise of deep learning nowadays, more and more target detection algorithms are applied to the elevator foreign matter detection system. However, most of the algorithms locate the foreign matters based on a horizontal target detection method, and the foreign matters have a certain rotation angle in an actual scene, so that the elevator foreign matter detection method is not accurate enough, and the monitoring capability of the elevator on the foreign matters is reduced.
Disclosure of Invention
In view of the problems in the background art, a first object of the present invention is to provide a method for detecting foreign matters in an elevator based on rotated-yolov5, which improves the detection accuracy and robustness of foreign matters entering the elevator and improves the safety of the elevator.
In order to achieve the purpose, the invention provides the following technical scheme:
a foreign matter detection method for an elevator based on rotated-yolov5 comprises the following steps:
(1) acquiring a video stream of a camera in an elevator, and decoding each frame of image;
(2) preprocessing the image to obtain [1, 3, 416, 416] dimensional data and feeding the data into a rotated-yolov5 neural network;
(3) outputting a foreign matter prediction box with an angle after forward reasoning calculation in a rotated-yolov5 network;
(4) inputting the prediction frame into a post-processing and non-maximum value suppression module, and filtering to obtain an effective foreign matter detection frame;
(5) and (4) judging whether the foreign matter detection frame falls into the safety area, if so, automatically alarming, and if not, decoding the next frame of image and repeating the steps (2) to (5).
As a further scheme of the invention: the rotated-yolov5 neural network needs to be obtained by model training, and the training steps are as follows:
(a) collecting a large number of elevator scene picture data sets, and manually marking out all foreign matter information in the pictures, wherein the information mainly comprises real frames and categories of the foreign matters;
(b) preprocessing the pictures uniformly and generating a structured binary data file;
(c) feeding binary data files into a rotated-yolov5 neural network model in batches, and obtaining a foreign body prediction box after forward reasoning calculation;
(d) comparing the prediction frame and the real frame of the foreign matter, and calculating a loss value between the prediction frame and the real frame;
(e) updating weight parameters in the rotated-yolov5 neural network through an error back propagation operation, and minimizing the loss value in (d) by using a gradient descent method;
(f) and (e) repeating the operations (c) to (e) until the number of iteration rounds is met, and finishing the training.
As a further scheme of the invention: in the step (a), the four vertex coordinates of the rectangular ROI of each foreign matter in the picture and the category of the foreign matter are labeled manually.
As a further scheme of the invention: the method for reading the image data and the label information of the foreign matters in the step (b) and preprocessing comprises the following steps: firstly, increasing noise and color disturbance to a picture, uniformly cutting and scaling the picture into 416x416 resolution, and converting a color channel into an RGB form; and coding the marking information of the foreign matters into real frame data, and generating a binary data file after preprocessing for training a rotated-yolov5 neural network.
As a further scheme of the invention: the step c comprises the following steps: after the forward reasoning calculation, 3 feature maps are obtained, namely high resolution, medium resolution and low resolution in sequence, the corresponding resolution sizes are respectively 52x52, 26x26 and 13x13, pixels on each feature map are provided with (4+1+180+ n) x3 channels, and the data of the first 4 channels respectively comprise the central coordinates x and y and the width and height w and h of the foreign matter in sequence; the 5 th channel represents probability information of whether foreign matters exist, the following 180 channels respectively represent the possibility that the foreign matters rotate in 0-179 angles in sequence, and the last n channels represent the probability that the foreign matters belong to the categories 0-n-1.
As a further scheme of the invention: the step d comprises the following steps:
step S41: the loss value between the predicted frame data and the real frame data of the foreign object is calculated by the loss function of the rotated-yolov5 neural network in such a manner that
The four items on the right of the equal sign are the positioning loss of the target, the confidence loss of the target, the rotation angle loss of the target and the classification loss of the target in turn,
it is calculated that the position loss between the positive sample box and the prediction box,
in order to make the model learn to distinguish whether there is a foreign object at the position,
with the softmax cross-entropy loss function,
using a two-class cross entropy loss function;
step S42: using a random gradient descent method pair
Is globally optimized, and a weight matrix of the rotated-yolov5 neural network is updated through back propagation calculation;
step S43: and adding 1 to the current training round number to finish the training of the current round.
As a still further scheme of the invention: the step (f) of judging whether the condition of the number of training rounds is satisfied is as follows: whether the current number of training rounds is equal to the total number of training rounds.
The second purpose of the invention is to provide an elevator foreign matter detection system based on a rotated-yolov5 neural network, which improves the detection precision and robustness of foreign matters entering an elevator and improves the use safety of the elevator.
The above object of the invention is achieved by the following technical scheme:
a rotated-yolov 5-based elevator foreign matter detection system comprises:
the acquisition module is used for acquiring and acquiring video information, image information and display information in the elevator;
a memory for storing the program of the elevator foreign matter detection method based on the rotated-yolov5 neural network;
the processor is loaded and executed by the program in the memory, and realizes the elevator foreign matter detection method based on the rotated-yolov5 neural network, and the processor comprises a preprocessing module, a rotated-yolov5 neural network module and a post-processing module, wherein the preprocessing module is used for processing the acquired information of the acquisition module.
A third object of the present invention is to provide a computer storage medium for implementing the monitoring accuracy and robustness of foreign objects in an elevator, and a method for improving the safety of elevator use.
The above object of the invention is achieved by the following technical scheme:
a computer storage medium comprising a program capable of being loaded and executed by a processor to implement the above-mentioned rotated-yolov5 neural network-based elevator foreign matter detection method.
A fourth object of the present invention is to provide an elevator based on rotated-yolov5, which can improve the monitoring precision and robustness of foreign matters in the elevator, thereby improving the safety of the elevator.
The above object of the invention is achieved by the following technical scheme:
an elevator based on rotated-yolov5 comprises
A processor for loading and executing a set of instructions; and a storage medium as described above.
By adopting the technical scheme, the IOU (interaction Over Unit) value of each detection frame and the safety region is calculated by calculating and analyzing the video information and the image information, if the IOU is higher than the manual set value, the system considers that foreign matters in the elevator enter the safety region, controls the elevator to be normally opened, automatically alarms and informs an attendant, and the use safety of the elevator is improved.
In conclusion, the beneficial effects of the invention are as follows:
(1) the aim of the invention can be realized by the minimum requirements of a common monocular camera and an edge calculation box, and the edge calculation box is internally provided with a corresponding system and a corresponding processing program.
(2) The invention can not only detect the position of the foreign body, but also detect the rotation angle of the foreign body, thereby greatly improving the precision and robustness of the detection algorithm of the foreign body in the elevator.
(3) The degree of automation is high, in case the foreign matter gets into the elevator can automatic alarm and inform the personnel on duty, improves the safety in utilization of elevator.
Drawings
Fig. 1 is a flow chart of elevator foreign matter detection according to the present invention.
FIG. 2 is a flow chart of the training of rotated-yolov 5.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, in the elevator foreign matter detection method based on rotated-yolov5 disclosed by the invention, a camera is installed in an elevator, objects in the elevator are monitored and analyzed in an artificial intelligence manner, dangerous goods in the elevator are detected, and the like, so that the opening and the keeping opening of the elevator door are controlled, an alarm is automatically given, a person on duty is notified, and the use safety of the elevator is improved.
Because the elevator foreign matter detection process mainly depends on a trained rotated-yolov5 neural network (called rotated-yolov5 for short), in the real-time case of the invention, firstly, the training process of the rotated-yolov5 neural network is described in detail, and the specific training process is as follows:
step S1: the image data Of different elevator scenes are collected in a large scale through the camera, a data set which is disclosed in the industry and has a good acknowledged effect can be downloaded from the Internet, and four vertex coordinates Of a rectangular ROI (region Of interest) Of each foreign matter in the image and the category Of the foreign matter are labeled manually.
The collected picture data should contain at least one foreign object, and meet the requirements of different viewing angles, multiple illumination changes and multiple scene changes as much as possible, and the pictures with distortion, single scene and gray scale should be discarded.
Because the types and the models of the elevators are different, the environment in the elevator can be recorded or photographed by selecting a common or high-definition monocular camera according to the field environment when the camera is installed, so that the scene information in the elevator is obtained, and the scene information is displayed in real time.
Step S2: reading the image data and the mark information of the foreign matters for preprocessing: firstly, increasing noise and color disturbance to a picture, uniformly cutting and scaling the picture into 416x416 resolution, and converting a color channel into an RGB form; and encoding the marking information of the foreign matter into real frame data. After the preprocessing is finished, a binary data file is generated and used for training a rotated-yolov5 neural network.
By artificially designing the backbone structure of the rotated-yolov5 neural network, it is possible to refer to the mainstream networks such as Resnet18, Resnet50, and HRnet. In order to balance the requirements of higher precision and lower computation, the depth parameter and the width parameter of the network are ranged between 0.95 and 1.05 after a lot of experiments, wherein the parameter is set to be 1.0 and is optimal.
Step S3: and (4) feeding the binary form data file obtained in the step (S2) into a rotated-yolov5 neural network, and obtaining the prediction box data of the foreign matters after forward propagation reasoning calculation.
After forward reasoning calculation, 3 feature maps (feature maps) are obtained, which are high resolution, medium resolution and low resolution in sequence, and the corresponding resolution sizes are 52x52, 26x26 and 13x13 respectively. The pixels on each feature map have (4+1+180+ n) x3 channels: the first 4 channel data respectively comprise the central coordinates x and y and the width and height w and h of the foreign matters in sequence; the 5 th channel represents probability information of whether foreign matters exist, the following 180 channels respectively represent the possibility that the foreign matters rotate in 0-179 angles in sequence, and the last n channels represent the probability that the foreign matters belong to the categories 0-n-1.
Step S4: a loss value between the predicted box data of step S3 and the real box data of step S2 is calculated, and the weight parameters of the rotated-yolov5 neural network are reversely updated by a stochastic gradient descent method.
The step S4 includes:
step S41: the loss value between the predicted frame data and the real frame data of the foreign object is calculated by the loss function of the rotated-yolov5 neural network in such a manner that
The four items on the right side of the equal sign are the positioning loss of the target, the confidence loss of the target, the rotation angle loss of the target and the classification loss of the target in turn.
It is calculated that the position loss between the positive sample box and the prediction box,
the purpose is to make the model learn to distinguish whether there is a foreign object at the position. One of the innovations of the present invention is that the prediction of the rotation angle is treated as a classification task, and a total of 180 classes are set and correspond to 0-179 angles, so that
A softmax cross entropy loss function is used. In order to solve the problem of missing detection due to the overlapping of a plurality of objects of the category,
a two-class cross entropy loss function is used.
Step S42: using a random gradient descent method pair
Is globally optimized and updates the weight matrix of the rotated-yolov5 neural network through back propagation calculations.
Step S43: and adding 1 to the current training round number to finish the training of the current round.
Step S5: and judging whether the training round number is met, if so, freezing the weight parameters of the rotated-yolov5 neural network, outputting the weight file of the model, and finishing the training. If not, step S3 is repeated.
In step S5, the condition for determining whether or not the number of training rounds is satisfied is: whether the current number of training rounds is equal to the total number of training rounds.
After steps S1-S5 are completed, the trained rotated-yolov5 neural network model is obtained.
The elevator foreign matter detection flow based on the rotated-yolov5 neural network of the invention is described as follows:
(1) the video camera is installed in the elevator, the camera pose of the video camera is fixed, and each frame of image is collected in real time.
(2) According to the shooting angle of the camera pose, a closed-loop polygonal area is artificially divided, and the area is set as a safety area.
(3) Inputting each frame of image into a preprocessing module to obtain an image with RGB color channels and 416x416 size, and normalizing the image into image data with [1, 3, 416, 416] dimension.
(4) And feeding the image data into a trained rotated-yolov5 neural network, and performing forward reasoning to obtain 10647 prediction boxes in total, wherein each prediction box has the characteristics of a central coordinate, a width and height dimension, a rotation angle, a confidence score and the like.
(5) And inputting all the prediction frames into a post-processing and non-maximum value suppression module, and filtering to obtain a final detection frame. Each detection box contains the coordinates Of the four vertices Of a rectangular ROI (region Of interest) region Of the detected alien material.
And calculating the IOU (interaction Over Unit) value of each detection frame and the safety area, and if the IOU value is higher than the manually set value, the system considers that foreign matters in the elevator enter the safety area and automatically alarms to notify operators on duty.
Based on the same inventive concept, the embodiment of the invention provides an elevator foreign matter detection system based on a rotated-yolov5 neural network, which comprises:
the acquisition module is used for acquiring and acquiring video information, image information and display information in the elevator;
a memory for storing a method as in FIGS. 1-2;
a processor, a program in memory capable of being loaded for execution by the processor and implementing the method as in fig. 1-2.
It will be clear to those skilled in the art that, for convenience and simplicity of description, the foregoing division of the functional modules is merely used as an example, and in practical applications, the above function distribution may be performed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to perform all or part of the above described functions. For the above-described system, the specific working process may refer to the corresponding process in the foregoing method embodiment, and details are not repeated here.
Based on the same inventive concept, the embodiment of the present invention also provides a computer storage medium, which can be loaded by a processor to execute the steps included in the flowcharts of fig. 1-2.
The computer storage medium includes, but is not limited to, various media that can store program codes, such as a usb disk, a removable hard disk, a Read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
Based on the same inventive concept, the embodiment of the invention also provides an elevator based on a rotated-yolov5 neural network, which comprises: a processor for loading and executing a set of instructions; and the storage medium described above.
By adopting the technical scheme, the IOU (interaction Over Unit) value of each detection frame and the safety region is calculated by calculating and analyzing the video information and the image information, if the IOU is higher than the manual set value, the system considers that foreign matters in the elevator enter the safety region, controls the elevator to be normally opened, automatically alarms and informs an attendant, the automation degree is high, and the use safety of the elevator is improved.
It will be clear to those skilled in the art that, for convenience and simplicity of description, the foregoing division of the functional modules is merely used as an example, and in practical applications, the above function distribution may be performed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to perform all or part of the above described functions. For the specific working processes of the system, the apparatus and the unit described above, reference may be made to the corresponding processes in the foregoing method embodiments, and details are not described here again.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, a network device, or the like) or a processor (processor) to execute all or part of the steps of the method according to the embodiments of the present application.
In the embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules or units is only one logical division, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
Claims (10)
1. A foreign matter detection method for an elevator based on rotated-yolov5 is characterized by comprising the following steps:
(1) acquiring a video stream of a camera in an elevator, and decoding each frame of image;
(2) preprocessing the image to obtain [1, 3, 416, 416] dimensional data and feeding the data into a rotated-yolov5 neural network;
(3) outputting a foreign matter prediction box with an angle after forward reasoning calculation in a rotated-yolov5 neural network;
(4) inputting the prediction frame into a post-processing and non-maximum value suppression module, and filtering to obtain an effective foreign matter detection frame;
(5) and (4) judging whether the foreign matter detection frame falls into the safety area, if so, automatically alarming, and if not, decoding the next frame of image and repeating the steps (2) to (5).
2. The method for detecting the foreign matters in the elevator based on the rotated-yolov5, according to claim 1, wherein the rotated-yolov5 neural network is obtained by model training, and the training steps are as follows:
(a) collecting a large amount of elevator scene picture data sets, and manually marking out all foreign matter information in the elevator scene pictures, including real frames and categories of the foreign matters;
(b) uniformly preprocessing the elevator scene pictures and generating a structured binary data file;
(c) feeding binary data files into a rotated-yolov5 neural network model in batches, and obtaining a foreign body prediction box after forward reasoning calculation;
(d) comparing the prediction frame and the real frame of the foreign matter, and calculating a loss value between the prediction frame and the real frame;
(e) updating weight parameters in the rotated-yolov5 neural network through an error back propagation operation, and minimizing the loss value in (d) by using a gradient descent method;
(f) and (e) repeating the operations (c) to (e) until the number of iteration rounds is met, and finishing the training.
3. The method for detecting the foreign matters in the elevator based on the rotated-yolov5 of claim 2, wherein the four vertex coordinates of the rectangular ROI of each foreign matter in the picture and the category of the foreign matter are manually marked in the step (a).
4. The method for detecting the foreign matters in the elevator based on the rotated-yolov5 of claim 2, wherein the method for reading the picture data and the label information of the foreign matters in the step (b) and preprocessing the picture data and the label information comprises the following steps: firstly, increasing noise and color disturbance to a picture, uniformly cutting and scaling the picture into 416x416 resolution, and converting a color channel into an RGB form; and coding the marking information of the foreign matters into real frame data, and generating a binary data file after preprocessing for training a rotated-yolov5 neural network.
5. The rotated-yolov 5-based elevator foreign matter detection method according to claim 2, wherein the step (c) comprises: after the forward reasoning calculation, 3 feature maps are obtained, namely high resolution, medium resolution and low resolution in sequence, the corresponding resolution sizes are respectively 52x52, 26x26 and 13x13, pixels on each feature map are provided with (4+1+180+ n) x3 channels, and the data of the first 4 channels respectively comprise the central coordinates x and y and the width and height w and h of the foreign matter in sequence; the 5 th channel represents probability information of whether foreign matters exist, the following 180 channels respectively represent the possibility that the foreign matters rotate in 0-179 angles in sequence, and the last n channels represent the probability that the foreign matters belong to the categories 0-n-1.
6. The rotated-yolov 5-based elevator foreign matter detection method according to claim 2, wherein the step (d) comprises:
step S41: the loss value between the predicted frame data and the real frame data of the foreign object is calculated by the loss function of the rotated-yolov5 neural network in such a manner that
The four items on the right of the equal sign are the positioning loss of the target, the confidence loss of the target, the rotation angle loss of the target and the classification loss of the target in turn,
it is calculated that the position loss between the positive sample box and the prediction box,
in order to make the model learn to distinguish whether there is a foreign object at the position,
using softmThe ax cross-entropy loss function is,
using a two-class cross entropy loss function;
step S42: using a random gradient descent method pair
Is globally optimized, and a weight matrix of the rotated-yolov5 neural network is updated through back propagation calculation;
step S43: and adding 1 to the current training round number to finish the training of the current round.
7. The rotated-yolov 5-based elevator foreign matter detection method according to claim 2, wherein the step (f) of determining whether the condition of the number of training rounds is satisfied is: whether the current number of training rounds is equal to the total number of training rounds.
8. An elevator foreign matter detection system based on rotated-yolov5, characterized by comprising:
the acquisition module is used for acquiring and acquiring video information, image information and display information in the elevator;
a memory for storing a program of the rotated-yolov5 neural network-based elevator foreign matter detection method according to any one of claims 1 to 7;
a processor, a program in the memory capable of being loaded and executed by the processor and implementing the method for detecting the foreign matter in the elevator based on the rotated-yolov5 neural network as claimed in any one of claims 1 to 7.
9. A computer storage medium characterized by comprising a program that is loadable by a processor and executable to implement the rotated-yolov5 neural network-based elevator foreign matter detection method according to any one of claims 1 to 7.
10. An elevator based on rotated-yolov5, comprising a processor for loading and executing a set of instructions; and a storage medium as claimed in claim 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110696320.5A CN113435296A (en) | 2021-06-23 | 2021-06-23 | Method, system, storage medium and elevator for detecting foreign matters based on rotated-yolov5 |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110696320.5A CN113435296A (en) | 2021-06-23 | 2021-06-23 | Method, system, storage medium and elevator for detecting foreign matters based on rotated-yolov5 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113435296A true CN113435296A (en) | 2021-09-24 |
Family
ID=77753465
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110696320.5A Pending CN113435296A (en) | 2021-06-23 | 2021-06-23 | Method, system, storage medium and elevator for detecting foreign matters based on rotated-yolov5 |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113435296A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115049948A (en) * | 2022-08-15 | 2022-09-13 | 深圳市万物云科技有限公司 | Unmanned aerial vehicle inspection method and device based on neural network model and related equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014001028A (en) * | 2012-06-15 | 2014-01-09 | Hitachi Building Systems Co Ltd | Sill groove foreign matter detection device for elevator |
| CN104933436A (en) * | 2014-03-19 | 2015-09-23 | 通用汽车环球科技运作有限责任公司 | Vision-based multi-camera factory monitoring including dynamic integrity grading |
| CN110540121A (en) * | 2019-09-09 | 2019-12-06 | 广州广日电梯工业有限公司 | system and method for identifying foreign matters in sliding groove of elevator sill |
| CN111444809A (en) * | 2020-03-23 | 2020-07-24 | 华南理工大学 | Power transmission line abnormal target detection method based on improved YO L Ov3 |
| CN112850396A (en) * | 2019-11-28 | 2021-05-28 | 宁波微科光电股份有限公司 | Elevator foreign matter detection method and system, computer storage medium and elevator |
-
2021
- 2021-06-23 CN CN202110696320.5A patent/CN113435296A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014001028A (en) * | 2012-06-15 | 2014-01-09 | Hitachi Building Systems Co Ltd | Sill groove foreign matter detection device for elevator |
| CN104933436A (en) * | 2014-03-19 | 2015-09-23 | 通用汽车环球科技运作有限责任公司 | Vision-based multi-camera factory monitoring including dynamic integrity grading |
| CN110540121A (en) * | 2019-09-09 | 2019-12-06 | 广州广日电梯工业有限公司 | system and method for identifying foreign matters in sliding groove of elevator sill |
| CN112850396A (en) * | 2019-11-28 | 2021-05-28 | 宁波微科光电股份有限公司 | Elevator foreign matter detection method and system, computer storage medium and elevator |
| CN111444809A (en) * | 2020-03-23 | 2020-07-24 | 华南理工大学 | Power transmission line abnormal target detection method based on improved YO L Ov3 |
Non-Patent Citations (1)
| Title |
|---|
| 卢琨: "监控场景下的行人检测研究", 《中国优秀硕士学位论文全文数据库（电子期刊）》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115049948A (en) * | 2022-08-15 | 2022-09-13 | 深圳市万物云科技有限公司 | Unmanned aerial vehicle inspection method and device based on neural network model and related equipment |
| CN115049948B (en) * | 2022-08-15 | 2022-11-22 | 深圳市万物云科技有限公司 | Unmanned aerial vehicle inspection method and device based on neural network model and related equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10699151B2 (en) | System and method for performing saliency detection using deep active contours | |
| CN108399386B (en) | Method and device for extracting information in pie chart | |
| Xie et al. | Multilevel cloud detection in remote sensing images based on deep learning | |
| WO2022100470A1 (en) | Systems and methods for target detection | |
| CN106951870B (en) | Intelligent detection and early warning method for active visual attention of significant events of surveillance video | |
| CN106327488B (en) | Self-adaptive foreground detection method and detection device thereof | |
| AU2020272936B2 (en) | Methods and systems for crack detection using a fully convolutional network | |
| Xiang et al. | Moving object detection and shadow removing under changing illumination condition | |
| CN110349188B (en) | Multi-target tracking method, device and storage medium based on TSK fuzzy model | |
| CN109685830B (en) | Target tracking method, device and equipment and computer storage medium | |
| Fier et al. | Automatic fish counting system for noisy deep-sea videos | |
| CN111242026A (en) | Remote sensing image target detection method based on spatial hierarchy perception module and metric learning | |
| US20170053172A1 (en) | Image processing apparatus, and image processing method | |
| CN111986156A (en) | Axe-shaped sharp tool detection method, system, device and storage medium | |
| CN113435296A (en) | Method, system, storage medium and elevator for detecting foreign matters based on rotated-yolov5 | |
| CN113409271B (en) | Method, device and equipment for detecting oil stain on lens | |
| Costa et al. | Genetic adaptation of segmentation parameters | |
| CN110929731B (en) | Medical image processing method and device based on pathfinder intelligent search algorithm | |
| CN111814653A (en) | Method, device, equipment and storage medium for detecting abnormal behaviors in video | |
| CN104778468B (en) | Image processing apparatus, image processing method, and monitoring device | |
| CN113869163B (en) | Target tracking method and device, electronic equipment and storage medium | |
| CN113255549B (en) | Intelligent recognition method and system for behavior state of wolf-swarm hunting | |
| CN115661860A (en) | Method, device and system for dog behavior and action recognition technology and storage medium | |
| CN117173075A (en) | Medical image detection method and related equipment | |
| CN112686162A (en) | Method, device, equipment and storage medium for detecting clean state of warehouse environment |
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: 20210924 |
|
| RJ01 | Rejection of invention patent application after publication |