CN113158962A - Swimming pool drowning detection method based on YOLOv4 - Google Patents
Swimming pool drowning detection method based on YOLOv4 Download PDFInfo
- Publication number
- CN113158962A CN113158962A CN202110488324.4A CN202110488324A CN113158962A CN 113158962 A CN113158962 A CN 113158962A CN 202110488324 A CN202110488324 A CN 202110488324A CN 113158962 A CN113158962 A CN 113158962A
- Authority
- CN
- China
- Prior art keywords
- swimming pool
- model
- swimming
- drowning
- detection
- 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
- 230000009182 swimming Effects 0.000 title claims abstract description 71
- 238000001514 detection method Methods 0.000 title claims abstract description 49
- 206010013647 Drowning Diseases 0.000 title claims abstract description 41
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000003064 k means clustering Methods 0.000 claims abstract description 4
- 230000009189 diving Effects 0.000 claims description 8
- 238000012937 correction Methods 0.000 claims description 6
- 230000006399 behavior Effects 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 2
- 238000000034 method Methods 0.000 abstract description 3
- 238000012544 monitoring process Methods 0.000 abstract description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 7
- 238000002372 labelling Methods 0.000 description 5
- 238000012795 verification Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000037007 arousal Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000009429 distress Effects 0.000 description 1
- 208000002173 dizziness Diseases 0.000 description 1
- 230000036651 mood Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000004962 physiological condition Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a swimming pool drowning detection method based on YOLOv4, which further restricts the detection object type by adding a swimming pool determination line to a YOLOv4 detection model, thereby improving the detection precision. An underwater camera arranged in the swimming pool collects images of all swimming people in the swimming pool and labels the images to obtain a special swimmer database; obtaining the sizes of prior frames by a swimmer database by adopting a K-means clustering algorithm, and clustering the prior frames with 9 sizes according to different scales; constructing a YOLOv4 network model, carrying out iterative training until a loss function is converged, and storing the trained network model; adding the trained network model into a swimming pool judging line model, so that the model can judge the recognized target type again; outputting a YOLOv 4-swimming pool determination line detection model meeting the requirements; and carrying out target detection on the underwater sequence images by using a YOLOv 4-swimming pool determination line detection model. The method has the advantages of high detection speed, high accuracy and low drowning detection false alarm rate, and meets the requirement of real-time monitoring.
Description
Technical Field
The invention relates to a computer vision artificial intelligence technology, in particular to a swimming pool drowning detection method based on YOLOv 4.
Background
With the arousal of national sports and body-building consciousness, swimming is one of the favorite sports of people. But at the same time, the drowning accident of the swimming pool presents a high emergence situation, and the lifeguard is difficult to detect and react at the first time once the drowning accident occurs. Therefore, the research on how to discover and help drowners in time has great practical value.
For the safety issues of public artificial natatoriums (pools), some of them use traditional human supervision mode. Each swimming pool is kept close to the water surface by 2-4 lifeguards to prevent and rescue drowning people. However, the supervision mode has poor reliability, weak ability of lifeguards to handle emergencies, slow rescue speed of drowners, and difficult effective guarantee of public safety in the venue. In addition, due to the limitation of human physiological conditions, the lifeguard is difficult to keep high concentration of long-time attention, and the lifeguard can also be in a dizzy state when watching the water surface of the swimming pool for a long time. For the season of swimming, the swimming pool is often large in number and noisy in environment, the distress action of a drowning person on the water surface is difficult to draw the attention of a rescuer in time, and the drowning person is more difficult to be noticed once sinking underwater. The other part venue uses the camera to control inside and under water of natatorium, but mostly is artifical the supervision, not only consumes the manpower, and the monitoring effect can receive factors influence such as people's mood and fatigue degree in addition, appears leaking easily and examines the phenomenon, and is not obvious to drowning person's prevention and succour effect.
Disclosure of Invention
The invention provides a swimming pool drowning detection method based on a YOLOv4 algorithm, which is characterized by further restricting the detection target type by adding a swimming pool judgment line model into a YOLOv4 detection model to improve the detection precision and reduce the detection false alarm rate of drowners and comprises the following steps:
s1: images of all swimming people in the swimming pool are collected and marked through an underwater camera arranged in the swimming pool, and a special swimmer database is obtained.
S2: the swimmer database obtains the sizes of the prior frames by adopting a K-means clustering algorithm, and the prior frames with 9 sizes are clustered according to different scales.
S3: setting network model parameters in a configuration file of a YOLOv4 network model, putting a YOLOv4 network structure into a computer in a configured environment, carrying out iterative training on a pre-training model by using a training set until a loss function is converged, and storing the trained network model.
S4: and adding the trained network detection model into a swimming pool judgment line model, and judging the recognized target type by the model again.
S5: and outputting a satisfactory YOLOv 4-swimming pool decision line detection model.
S6: and (5) carrying out target detection on the sequence image by using the YOLOv 4-swimming pool determination line detection model which meets the requirements in the step S5, outputting a detection result and alarming for drowning.
Further, the swimmer database described in step S1 is a request for: the underwater camera collects videos of swimming crowds in the swimming pool, converts the collected videos into picture formats, and utilizes Labelimg software to mark the videos into VOC formats to obtain a swimmer database, wherein actions of free swimming, backstroke, breaststroke, butterfly swimming, diving, water treading and the like of a swimmer are marked as normal swimming (swiming), and actions of diving, struggling, drowning and the like are marked as drowning (sinking).
Further, the configuration file in step S3 is required as follows:
s3-1: when the size of an input image is 416x416, the parameter random is 0, the batch _ size parameter is 16, the batch subdivision is 16, the iteration number is 6000, the initial learning rate is 0.001, and the type of the detected object is 2, the user uses a cpu training model and at least needs 2GB of memory; the model was trained using a single GPU, requiring at least 4 GB.
S3-2: when the size of an input image is 608x608, the parameter random is 0, the batch _ size parameter is 16, the batch subdivision is 16, the iteration number is 6000, the initial learning rate is 0.001, and the type of the detected object is 2, the user uses a cpu training model and at least needs 3GB of memory; the model is trained using a single GPU, requiring at least 6 GB.
Further, the swimming pool determination line model in step S4 specifically includes the following steps: dividing the picture according to the position of the underwater footrest of the swimming pool in the picture, wherein the position of the footrest in the picture is a determination line, dividing the picture into an upper part and a lower part according to the divided determination line, and the upper part of the determination line is a normal swimming area and the lower part of the determination line is a drowning area. When the detected target center position is located in a normal swimming area, if the target label is drowning (sinking), the label is corrected to be normal swimming (swiming), and if the target label is normal swimming (swiming), the correction is not needed; when the detected target center position is located in the drowning area, if the target tag is normal swimming (swimming), the tag is corrected to be drowned (descending), and if the target tag is drowned (descending), the correction is not needed.
Further, the step S5 is performed to evaluate the performance of the YOLOv 4-pool determination line detection model, so that the mAp0.5 reaches 92% or more.
Further, the alarm request in step S6 is: and a drowning alarm signal appears within 5 seconds continuously.
Drawings
FIG. 1 is a flow chart of a YOLOv 4-swimming pool determination line detection method.
Fig. 2 is a network structure diagram of YOLOv 4.
Fig. 3 is a graph of a swimming pool determination line.
Figure 4 is a view of underwater target detection of the swimming pool.
Detailed Description
The invention is described in detail below with reference to fig. 1 and the specific examples. The present embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation manner and a specific operation process are given, but the scope of the present invention is not limited to the following embodiments.
A method for detecting drowning in a swimming pool based on YOLOv4 algorithm, as shown in fig. 1 to 4, includes:
s1-1: acquiring an image:
the underwater camera arranged in the swimming pool is used for collecting data of all swimming people in the swimming pool, the frame rate of the video is about 30 pictures in 1 second generally, the frame taking frequency is set to be 1 second and 3 frames are taken, and the total number of samples is at least more than 20 videos, namely the number of samples is about 15000. And naming the acquired images according to the format of the Pascal VOC data set, simultaneously creating three folders named as options, ImageSets and JPEGImages, and creating a new Main folder under the ImageSets. And copying the collected data set picture to a JPEGImages directory.
Labeling of S1-2 image:
and labeling the picture by using a labeling tool labellimg to generate an xml labeling file. Randomly numbering about 15000 video frame pictures, and coding a reasonable sequence number for the pictures, such as 000001-000999; and labeling data by using labelimg software, wherein each picture name corresponds to an xml label file with a corresponding name, such as a picture 000001.jpg, and the label file is 000001. xml. The range of labels includes: the position of the image, the name of the image (such as 000001.jpg), the width and height of the image, the dimension of the image, the name of the annotated object and the coordinate value of bbox; the swimmer behaviors comprise free swimming, backstroke, breaststroke, butterfly stroke, diving, treading, diving, struggling and drowning, wherein the swimmer free swimming, backstroke, breaststroke, butterfly stroke, diving, treading and other behaviors are marked as normal swimming (swimming), and the diving, struggling, drowning and other behaviors are marked as drowning (sinking). The image data is divided into a training data set and a verification data set, wherein the training data set accounts for 80% and the verification data set accounts for 20%.
S2: and (3) obtaining the size of a prior frame by adopting K-means clustering on the picture data set:
3 prior boxes are set for each downsampling scale, and 9 sizes of prior boxes are clustered in total. The 9 prior boxes of the data set are: (68x118), (93x137), (114x84), (139x220), (158x135), (173x72), (213x300), (237x104), (291x 159).
TABLE 1 eigenmap multiscale assignment
S3: YOLOv4 model training step: training learning is performed on the training data set by using Yolov4, and the operation is as follows:
the YOLOv4 target detection network takes CSPDarknet53 as a backbone and comprises 5 CSP modules, the size of a convolution kernel in front of each module is 3 multiplied by 3, the step length is 2, and the network learning capability can be further enhanced; a path aggregation network (PANet) is used as a neck, a Spatial Pyramid Pool (SPP) additional module is added, and a 1 × 1, 5 × 5, 9 × 9 and 13 × 13 maximum pooling mode is adopted, so that a sensing area can be increased, and more important context characteristics can be separated; the YOLOv3 detection head is used as the head. And finally, outputting a training result through the full connection layer, wherein the training result comprises a frame regression coordinate, a target classification result and a confidence degree.
S3-1: when the size of an input image is 416 × 416, the parameter random is 0, the batch _ size parameter is 16, the batch subdivision is 16, the iteration times are 6000, the initial learning rate is 0.001, and the type of the detected object is 2, the user uses a cpu training model and at least needs 2GB of memory; the model was trained using a single GPU, requiring at least 4 GB.
S3-2: when the size of an input image is 608 × 608, the parameter random is 0, the batch _ size parameter is 16, the batch subdivision is 16, the iteration number is 6000, the initial learning rate is 0.001, and the detected object type is 2, a user uses a cpu training model and at least needs 3GB of memory; the model is trained using a single GPU, requiring at least 6 GB.
S4: the swimming pool judges line model specifically includes:
s4-1: dividing the picture according to the position of the underwater footrest of the swimming pool in the picture, wherein the position of the footrest in the picture is a determination line, dividing the picture into an upper part and a lower part according to the divided determination line, and the upper part of the determination line is a normal swimming area and the lower part of the determination line is a drowning area. When the detected target center position is located in a normal swimming area, if the target label is drowning (sinking), the label is corrected to be normal swimming (swiming), and if the target label is normal swimming (swiming), the correction is not needed; when the center position of the detected target is located in a drowning area, if the target label is swamming, the label is corrected to be drowning (descending), and if the target label is drowning (descending), the correction is not needed.
S4-2: for calculation, a plane rectangular coordinate system is established, the origin of coordinates is located at the upper left corner of the image, and the coordinate values are continuously increased along with the edge of the image towards the right and vertically downwards. Further, the coordinates (x, y) in the coordinate system are related to the picture resolution, x ∈ [0,1920], y ∈ [0,1080 ].
S4-3: judging the line expression as a piecewise function, specifically:
wherein A is1-An,B1-Bn is the coefficient of the equation, R1Rn is the domain of x, disjoint and E [0,1920]。
S4-4: and (3) judging the central coordinates (x, y) of the detection frame output by the YOLOv4 model, wherein the detection target label is swift if the central coordinates (x, y) are above the judgment line, and the detection target label is descending if the central coordinates (x, y) are below the judgment line.
Step S5: YOLOv 4-swimming pool decision line model verification step:
and verifying the YOLOv 4-swimming pool decision line detection model through a verification data set to obtain a model score, evaluating the model, and screening out the model with the optimal prediction performance through model evaluation.
The average Precision average (mAP) is an important index for measuring the target detection efficiency and is determined by Precision (Precision) and Recall (Recall). A curve with Recall as a horizontal axis and Precision as a vertical axis is called a P-R curve for short, the area below the P-R curve is marked as a Precision mean value, the mean value of the Precision mean values of all target categories is an mAP value, and the larger the value is, the better the effect of the neural network model is.
Their calculation formulas are respectively as follows:
in the formula:
TP (true Positives) -positive class judged as positive class;
FP (false positives) -a negative class that is judged to be a positive class;
FN (false negatives) -a positive class that is judged to be a negative class.
AP: area under PR curve (PR curve: Precision-Recall curve) measures whether detection is good or bad for one class, and mAP measures whether detection is good or bad for a plurality of classes.
TABLE 2 precision values of the test results
And a target detection step: utilize the YOLOv 4-swimming pool determination line detection model that finally obtains to monitor the scene under water of swimming pool, as the input of the model through the camera under water, when discerning swimmer and drowned person, when appearing drowning alarm signal in 5 seconds in succession, propelling movement alarm information.
Claims (6)
1. A swimming pool drowning detection method based on YOLOv4 comprises the following steps:
s1: collecting images of all swimming people in the swimming pool through an underwater camera arranged in the swimming pool and marking the images to obtain a swimmer database;
s2: obtaining the sizes of prior frames by a swimmer database by adopting a K-means clustering algorithm, and clustering the prior frames with 9 sizes according to different scales;
s3: setting network model parameters in a configuration file of a YOLOv4 network model, putting a YOLOv4 network structure into a computer in a configured environment, carrying out iterative training on a pre-training model by using a training set until a loss function is converged, and storing a trained network detection model;
s4: adding the trained network detection model into a swimming pool judgment line model, and judging the recognized target type again by the model;
s5: outputting a YOLOv 4-swimming pool determination line detection model meeting the requirements;
s6: and (5) carrying out target detection on the sequence image by using the YOLOv 4-swimming pool determination line detection model which meets the requirements in the step S5, outputting a detection result and alarming for drowning.
2. The swimming pool drowning detection method based on YOLOv4 as claimed in claim 1, wherein in the swimmer database of S1, the underwater camera collects the video of the swimming population in the swimming pool, converts the collected video into a picture format, and marks the collected video into a VOC format by using Labelimg software to obtain the swimmer database, wherein the swimmer free-swimming, backstroke, breaststroke, butterfly stroke, diving and treading behaviors are marked as normal swimming, and the diving, struggling and drowning behaviors are marked as drowning.
3. The swimming pool drowning detection method based on YOLOv4 as claimed in claim 1, wherein the configuration file in S3 is as follows:
s3-1: when the size of an input image is 416x416, the parameter random is 0, the batch _ size parameter is 16, the batch subdivision is 16, the iteration number is 6000, the initial learning rate is 0.001, and the type of the detected object is 2, the user uses a cpu training model and at least needs 2GB of memory; training the model by using a single GPU, wherein at least 4GB is required;
s3-2: when the size of an input image is 608x608, the parameter random is 0, the batch _ size parameter is 16, the batch subdivision is 16, the iteration number is 6000, the initial learning rate is 0.001, and the type of the detected object is 2, the user uses a cpu training model and at least needs 3GB of memory; the model is trained using a single GPU, requiring at least 6 GB.
4. The swimming pool drowning detection method based on YOLOv4 as claimed in claim 1, wherein the swimming pool decision line model in S4 specifically includes: dividing the picture according to the position of the underwater footrest of the swimming pool in the picture, wherein the position of the footrest in the picture is a determination line, dividing the picture into an upper part and a lower part according to the divided determination line, and the upper part of the determination line is a normal swimming area and the lower part of the determination line is a drowning area; when the central position of the detected target is positioned in the normal swimming area, if the target label is drowned, the label is corrected to be normal swimming, and if the target label is normal swimming, the correction is not needed; when the center position of the detected target is located in a drowning area, if the target label is normal swimming, the label is corrected to be drowned, and if the target label is drowned, the correction is not needed.
5. The drowning detection method of swimming pool based on YOLOv4 of claim 1, wherein the meeting requirement in S5 is that the performance evaluation of YOLOv 4-swimming pool determination line detection model, mAP @0.5 reaches 92% or more.
6. The swimming pool drowning detection method based on YOLOv4 as claimed in claim 1, wherein the alarm request in S6 is: and a drowning alarm signal appears within 5 seconds continuously.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110488324.4A CN113158962A (en) | 2021-05-06 | 2021-05-06 | Swimming pool drowning detection method based on YOLOv4 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110488324.4A CN113158962A (en) | 2021-05-06 | 2021-05-06 | Swimming pool drowning detection method based on YOLOv4 |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113158962A true CN113158962A (en) | 2021-07-23 |
Family
ID=76871495
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110488324.4A Pending CN113158962A (en) | 2021-05-06 | 2021-05-06 | Swimming pool drowning detection method based on YOLOv4 |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113158962A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114170317A (en) * | 2022-01-10 | 2022-03-11 | 杭州巨岩欣成科技有限公司 | Method and device for judging position of drowning-proof human head of swimming pool, computer equipment and storage medium thereof |
CN114359967A (en) * | 2022-01-10 | 2022-04-15 | 杭州巨岩欣成科技有限公司 | Method and device for detecting drowning-preventing human body target in swimming pool, computer equipment and storage medium |
CN114560059A (en) * | 2022-03-08 | 2022-05-31 | 三峡大学 | Underwater lifesaving robot and rescuing method |
CN114735165A (en) * | 2022-03-08 | 2022-07-12 | 三峡大学 | Intelligent underwater lifesaving system and drowning detection and rescue method |
CN115994911A (en) * | 2023-03-24 | 2023-04-21 | 山东上水环境科技集团有限公司 | Natatorium target detection method based on multi-mode visual information fusion |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200057935A1 (en) * | 2017-03-23 | 2020-02-20 | Peking University Shenzhen Graduate School | Video action detection method based on convolutional neural network |
CN111709489A (en) * | 2020-06-24 | 2020-09-25 | 广西师范大学 | Citrus identification method based on improved YOLOv4 |
CN111985451A (en) * | 2020-09-04 | 2020-11-24 | 南京航空航天大学 | Unmanned aerial vehicle scene detection method based on YOLOv4 |
CN112052826A (en) * | 2020-09-18 | 2020-12-08 | 广州瀚信通信科技股份有限公司 | Intelligent enforcement multi-scale target detection method, device and system based on YOLOv4 algorithm and storage medium |
-
2021
- 2021-05-06 CN CN202110488324.4A patent/CN113158962A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200057935A1 (en) * | 2017-03-23 | 2020-02-20 | Peking University Shenzhen Graduate School | Video action detection method based on convolutional neural network |
CN111709489A (en) * | 2020-06-24 | 2020-09-25 | 广西师范大学 | Citrus identification method based on improved YOLOv4 |
CN111985451A (en) * | 2020-09-04 | 2020-11-24 | 南京航空航天大学 | Unmanned aerial vehicle scene detection method based on YOLOv4 |
CN112052826A (en) * | 2020-09-18 | 2020-12-08 | 广州瀚信通信科技股份有限公司 | Intelligent enforcement multi-scale target detection method, device and system based on YOLOv4 algorithm and storage medium |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114170317A (en) * | 2022-01-10 | 2022-03-11 | 杭州巨岩欣成科技有限公司 | Method and device for judging position of drowning-proof human head of swimming pool, computer equipment and storage medium thereof |
CN114359967A (en) * | 2022-01-10 | 2022-04-15 | 杭州巨岩欣成科技有限公司 | Method and device for detecting drowning-preventing human body target in swimming pool, computer equipment and storage medium |
CN114170317B (en) * | 2022-01-10 | 2024-04-05 | 杭州巨岩欣成科技有限公司 | Swimming pool drowning prevention head position judging method and device and computer equipment |
CN114560059A (en) * | 2022-03-08 | 2022-05-31 | 三峡大学 | Underwater lifesaving robot and rescuing method |
CN114735165A (en) * | 2022-03-08 | 2022-07-12 | 三峡大学 | Intelligent underwater lifesaving system and drowning detection and rescue method |
CN115994911A (en) * | 2023-03-24 | 2023-04-21 | 山东上水环境科技集团有限公司 | Natatorium target detection method based on multi-mode visual information fusion |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113158962A (en) | Swimming pool drowning detection method based on YOLOv4 | |
CN111723748A (en) | Infrared remote sensing image ship detection method | |
Long et al. | Safety helmet wearing detection based on deep learning | |
CN110569772B (en) | Method for detecting state of personnel in swimming pool | |
CN110427813A (en) | Pedestrian's recognition methods again based on the twin production confrontation network that posture instructs pedestrian image to generate | |
CN111091072A (en) | YOLOv 3-based flame and dense smoke detection method | |
CN112861635B (en) | Fire disaster and smoke real-time detection method based on deep learning | |
CN111814638B (en) | Security scene flame detection method based on deep learning | |
CN115661943B (en) | Fall detection method based on lightweight attitude assessment network | |
CN110598693A (en) | Ship plate identification method based on fast-RCNN | |
CN111626162A (en) | Overwater rescue system based on space-time big data analysis and drowning warning situation prediction method | |
CN106529398A (en) | Quick and accurate face detection method based on cascade structure | |
CN106981063A (en) | A kind of grid equipment state monitoring apparatus based on deep learning | |
CN111488850B (en) | Neural network-based old people falling detection method | |
CN114283469A (en) | Lightweight target detection method and system based on improved YOLOv4-tiny | |
US8094971B2 (en) | Method and system for automatically determining the orientation of a digital image | |
CN113139481B (en) | Classroom people counting method based on yolov3 | |
CN109886086A (en) | Pedestrian detection method based on HOG feature and Linear SVM cascade classifier | |
CN115527234A (en) | Infrared image cage dead chicken identification method based on improved YOLOv5 model | |
Lei et al. | Drowning behavior detection in swimming pool based on deep learning | |
CN113822185A (en) | Method for detecting daily behavior of group health pigs | |
CN111507353B (en) | Chinese field detection method and system based on character recognition | |
CN115546904A (en) | Method for tracking and identifying danger of fallen personnel based on target detection time sequence | |
CN110321775A (en) | A kind of drowning man's autonomous classification method waterborne based on multi-rotor unmanned aerial vehicle | |
CN113139497A (en) | System and method for identifying water surface object and application based on 5G MEC |
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 |