CN110765900B - Automatic detection illegal building method and system based on DSSD - Google Patents
Automatic detection illegal building method and system based on DSSD Download PDFInfo
- Publication number
- CN110765900B CN110765900B CN201910956710.4A CN201910956710A CN110765900B CN 110765900 B CN110765900 B CN 110765900B CN 201910956710 A CN201910956710 A CN 201910956710A CN 110765900 B CN110765900 B CN 110765900B
- Authority
- CN
- China
- Prior art keywords
- image
- dssd
- neural network
- building
- training set
- 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.)
- Active
Links
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- 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
- 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
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention relates to the field of illegal building image recognition, in particular to an automatic illegal building detection method and system based on DSSD, an image workstation, a detection system and a system terminal, wherein the image workstation is responsible for illegal building images, the detection system comprises training set and test set preprocessing, labeling and neural network discrimination, and the training set and test set preprocessing comprises multi-scale image denoising and enhancement. The method comprises the steps of acquiring and collecting a violation building graph for an image workstation, discriminating by using a DSSD neural network, and finally transmitting a result obtained by processing a detection system to a system terminal. The invention completes an automatic detection illegal building system based on an intelligent image processing technology and a neural network, can reduce the workload of manual identification to a certain extent, and has important significance for urban dynamic monitoring, illegal building identification, geographic information updating and the like.
Description
Technical Field
The invention relates to the technical field of illegal building image recognition, in particular to a method and a system for automatically detecting illegal buildings based on DSSD.
Background
Building change detection is one of important content of geographic national condition monitoring, and has important significance for illegal building identification, urban dynamic monitoring, geographic information updating and the like. Taking urban illegal building detection as an example, along with the continuous development of the economic society of China, the urban process is continuously accelerated, the urban buildings are continuously increased, and the number and the scale of illegal buildings are continuously increased, so that the phenomenon not only damages urban planning and urban landscapes, but also affects urban image and resident life, is a hot spot problem of common people, is a difficult problem of urban management, and is one of negative factors affecting social harmony. At present, the 'illegal cost is low, the law enforcement cost is high', which is one of the main reasons for the frequent occurrence of illegal buildings, the detection of the illegal buildings is weak besides the lack of relevant legal links, and the manual inspection mode is utilized to have a plurality of defects due to the lack of automatic monitoring means for the illegal buildings, so that the discovery process has longer period, and the large-scale monitoring cost is high. In recent years, cities such as Beijing and the like try to detect illegal buildings by using satellite image data, but an automatic analysis technology of image information is still not mature enough, and the specific gravity of manual identification and verification participation in the process is large. The land law enforcement and city management in the country annually have a billion investment in manpower and materials for this task. The method with high automation degree, robustness and reliability is urgently needed in the market to detect the urban illegal buildings, so that the improvement process of the urban illegal buildings is promoted.
Disclosure of Invention
The invention aims to provide a method and a system for automatically detecting illegal buildings based on DSSD, which are used for solving the problems in the background technology.
In order to achieve the above purpose, the present invention provides the following technical solutions:
a method for automatically detecting illegal building based on DSSD comprises the following steps:
step 1: the image workstation is responsible for collecting images of a building area, firstly carrying out image definition pretreatment on the images, and then sending the images into the detection system for image processing;
step 2: the detection system divides the collected images into a training set and a testing set for processing and analysis, wherein: the training set and the testing set comprise preprocessing, labeling and neural network discrimination;
step 3: and sending the identification result to a system terminal as an auxiliary diagnosis result to a user for final diagnosis.
Preferably, the preprocessing of the training set and the testing set comprises denoising and enhancing of the multi-scale image, marking the training set and the testing set image, and putting the training set and the testing set image into a DSSD neural network for training and identifying.
Preferably, the DSSD neural network includes a feature layer, a deconvolution module, a deconvolution layer, and a prediction module.
Preferably, the feature layer is 6 layers, which is composed of ResNet101 as a base network, conv3_x layers as convolution layers in ResNet101, followed by a series of progressively smaller convolution layers; the prediction module is a new prediction module (b) formed by adding a residual unit on the basis of the SSD prediction module (a) and performing an inter-channel addition on the original characteristic diagram after performing convolution processing on the original characteristic diagram and the characteristic diagram of the network trunk by a residual bypass.
Preferably, the DSSD neural network identifies lesions based on a regression and classification process, wherein: generating the recognition frame is a regression process, and judging that the focus in the recognition frame belongs to a classification process, so that the overall objective loss function is a weighted sum of the positioning loss and the confidence loss:
wherein N is the number of default frames corresponding to the real frames; c is the confidence of each class; alpha is the weight term set to 1; loss of positioning L loc (x, l, g) is the smooth between the prediction box l and the real target box g L1 Loss. l= (l) x ,l y ,l w ,l h ) Each item respectively represents the central position (x, y) of the predicted frame and the width w and the height h of the frame; g= (g) x ,g y ,g w ,g h ) Each item respectively represents the center position (x, y) of the real target and the width w and the height h of the frame thereof,
confidence loss L conf (x, c) is a Softmax penalty for multiple classes of confidence:
preferably, K cluster centers of the lesion aspect ratio are obtained using a K-means clustering algorithm, and the optimal K value is determined using an elbow method, whose core index is SSE (sum of the squared errors, sum of squares of errors):
wherein c i Is the ith cluster, p is c i Sample points m in (1) i Is c i Centroid (c) i The average value of all samples in (a), SSE is the clustering error of all samples, and represents the quality of the clustering effect, when K is smaller than the true clustering number, the descending amplitude of SSE is large, and when K reaches the true clustering number, the descending amplitude of SSE is suddenly reduced, so that the width and the height of a focus are obtained from a label file to obtain the aspect ratio, and then the K-means algorithm is operated by taking the aspect ratio as the characteristic.
The invention also provides a method for automatically detecting the illegal building based on the DSSD, which comprises the following steps:
the image workstation is used for collecting the building area image and uploading the collected image to the detection system for image processing;
the detection system divides the collected images into a training set and a testing set for processing and analysis, and then sends the images into a DSSD neural network;
the DSSD neural network trains and identifies the processing analysis image uploaded by the detection system, and then sends the identification result to the system terminal;
and the detection system receives the identification result and transmits the identification result to a user as an auxiliary diagnosis result for final diagnosis.
Preferably, the detection system comprises training set and test set preprocessing, labeling and DSSD neural network discrimination.
Preferably, the training set and test set preprocessing includes multi-scale image denoising and enhancement.
Preferably, the image workstation comprises photographing equipment and image scanning equipment, the photographing equipment photographs the building area, and the image scanning equipment performs image definition preprocessing on pictures obtained by the photographing equipment; wherein: the photographing device can adopt an unmanned plane, the image scanning device is based on a PC, and Matlab710 based on a Retinex image enhancement algorithm is loaded in the PC.
Compared with the prior art, the invention has the beneficial effects that:
the invention completes an automatic detection illegal building system based on intelligent image processing technology and neural network, reduces the workload of manual identification to a certain extent, and has important significance for urban dynamic monitoring, illegal building identification, geographic information updating and the like.
Drawings
FIG. 1 is a schematic diagram of a system of the present invention;
FIG. 2 is a system workflow diagram of the present invention;
FIG. 3 is a schematic diagram of a DSSD neural network according to the present invention;
FIG. 4 is a schematic diagram of a deconvolution module of a DSSD neural network according to the present invention;
fig. 5 is a schematic diagram of a prediction module structure of the DSSD neural network in the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1 to 5, the present invention provides a technical solution:
referring to fig. 1, a DSSD-based automatic detection violation building system includes an image workstation, a detection system, and a system terminal. The image workstation is responsible for collecting the building area image. The detection system comprises training set and test set preprocessing, labeling and neural network discrimination. The training set and the test set preprocessing comprise multi-scale image denoising and enhancement. The discrimination mainly utilizes a DSSD neural network.
By adopting the DSSD neural network, the DSSD designs the deconvolution module to fuse semantic information of each feature extraction layer, so that the DSSD neural network has more excellent performance in detection precision and small focus target detection capability. The structure is shown in fig. 2, the basic network is ResNet101, wherein the conv3_x layer is the convolution layer in ResNet101, and five layers are then a series of progressively smaller convolution layers that together act as the characteristic layer (6 layers in total) of the DSSD network. After this DSSD adds a series of deconvolution layers and designs a deconvolution module to fuse the previous feature layers (conv3_x, convolution layers) and deconvolution layers. In addition, DSSD also designs new prediction modules and moves predictions onto the deconvolution layer.
The deconvolution module has the function of fusing the high-level feature mapping information with the low-level feature mapping information, and the structure of the deconvolution module is shown in fig. 2. It can be seen that DSSD uses a learned deconvolution layer instead of bilinear upsampling and adds a normalization layer after each convolution layer. In addition, DSSD uses a method based on elemental dot-product to obtain better accuracy when combining higher-level feature mapping and lower-level feature mapping.
The prediction module of DSSD is shown in fig. 4 below, which is a new prediction module (b) formed by adding a residual unit based on the SSD prediction module (a) and performing the convolution processing on the original feature map and then performing the inter-channel addition with the feature map of the network trunk by the residual bypass.
DSSD identification of lesions is a regression and classification process. Generating the identification frame is a regression process, and judging the category of the focus in the identification frame is a classification process. Thus, the overall objective loss function is a weighted sum of the location loss and the confidence loss:
wherein N is the number of default frames corresponding to the real frames; c is the confidence of each class; alpha is the weight term set to 1; loss of positioning L loc (x, l, g) is the smooth between the prediction box l and the real target box g L1 Loss. l= (l) x ,l y ,l w ,l h ) Each item respectively represents the central position (x, y) of the predicted frame and the width w and the height h of the frame; g= (g) x ,g y ,g w ,g h ) Each item respectively represents the center position (x, y) of the real target and the width w and the height h of the frame of the real target.
Confidence loss L conf (x, c) is a Softmax penalty for multiple classes of confidence:
as can be seen from the DSSD identification process, the greater the number of aspect ratios, the greater the number of prior frames, so that the prior frames which are more matched with the real target can be found, thereby improving the detection accuracy, but more time is spent in predicting and performing NMS. It is important to choose an appropriate aspect ratio. We therefore reselect the aspect ratio of the a priori frame for their own dataset.
K cluster centers for the lesion aspect ratio are obtained using a K-means clustering algorithm, and the optimal K-value is determined using an elbow method. The core index of the elbow method is SSE (sum of the squared errors, sum of squares error):
wherein c i Is the ith cluster, p is c i Sample points m in (1) i Is c i Centroid (c) i Average value of all samples in (a), SSE is a clustering error of all samples, and represents the quality of the clustering effect. When k is smaller than the true cluster number, the magnitude of the drop in SSE will be large, while when k reaches the true cluster number, the magnitude of the drop in SSE will be drastically reduced. Thus, the lesion width and height are obtained from the label file to obtain the aspect ratio, and then the K-means algorithm is run featuring the aspect ratio.
And finally, sending the identification result to a system terminal, and giving the identification result to a user (user) as an auxiliary diagnosis result for final identification to obtain the position of the illegal building.
The technical scheme of the invention is as follows:
a method for automatically detecting illegal building based on DSSD comprises an image workstation, a detection system and a system terminal. The image workstation is responsible for acquiring images of the building area. The detection system comprises training set and test set preprocessing, labeling and neural network discrimination. The training set and the test set preprocessing comprise multi-scale image denoising and enhancement. The discrimination mainly utilizes a DSSD neural network. According to the invention, the unmanned aerial vehicle is adopted as photographing equipment, and the unmanned aerial vehicle can be used for photographing in a low altitude, so that the manual photographing difficulty and workload are reduced. The unmanned aerial vehicle shoots buildings (including illegal buildings) in cities or other areas according to a preset route, the obtained picture images are input into an image scanning device for preprocessing, the preprocessing is performed under a Windows XP or above operating system in a computer (PC), matlab710 based on a Retinex image enhancement algorithm is installed in the Windows XP, and the picture or the photo shot by the unmanned aerial vehicle can be subjected to sharpening processing based on the Retinex algorithm. Because unmanned aerial vehicle is taking a photograph in the in-process of collecting evidence, there is weather or other factor's influence, for example have haze condition or receive wind speed interference and make the photo of taking a photograph greatly reduced on the whole contrast and the luminance of image, the image color distortion. The method adopts the post-processing of the reset image of Retinex, can enhance the detail information of the image while enhancing the whole image, has clear edges, further enhances the color information of the image, better restores the color of the image and achieves the purpose of enhancing the highlight region. The Matlab710 based on the Retinex image enhancement algorithm is the prior art, and specific reference may be made to the image enhancement algorithm-article number based on the Retinex principle: 1009-3044 (2018) 11-0185-02, a method for improving foggy day image clarity-article number 100325060 (2011) 0120083204, etc.
The working principle and working process of the invention are as follows:
1. the image workstation is responsible for collecting the building area image, firstly carries out image definition preprocessing on the image, and then sends the image to the detection system for image processing.
2. The detection system divides the collected images into a training set and a testing set for processing and analysis. The method comprises preprocessing, labeling and neural network discrimination of a training set and a testing set. Wherein training set and testing set preprocessing includes multi-scale image denoising and enhancement. And marking the training set and the test set images, and putting the training set and the test set images into a DSSD neural network for training and recognition.
3. And the identification result is sent to the system terminal as an auxiliary diagnosis result to be transmitted to a user for final diagnosis, so that the exact position of the illegal building in the city or other areas can be obtained, and the law enforcement department can conveniently dismount the illegal building.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (8)
1. The automatic detection violation building method based on the DSSD is characterized by comprising the following steps of:
step 1: the image workstation is responsible for collecting images of a building area, firstly carrying out image definition pretreatment on the images, and then sending the images into the detection system for image processing;
step 2: the detection system divides the collected images into a training set and a testing set for processing and analysis, wherein: the training set and the testing set comprise preprocessing, labeling and neural network discrimination;
step 3: the identification result is sent to a system terminal and used as an auxiliary diagnosis result to be sent to a user for final diagnosis;
preprocessing a training set and a testing set, namely denoising and enhancing a multi-scale image, marking the training set and the testing set image, and putting the training set and the testing set image into a DSSD neural network for training and identifying;
DSSD neural networks identify lesions based on regression and classification processes, wherein:
generating the recognition frame is a regression process, and judging that the focus in the recognition frame belongs to a classification process, so that the overall objective loss function is a weighted sum of the positioning loss and the confidence loss:
wherein N is the number of default frames corresponding to the real frames; c is the confidence of each class; alpha is the weight term set to 1; loss of positioning L loc (x, l, g) is the smooth between the prediction box l and the real target box g L1 Loss. l= (l) x ,l y ,l w ,l h ) Each item respectively represents the middle of the predicted frameThe heart position (x, y) and the width w and height h of the frame of the heart position (x, y); g= (g) x ,g y ,g w ,g h ) Each item respectively represents the center position (x, y) of the real target and the width w and the height h of the frame thereof,
confidence loss L conf (x, c) is a Softmax penalty for multiple classes of confidence:
2. the DSSD-based automatic detection violation building method according to claim 1, wherein: the DSSD neural network comprises a feature layer, a deconvolution module, a deconvolution layer and a prediction module.
3. The DSSD-based automatic detection violation building method according to claim 2, wherein: the feature layer is 6 layers, the structure is based on ResNet101, the conv3_x layer is used as a convolution layer in ResNet101, and five layers are a series of gradually smaller convolution layers; the prediction module is a new prediction module (b) formed by adding a residual unit on the basis of the SSD prediction module (a) and performing an inter-channel addition on the original characteristic diagram after performing convolution processing on the original characteristic diagram and the characteristic diagram of the network trunk by a residual bypass.
4. The DSSD-based automatic detection violation building method according to claim 1, wherein: k cluster centers of the lesion aspect ratio were obtained using a K-means clustering algorithm, and the optimal K values were determined using the elbow method, whose core index is SSE (sum of the squared errors, sum of squares of errors):
wherein c i Is the ith cluster, p is c i Sample points m in (1) i Is c i Centroid (c) i The average value of all samples in (a), SSE is the clustering error of all samples, and represents the quality of the clustering effect, when K is smaller than the true clustering number, the descending amplitude of SSE is large, and when K reaches the true clustering number, the descending amplitude of SSE is suddenly reduced, so that the width and the height of a focus are obtained from a label file to obtain the aspect ratio, and then the K-means algorithm is operated by taking the aspect ratio as the characteristic.
5. A DSSD-based automated detection violation building system according to any of claims 1 to 4, comprising:
the image workstation is used for collecting the building area image and uploading the collected illegal building image to the detection system for image processing;
the detection system divides the collected images into a training set and a testing set for processing and analysis, and then sends the images into a DSSD neural network;
the DSSD neural network trains and identifies the processing analysis image uploaded by the detection system, and then sends the identification result to the system terminal;
and the detection system receives the identification result and transmits the identification result to a user as an auxiliary diagnosis result for final diagnosis.
6. The DSSD-based automated detection violation building system according to claim 4, wherein: the detection system comprises a training set and a test set for preprocessing, labeling and DSSD neural network discrimination.
7. The DSSD-based automated detection violation building system according to claim 6, wherein: training set and test set preprocessing includes multi-scale image denoising and enhancement.
8. The DSSD-based automated detection violation building system according to claim 6, wherein: the image workstation comprises photographing equipment and image scanning equipment, wherein the photographing equipment photographs a building area, and the image scanning equipment performs image definition preprocessing on a picture obtained by the photographing equipment; wherein: the photographing device can adopt an unmanned plane, the image scanning device is based on a PC, and Matlab710 based on a Retinex image enhancement algorithm is loaded in the PC.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910956710.4A CN110765900B (en) | 2019-10-10 | 2019-10-10 | Automatic detection illegal building method and system based on DSSD |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910956710.4A CN110765900B (en) | 2019-10-10 | 2019-10-10 | Automatic detection illegal building method and system based on DSSD |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110765900A CN110765900A (en) | 2020-02-07 |
| CN110765900B true CN110765900B (en) | 2023-09-05 |
Family
ID=69331587
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910956710.4A Active CN110765900B (en) | 2019-10-10 | 2019-10-10 | Automatic detection illegal building method and system based on DSSD |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110765900B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111582117A (en) * | 2020-04-29 | 2020-08-25 | 长江大学 | Unmanned aerial vehicle illegal building inspection method, equipment and storage medium |
| CN113033289A (en) * | 2021-01-29 | 2021-06-25 | 南瑞集团有限公司 | Safety helmet wearing inspection method, device and system based on DSSD algorithm |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711348A (en) * | 2018-12-28 | 2019-05-03 | 湖南航天远望科技有限公司 | Intelligent monitoring method and system based on the long-term real-time architecture against regulations in hollow panel |
| CN110211137A (en) * | 2019-06-08 | 2019-09-06 | 西安电子科技大学 | Satellite Image Segmentation method based on residual error network and U-Net segmentation network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019133052A1 (en) * | 2017-12-28 | 2019-07-04 | Yang Shao Wen | Visual fog |
-
2019
- 2019-10-10 CN CN201910956710.4A patent/CN110765900B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711348A (en) * | 2018-12-28 | 2019-05-03 | 湖南航天远望科技有限公司 | Intelligent monitoring method and system based on the long-term real-time architecture against regulations in hollow panel |
| CN110211137A (en) * | 2019-06-08 | 2019-09-06 | 西安电子科技大学 | Satellite Image Segmentation method based on residual error network and U-Net segmentation network |
Non-Patent Citations (1)
| Title |
|---|
| 钟波 ; .应用无人机对于违法建筑物的影像对比判别.城市地理.2017,(22),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110765900A (en) | 2020-02-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108596277B (en) | Vehicle identity recognition method and device and storage medium | |
| CN110674861B (en) | Intelligent analysis method and device for power transmission and transformation inspection images | |
| CN107123131B (en) | Moving target detection method based on deep learning | |
| CN110008854B (en) | Unmanned aerial vehicle image highway geological disaster identification method based on pre-training DCNN | |
| CN111754498A (en) | Conveyor belt carrier roller detection method based on YOLOv3 | |
| CN110796009A (en) | Method and system for detecting marine vessel based on multi-scale convolution neural network model | |
| CN111079640B (en) | Vehicle type identification method and system based on automatic amplification sample | |
| CN105574550A (en) | Vehicle identification method and device | |
| CN106709903B (en) | PM2.5 concentration prediction method based on image quality | |
| CN110852164A (en) | YOLOv 3-based method and system for automatically detecting illegal building | |
| CN112287827A (en) | Complex environment pedestrian mask wearing detection method and system based on intelligent lamp pole | |
| CN113435407B (en) | Small target identification method and device for power transmission system | |
| CN110765900B (en) | Automatic detection illegal building method and system based on DSSD | |
| CN109086803A (en) | A kind of haze visibility detection system and method based on deep learning and the personalized factor | |
| CN111540203B (en) | Method for adjusting green light passing time based on fast-RCNN | |
| CN114596316A (en) | Road image detail capturing method based on semantic segmentation | |
| CN112668375B (en) | Tourist distribution analysis system and method in scenic spot | |
| CN115131747A (en) | Knowledge distillation-based power transmission channel engineering vehicle target detection method and system | |
| CN110991374A (en) | Fingerprint singular point detection method based on RCNN | |
| CN112785610B (en) | Lane line semantic segmentation method integrating low-level features | |
| CN113255580A (en) | Method and device for identifying sprinkled objects and vehicle sprinkling and leaking | |
| CN113378642A (en) | Method for detecting illegal occupation buildings in rural areas | |
| CN112836590A (en) | Flood disaster monitoring method and device, electronic equipment and storage medium | |
| CN111897993A (en) | Efficient target person track generation method based on pedestrian re-recognition | |
| CN110163081A (en) | Regional invasion real-time detection method, system and storage medium based on SSD |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |