CN115331178A - Natural resource monitoring method and system based on iron tower video - Google Patents
Natural resource monitoring method and system based on iron tower video Download PDFInfo
- Publication number
- CN115331178A CN115331178A CN202211196051.7A CN202211196051A CN115331178A CN 115331178 A CN115331178 A CN 115331178A CN 202211196051 A CN202211196051 A CN 202211196051A CN 115331178 A CN115331178 A CN 115331178A
- Authority
- CN
- China
- Prior art keywords
- module
- target
- target detection
- detection result
- natural resource
- 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
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
- G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
- G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/176—Urban or other man-made structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a natural resource monitoring method and a system based on an iron tower video, wherein the method comprises the following steps: training the improved YOLOv5s network model to obtain an optimal model, acquiring a video of the iron tower in real time, determining a target to be detected, selecting iron tower video images of preset points at a time point T1 and a time point T2, detecting the target to be detected by using the optimal model, and respectively obtaining a first target detection result and a second target detection result; and carrying out automatic difference processing on the first target detection result and the second target detection result to obtain newly increased and/or decreased detection target results. The method rapidly discovers suspected illegal/illegal occupation/natural resource destruction behaviors of the iron tower video by using the functions of target extraction and change detection based on the deep learning model, and can greatly perfect and support the working mechanism of early discovery, early prevention and strict check of the illegal/illegal occupation/natural resource destruction behaviors.
Description
Technical Field
The invention belongs to the field of natural resource monitoring and supervision, and particularly relates to a natural resource monitoring method and system based on an iron tower video.
Background
In recent years, the phenomenon of illegal/illegal abuse of natural resources is prominent, and various problems of unreasonable occupation or damage of natural resources are increasingly serious, such as illegal construction of urban development, illegal occupation of cultivated land, protected land and the like, and the actual problems of ecological imbalance, natural resource disasters, grain crisis and the like are caused if the cultivated land is damaged or the forest is destroyed if the urban development is illegal, or the like. At present, an intelligent monitoring method for natural resources is mostly based on remote sensing satellite image data, but the method has the problems of low monitoring frequency, low monitoring granularity, easy influence of rainy weather and the like.
The iron tower video is a near-ground and real-time ground monitoring means, can effectively supplement the defects of satellite remote sensing monitoring, and realizes the monitoring of natural resources with high frequency and fine granularity. The range and angle of a view field of the iron tower camera and the distance of an observed target can be controlled by the cradle head, but along with the change of the pitching angle and the focal length zoom factor of the iron tower camera, the target object to be detected has the characteristic of spanning multiple scale ranges in a video, and particularly when the pitching angle is small (the sight line is far) and the zoom scale is small, the target object is often small, so that the difficulty is increased for target detection work.
Disclosure of Invention
In order to solve the problems, the invention provides an intelligent natural resource monitoring method based on an iron tower video, which is used for rapidly finding suspected illegal/illegal natural resource occupation/damage behaviors of the iron tower video by using small target extraction and change detection functions based on an improved YOLOv5s network model. The method specifically comprises the following steps:
training an improved YOLOv5s network model, firstly, improving a Mosaic image enhancement method in an Input module of the YOLOv5s network model into a Mosaic-9 enhancement mode, secondly, embedding an SKNet attention mechanism module behind a Backbone module of the YOLOv5s network model to form an improved YOLOv5s network model, and finally, sending a training set sample into the improved YOLOv5s network model for training to obtain an optimal model;
acquiring an iron tower video in real time, and determining a target to be detected;
selecting an iron tower video image with preset point positions at the time point T1, detecting the target to be detected by using the optimal model to obtain a first target detection result, and storing the first target detection result in a warehouse;
selecting an iron tower video image of a preset point position at a time point T2, and detecting the target to be detected by using the optimal model to obtain a second target detection result;
and carrying out automatic difference processing on the first target detection result and the second target detection result to obtain newly increased and/or decreased detection targets, and generating detection results of suspected violation, illegal occupation and natural resource damage behaviors.
Further, the improved YOLOv5s network model comprises an Input module, a backhaul module, an SKNet attention mechanism module, a Neck module and a head module.
Further, the method for enhancing the Mosaic image in the Input module of the YOLOv5s network model is improved to a Mosaic-9 enhancement mode, and 9 randomly selected detection targets are randomly cut and spliced to form new sample data;
further, an SKNet attention mechanism module is embedded behind a YOLOv5s network model Backbone module, the Backbone module comprises a Focus module, a CSP module, an SPPF module and an SKNet module, and the CSP module is used for extracting main information in an input sample; the SPPF modules are connected in series through maximum pooling and concat thereof, so that the receptive field is increased; and the SKNet module performs importance weighting on the characteristic channels, so that the small target detection precision is improved.
Further, the SKNet attention mechanism module is composed of three parts: the Split part is obtained by convolving the original characteristic diagram by convolution kernel parts with different sizes
(ii) a The Fuse part integrates information from all branches by using a gate, and the characteristics of the two parts are summed to obtain
(ii) a The Select part uses two weight matrix pairs of a and b
And performing weighting operation, and adding to obtain an output V.
Furthermore, the training set sample comprises an engineering vehicle, a tower crane, a pile driver and a manual pile pushing and earth surface target excavating device.
The invention also provides a natural resource monitoring system based on the iron tower video, which comprises a model training module, a video acquisition module, a first target detection module, a second target detection module and an execution change detection module.
The model training module is used for training an improved YOLOv5s network model, firstly, a Mosaic image enhancement method in an Input module of the YOLOv5s network model is improved into a Mosaic-9 enhancement mode, secondly, an SKNet attention mechanism module is embedded behind a Backbone module of the YOLOv5s network model to form an improved YOLOv5s network model, and finally, a training set sample is sent into the advanced YOLOv5s network model for training to obtain an optimal model;
the video acquisition module acquires a tower video in real time and determines a target to be detected;
the first target detection module selects an iron tower video image with preset point positions at the time point T1, detects the target to be detected by using the optimal model to obtain a first target detection result, and stores the first target detection result in a warehouse;
the second target detection module selects an iron tower video image of a preset point position at a time point T2, and detects the target to be detected by using the optimal model to obtain a second target detection result;
and the execution change detection module performs automatic difference processing on the first target detection result and the second target detection result to obtain newly increased and/or decreased detection targets and generates detection results of suspected violation, illegal occupation and natural resource damage behaviors.
The invention has the capability of detecting the change of natural resources based on videos, the used improved YOLOv5s network model greatly enriches the sample diversity and enhances the detection of small targets through the Mosaic-9 in the Input module, and simultaneously, after the backsbone part of the YOLOv5s network model is extracted, the SKNet attention mechanism module is embedded, so that the extracted features can be weighted in importance, the detection precision of the targets is improved, the detection and identification performance is improved, suspected behaviors of violation, illegal occupation, natural resource damage and the like of the iron tower videos can be rapidly discovered, and the working mechanism of 'early discovery, early inhibition and strict investigation' of the behaviors of violation/illegal occupation/damage of natural resources can be greatly perfected and supported.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the structures shown in the drawings without creative efforts.
Fig. 1 is a flow chart of an intelligent natural resource monitoring method based on an iron tower video.
Fig. 2 is a diagram of an improved YOLOv5s network model.
FIG. 3 is a schematic diagram of the Mosaic-9.
Fig. 4 is a diagram of an attention mechanism of SKNet.
Fig. 5a is an example of small object detection effect.
Fig. 5b is an example of large object detection effect.
Fig. 6 is a flow chart of natural resource change detection based on the optimal model.
FIG. 7a shows the results of the detection of the basal stage target.
FIG. 7b shows a second target detection result.
FIG. 7c shows the result of the change detection.
Fig. 8 is a block diagram of a natural resource monitoring system based on tower video.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
Fig. 1 is a flow chart of an intelligent natural resource monitoring method based on an iron tower video, which rapidly discovers suspected violation/illegal occupation/natural resource destruction behaviors of the iron tower video by using a target extraction and change detection function based on improved YOLOv5s network model training, and specifically includes:
in the first step, an improved YOLOv5s network model is trained.
Firstly, a Mosaic image enhancement method in an Input module of a YOLOv5s network model is improved to a Mosaic-9 enhancement mode, secondly, a SKNet attention mechanism module is embedded behind a Backbone module of the YOLOv5s network model to form an improved YOLOv5s network model, and finally, a training set sample is sent into the improved YOLOv5s network model to be trained to obtain an optimal model.
Fig. 2 is an improved YOLOv5s network model, which includes an Input module, a backhaul module, an SKNet attention mechanism module, a Neck module, and a head module, wherein the Input module uses Mosaic-9 to perform data enhancement on Input images, and randomly cuts and splices 9 randomly selected Input image data to form new sample data, so that sample diversity can be enriched, small target detection can be enhanced, and difficulty of the model in learning diversity can be reduced. And the SKNet attention mechanism module is embedded behind the backhaul module and carries out importance weighting on the characteristic channel.
As an engineering vehicle, a tower crane, a pile driver, a manual pile pushing and earth surface digging are selected as detection targets, the types are multiple and unbalanced, and the Mosaic-9 is used for data enhancement, the Mosaic-9 data enhancement method is an enhanced version of the Mosaic mode in a YOLOv5s model, as shown in figure 3, nine pictures are cut randomly and zoomed and mixed to form a picture for training, the number of data information and small target objects can be increased, and the nine pictures are calculated during normalization operation without depending on batch processing parameters, so that the calculation amount is reduced.
The SKNet attention mechanism module is an algorithm for realizing the self-adaptive receptive field size of the neuron through a nonlinear multi-core information aggregation method, and is a lightweight embedded module based on the channel domain attention. The SKNet module can weight the importance of the feature channel, so that the finally obtained feature graph focuses more on the feature channel which is useful for tasks such as detection and identification, and the module can automatically select and adjust an optimal convolution operator, so that the SKNet is embedded after feature extraction, the importance of the feature channel can be weighted, and then the feature channel is sent to a target detection layer, and the performance of detection and identification can be theoretically improved. As shown in fig. 4, SKNet is composed of Split, fuse, and Select. The Split is a process of convolving original feature maps by convolution kernels with different sizes, in order to further improve efficiency, a 5x5 kernel is replaced by a 3x3 kernel and the step pitch is 2 to obtain expansion convolution
(ii) a The basic idea of Fuse is to use gates to control the flow of information down multiple branches carrying different information scalesOne layer of neurons, so that the neurons can adaptively adjust the size of the receptive field according to the stimulation signals, and in order to achieve the purpose, the gate needs to integrate information from all branches and sum the characteristics of the two parts
In FIG. 4
For the global average pooling operation, that is, averaging along the H and W dimensions, the information about the channel is finally obtained, which is a one-dimensional vector of C × 1 × 1;
two fully-connected layers for firstly reducing and then increasing dimensions are adopted, namely information of channel dimensions is extracted, then Softmax is used for normalization, each channel corresponds to a weight coefficient at this time, and two weight matrixes a and b output represent the importance degree of the channel (channel). Where matrix b is a redundant matrix, b =1-a in the case of the two branches of fig. 4; selecting different information space scales in a self-adaptive mode by using cross-channel soft attention, calculating the weight of each convolution kernel by using a Softmax function according to the direction of a channel (channel), and using a weight matrix pair of a and b
And performing weighting operation, and adding to obtain an output V.
Training the improved YOLOv5s network model by using the training samples to obtain an optimal model, and performing target detection based on the optimal model, wherein the optimal model has a good detection effect on both large targets and small targets, as shown in FIGS. 5a and 5 b.
And step two, acquiring the tower video in real time and determining the target to be detected.
In order to realize monitoring of more natural resource scenes, engineering vehicles, tower cranes and pile drivers are selected, manual pile pushing and earth surface digging are taken as monitoring targets to be used as detection targets for monitoring behaviors of suspected violation, illegal occupation and natural resource damage, the applicability of the model for monitoring the natural resources is improved, the behaviors of violation/illegal occupation/natural resource damage of large-scale engineering using mechanical equipment such as engineering vehicles and tower cranes can be monitored, and the behaviors of violation/illegal occupation/natural resource damage of small-scale and light-weight engineering can also be monitored.
The iron tower video image refers to an image of an iron tower camera in an appointed PTZ direction and is called as a preset point position, each camera can be appointed with a plurality of preset point positions, and pictures of corresponding positions can be automatically obtained through the appointed preset point positions. The images acquired at each time point are automatically named: camera ID _ PiTiZi _ T1,
such as 430103500000001617 (u P315T10Z4 (u 20220400200. Jpg).
And thirdly, selecting the iron tower video image with the preset point position at the time point T1, detecting the target to be detected by using the optimal model to obtain a first target detection result, taking the first target detection result as a base period target detection result, and storing the first target detection result in a warehouse.
And thirdly, obtaining a video image of the iron tower at a preset point position at the time point T2, and detecting the target to be detected by using the optimal model to obtain a second target detection result.
And fourthly, automatically performing difference processing on the second target detection result and the base-period target detection result to obtain newly increased and/or decreased detection targets and obtain suspected behaviors of violation, illegal occupation and natural resource destruction.
Fig. 6 is a flow chart of natural resource change detection based on an optimal model, in which change detection is performed through automatic differencing processing based on target detection results at two time points, and suspected violation/illegal occupation/natural resource damage behaviors of an iron tower video are quickly discovered. The automatic differencing processing is to automatically match the second target detection result with the base stage target detection result by using the image name as a search condition, calculate the area intersection ratio of the second target detection result and the base stage target detection result, determine that the second target detection result changes when the area intersection ratio is less than 40%, and output newly increased and/or decreased detection targets to obtain behavior detection results of suspected violation, illegal occupation and natural resource destruction, as shown in fig. 7a, 7b and 7 c.
Fig. 8 is a block diagram of a natural resource monitoring system based on a tower video, where the system includes a model training module, a video acquisition module, a first target detection module, a second target detection module, and an execution change detection module.
The model training module is used for training an improved YOLOv5s network model, firstly, a Mosaic image enhancement method in an Input module of the YOLOv5s network model is improved into a Mosaic-9 enhancement mode, secondly, an SKNet attention mechanism module is embedded behind a Backbone module of the YOLOv5s network model to form an improved YOLOv5s network model, and finally, a training set sample is sent into the advanced YOLOv5s network model for training to obtain an optimal model;
the video acquisition module acquires a tower video in real time and determines a target to be detected;
the first target detection module selects a video image of the iron tower at a preset point at a time point T1, detects the target to be detected by using the optimal model to obtain a first target detection result, and stores the first target detection result in a warehouse;
the second target detection module selects an iron tower video image of a preset point position at a time point T2, and detects the target to be detected by using the optimal model to obtain a second target detection result;
and the execution change detection module performs automatic difference processing on the first target detection result and the second target detection result to obtain newly increased and/or decreased detection targets and generates detection results of suspected violation, illegal occupation and natural resource damage behaviors.
While embodiments in accordance with the invention have been described above, these embodiments are not intended to be exhaustive or to limit the invention to the precise embodiments described. Many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated. The invention is limited only by the claims and their full scope and equivalents.
Claims (7)
1. A natural resource monitoring method based on an iron tower video is characterized by comprising the following steps:
training an improved YOLOv5s network model, firstly, improving a Mosaic image enhancement method in an Input module of the YOLOv5s network model into a Mosaic-9 enhancement mode, secondly, embedding an SKNet attention mechanism module behind a Backbone module of the YOLOv5s network model to form an improved YOLOv5s network model, and finally, sending a training set sample into the improved YOLOv5s network model for training to obtain an optimal model;
acquiring an iron tower video in real time, and determining a target to be detected;
selecting an iron tower video image with preset point positions at the time point T1, detecting the target to be detected by using the optimal model to obtain a first target detection result, and storing the first target detection result in a warehouse;
selecting an iron tower video image of a preset point position at a time point T2, and detecting the target to be detected by using the optimal model to obtain a second target detection result;
and carrying out automatic difference processing on the first target detection result and the second target detection result to obtain newly increased and/or decreased detection targets, and generating detection results of suspected violation, illegal occupation and natural resource damage behaviors.
2. The natural resource monitoring method of claim 1, wherein the improved YOLOv5s network model comprises an Input module, a backhaul module, a SKNet attention mechanism module, a Neck module, and a head module.
3. The natural resource monitoring method of claim 1, wherein the Mosaic-9 enhancement mode is to randomly cut and splice 9 detection targets selected at random to form new sample data.
4. The natural resource monitoring method of claim 1, wherein the SKNet attention mechanism module is embedded behind a backhaul module to weight the importance of the feature channel.
5. The natural resource monitoring method of claim 1, wherein the SKNet attention mechanism module comprises three parts, namely Split, fuse and Select, and the Split part is obtained by performing convolution on an original feature map through convolution kernel parts with different sizes
(ii) a The Fuse part integrates information from all branches by using a gate, and the characteristics of the two parts are summed to obtain
(ii) a The Select part uses two weight matrix pairs of a and b
And performing weighting operation, and adding to obtain an output V.
6. A natural resource monitoring method as claimed in any one of claims 1 to 5, wherein the monitoring targets are engineering vehicles, tower cranes, pile drivers, and manual pile pushing and earth surface excavation.
7. A natural resource monitoring system based on an iron tower video is characterized by comprising a model training module, a video acquisition module, a first target detection module, a second target detection module and an execution change detection module,
the model training module is used for training an improved YOLOv5s network model, firstly, a Mosaic image enhancement method in an Input module of the YOLOv5s network model is improved into a Mosaic-9 enhancement mode, secondly, an SKNet attention mechanism module is embedded behind a Backbone module of the YOLOv5s network model to form an improved YOLOv5s network model, and finally, a training set sample is sent into the advanced YOLOv5s network model for training to obtain an optimal model;
the video acquisition module acquires a tower video in real time and determines a target to be detected;
the first target detection module selects a video image of the iron tower at a preset point at a time point T1, detects the target to be detected by using the optimal model to obtain a first target detection result, and stores the first target detection result in a warehouse;
the second target detection module selects an iron tower video image of a preset point position at a time point T2, and detects the target to be detected by using the optimal model to obtain a second target detection result;
and the execution change detection module performs automatic difference processing on the first target detection result and the second target detection result to obtain newly increased and/or decreased detection targets and generates detection results of suspected violation, illegal occupation and natural resource damage behaviors.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211196051.7A CN115331178A (en) | 2022-09-29 | 2022-09-29 | Natural resource monitoring method and system based on iron tower video |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211196051.7A CN115331178A (en) | 2022-09-29 | 2022-09-29 | Natural resource monitoring method and system based on iron tower video |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115331178A true CN115331178A (en) | 2022-11-11 |
Family
ID=83914302
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211196051.7A Pending CN115331178A (en) | 2022-09-29 | 2022-09-29 | Natural resource monitoring method and system based on iron tower video |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115331178A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117874273A (en) * | 2024-03-11 | 2024-04-12 | 湖南省第二测绘院 | Iron tower video image classification identification method and device based on geographic mapping |
| CN118134936A (en) * | 2024-05-09 | 2024-06-04 | 湖南省第二测绘院 | Method and device for detecting changes of illegal buildings through video images of iron towers |
-
2022
- 2022-09-29 CN CN202211196051.7A patent/CN115331178A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117874273A (en) * | 2024-03-11 | 2024-04-12 | 湖南省第二测绘院 | Iron tower video image classification identification method and device based on geographic mapping |
| CN117874273B (en) * | 2024-03-11 | 2024-05-10 | 湖南省第二测绘院 | Iron tower video image classification identification method and device based on geographic mapping |
| CN118134936A (en) * | 2024-05-09 | 2024-06-04 | 湖南省第二测绘院 | Method and device for detecting changes of illegal buildings through video images of iron towers |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115331178A (en) | Natural resource monitoring method and system based on iron tower video | |
| CN111461291B (en) | Long-distance pipeline inspection method based on YOLOv3 pruning network and deep learning defogging model | |
| US10621476B1 (en) | Learning method and learning device for object detector based on reconfigurable network for optimizing customers' requirements such as key performance index using target object estimating network and target object merging network, and testing method and testing device using the same | |
| CN113177183B (en) | Seawater pollution monitoring and early warning method and system based on ocean remote sensing image | |
| KR102337367B1 (en) | Learning method and learning device for object detector with hardware optimization based on cnn for detection at distance or military purpose using image concatenation, and testing method and testing device using the same | |
| US10402692B1 (en) | Learning method and learning device for fluctuation-robust object detector based on CNN using target object estimating network adaptable to customers' requirements such as key performance index, and testing device using the same | |
| CN113420819B (en) | Lightweight underwater target detection method based on CenterNet | |
| CN111079604A (en) | Method for quickly detecting tiny target facing large-scale remote sensing image | |
| CN110633633B (en) | Remote sensing image road extraction method based on self-adaptive threshold | |
| CN115439748B (en) | Coastline erosion degree monitoring method and device and electronic equipment | |
| CN114299011A (en) | Remote sensing target quadrilateral frame rapid detection method based on deep learning | |
| Zeng et al. | The equipment detection and localization of large-scale construction jobsite by far-field construction surveillance video based on improving YOLOv3 and grey wolf optimizer improving extreme learning machine | |
| CN116363526B (en) | MROCNet model construction and multisource remote sensing image change detection method and system | |
| CN115035295A (en) | Remote sensing image semantic segmentation method based on shared convolution kernel and boundary loss function | |
| CN116665040A (en) | Building change detection method based on attention mechanism and with multi-scale input and output | |
| CN116452991A (en) | Attention enhancement and multiscale feature fusion artificial disturbance ground remote sensing extraction method | |
| CN116385911A (en) | Lightweight target detection method for unmanned aerial vehicle inspection insulator | |
| CN118212572A (en) | Road damage detection method based on improvement YOLOv7 | |
| CN117746264B (en) | Multitasking implementation method for unmanned aerial vehicle detection and road segmentation | |
| CN114219998A (en) | Sonar image real-time detection method based on target detection neural network | |
| CN116719031B (en) | Ocean vortex detection method and system for synthetic aperture radar SAR image | |
| Katiyar et al. | Flood Area Detection Using SAR Images with Deep Neural Network during, 2020 Kyushu Flood Japan | |
| CN117392508A (en) | Target detection method and device based on coordinate attention mechanism | |
| CN116895036A (en) | Deep learning-based farmland protection early warning method and device | |
| CN116452900A (en) | Target detection method based on lightweight neural network |
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 |