CN111523613B - Image analysis anti-interference method under complex environment of hydraulic engineering - Google Patents
Image analysis anti-interference method under complex environment of hydraulic engineering Download PDFInfo
- Publication number
- CN111523613B CN111523613B CN202010385793.9A CN202010385793A CN111523613B CN 111523613 B CN111523613 B CN 111523613B CN 202010385793 A CN202010385793 A CN 202010385793A CN 111523613 B CN111523613 B CN 111523613B
- Authority
- CN
- China
- Prior art keywords
- result
- image
- rectangle
- area
- interference
- 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
Images
Classifications
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- 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
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
- Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping
Abstract
The invention discloses an image analysis anti-interference method in a hydraulic engineering complex environment, and the method comprises the following steps of S1, roughly screening foreground connected areas; s2, removing redundant communication areas by a non-maximum suppression method; s3, extracting image features; s4, based on classification processing of the SVM, after the image characteristic value is obtained in the S3, distinguishing whether the foreground area is environmental interference or not by using the image characteristic value; s5, removing the environmental interference by using the classification model, outputting a judgment result to each circumscribed rectangular area by using the classification model, judging that the circumscribed rectangular area is the environmental interference if the output result is 0, and deleting the circumscribed rectangular area from the result; if the output result is 1, it is retained. The method effectively reduces the influence of external interference factors on the hydraulic engineering site on the algorithm result, reduces the false alarm of the algorithm, and improves the accuracy rate of dangerous case identification; the method can quickly establish a new model in a short time and put into use, and has better expansibility.
Description
Technical Field
The invention relates to an image analysis anti-interference method, in particular to an image analysis anti-interference method in a water conservancy project complex environment.
Background
Hydraulic engineering is mostly in field areas, the field environment is complex, and various external non-hydrological or geological interference factors such as animals, plants, personnel, vehicles and the like often appear in engineering scenes. Therefore, the water conservancy project dangerous case information is monitored by using the foreground target detection technology only, and the external interference factors and the actual project dangerous case cannot be effectively distinguished, so that a large amount of false alarm information is formed, and the normal operation of safety maintenance and emergency work of the water conservancy project is directly influenced.
Disclosure of Invention
The invention aims to provide an image analysis anti-interference method in a water conservancy project complex environment.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention relates to an image analysis anti-interference method under a water conservancy project complex environment, which comprises the following steps:
s1, roughly screening foreground connected regions: firstly, extracting each connected region outline from a foreground binary image processed by a monitored field video stream image, and then calculating the circumscribed rectangular area s of each connected region outline; setting an area threshold value t, if s is larger than t, storing and entering the next method for judgment, and if s is smaller than or equal to t, judging that the noise information is deleted from the result;
s2, removing redundant communication areas by a non-maximum suppression (NMS) method: the result obtained in the step S1 includes a plurality of external rectangles, and the plurality of external rectangles may cause a situation that the external rectangles include or cross a large part of regions, so that redundant rectangular windows should be removed;
s3, image feature extraction: using the HS histogram feature and the HOG gradient feature as image description features; the HS histogram feature only selects one dimension of hue to calculate a histogram, and the hue histogram parameter bin =180; the HOG gradient feature calculation parameters are set as follows: image size =64 × 128, sliding window size =64 × 128, block size =16 × 16, block _stride = (8,8), cell size =8 × 8, histogram bin =9, total feature dimension 3780; calculating each external rectangular image area in the S2 result by using the two image characteristic operators to obtain an image characteristic value of each external rectangular area;
s4, classification processing based on SVM: after S3, distinguishing whether the foreground area is environmental interference or not by using the image characteristic value; the SVM is a two-classification model, and the basic model of the SVM is a linear classifier with the maximum interval defined on a feature space; the basic idea of SVM learning is to solve a separation hyperplane which can correctly divide a training data set and has the maximum geometric interval; for linearly separable data sets, there are an infinite number of such hyperplanes, but the most geometrically spaced apart hyperplane is unique; in the method, an SVM is used as a basic classifier for classification, and the training and learning steps of the SVM are as follows:
s41, collecting a sample image: in the method, 500 positive and negative sample images are collected respectively, and the collected sample patterns are subjected to S3 feature extraction to form a sample feature set and labeled, wherein the labeling refers to: the environmental interference image is marked as 0, and the non-interference image is marked as 1;
s42, SVM learning parameters are set: after grid searching and tuning, the SVM learning parameters are set as: kernel method = polynomial kernel, polynomial degree =2, penalty coefficient C =0.01;
s43, importing the sample feature set into a trainer for training to finally obtain a classification model;
s5, removing environmental interference by using the classification model: using the classification model in the calculation result of S3, wherein the classification model outputs a judgment result to each circumscribed rectangular area, if the output result is 0, the circumscribed rectangular area is judged to be environmental interference, and the circumscribed rectangular area is deleted from the result; if the output result is 1, it is retained.
In S2, the step of removing redundant rectangular windows comprises the following steps:
s21, taking the area of the external rectangle as a confidence score, wherein the larger the area is, the larger the confidence score of the external rectangle is; calculating S1 to obtain confidence scores of all circumscribed rectangles in the result, and then arranging the confidence scores from large to small;
s22, taking the circumscribed rectangle with the largest confidence score in the result as a reference rectangle, taking the rest circumscribed rectangles as non-reference rectangles, then calculating the intersection ratio (IOU) of each non-reference rectangle and the reference rectangle one by one from large to small, deleting the current circumscribed rectangle from the result if the intersection ratio is more than or equal to 25%, and keeping if the intersection ratio is less than 25%, and calculating in sequence until the complete result is traversed;
s23, selecting the circumscribed rectangle with the maximum confidence score from the non-reference rectangles in the result of S1 as the reference rectangle, and repeating S22 until the result only contains the reference rectangle, thereby removing all redundant rectangle windows in the result.
In S1, the value range of the area threshold t is selected from 50-200 according to the actual engineering requirement.
The method utilizes digital image feature extraction and machine learning technology to further analyze and process the foreground target, uses various graphical feature operators to extract the image features of the foreground target, performs data labeling and feature engineering processing on the image features, and uses artificial intelligence technology to train and learn the feature samples to obtain a classification model. The classification model can automatically judge the foreground target, automatically eliminate the area with the judgment result as the external interference factor, and only reserve the area which is possibly the actual engineering dangerous case. Its advantages are:
1. by extracting and screening the features of the foreground connected regions of the foreground target detection algorithm one by one, the influence of external interference factors on the hydraulic engineering site on the algorithm result is effectively reduced, the false alarm of the algorithm is reduced, and the accuracy of dangerous case identification is improved.
2. The method introduces a machine learning mode to judge the characteristics, can automatically adapt to the change of data, realizes system personalization according to different environments, finds potential rules and values of data from massive data, and can realize self-iterative upgrade of the algorithm by continuously increasing a sample learning mode.
3. The model training method has the advantages that the SVM and digital image characteristics are used for processing, high efficiency and low cost are achieved in model training, real-time performance is achieved in computational efficiency, new models can be quickly built and put into use in a short time according to different use scenes, and good expansibility is achieved.
Drawings
FIG. 1 is a flow chart of the present invention.
Detailed Description
The following describes embodiments of the present invention in detail with reference to the drawings, which are implemented on the premise of the technical solution of the present invention, and detailed embodiments and specific operation procedures are provided, but the scope of the present invention is not limited to the following embodiments.
As shown in FIG. 1, the image analysis anti-interference method in the complex environment of the hydraulic engineering comprises the following steps:
s1, roughly screening foreground connected regions: firstly, extracting each connected region outline from a foreground binary image processed by a monitored field video stream image, and then calculating the circumscribed rectangular area s of each connected region outline; setting an area threshold value t, selecting the value range of the area threshold value t between 50 and 200 according to the actual engineering requirements, if s is greater than t, storing and entering the next step for judgment, and if s is less than or equal to t, considering that noise information is deleted from the result;
s2, removing redundant communication areas by a non-maximum suppression (NMS) method: the result obtained in the step S1 includes a plurality of external rectangles, and the plurality of external rectangles may cause a situation that the external rectangles include or cross a large part of regions, so that redundant rectangular windows should be removed; the step of removing the redundant rectangular window is as follows:
s21, taking the area of the external rectangle as a confidence score, wherein the larger the area is, the larger the confidence score of the external rectangle is; calculating S1 to obtain confidence scores of all circumscribed rectangles in the result, and then arranging the confidence scores from large to small;
s22, taking the circumscribed rectangle with the largest confidence score in the result as a reference rectangle, taking the rest circumscribed rectangles as non-reference rectangles, then calculating the intersection ratio (IOU) of each non-reference rectangle and the reference rectangle one by one from large to small, deleting the current circumscribed rectangle from the result if the intersection ratio is more than or equal to 25%, and keeping if the intersection ratio is less than 25%, and calculating in sequence until the complete result is traversed;
s23, selecting the circumscribed rectangle with the maximum confidence score from the non-reference rectangles in the result of S1 as the reference rectangle, and repeating S22 until the result only contains the reference rectangle, thereby removing all redundant rectangle windows in the result.
S3, image feature extraction: in order to be suitable for hydraulic engineering, through multiple experiments and screening, the HS histogram feature and the HOG gradient feature are finally selected and used as image description features;
HSV color characteristics: HSV is a method of representing points in the RGB color space in an inverted cone. HSV is Hue (Hue), saturation (Saturation), and lightness (Value), also known as HSB (Brightness). Hue is a basic attribute of color, that is, the name of a common color, such as red, yellow, etc. The saturation (S) is the purity of the color, and the higher the color is, the more pure the color is, and the lower the color is, the gray gradually becomes, and the value is 0 to 100%. In the method, only one dimension of Hue (Hue) is selected as a calculation histogram, and the Hue histogram parameter bin =180;
the HOG gradient feature descriptor focuses on the structure or shape of an object, and in one image, the appearance and shape of a local target can be well described by the directional density distribution of the gradient or the edge; the HOG gradient is characterized by the following nature: statistical information of the gradient, while the gradient exists mainly at the edge; in the method, the calculation parameters of the HOG gradient feature are set as follows: image size =64x128, sliding window size =64x128, block size =16x16, block _stride = (8,8), cell size =8x8, histogram bin =9. The total feature dimension is 3780;
calculating each external rectangular image area in the S2 result by using the two image characteristic operators to obtain an image characteristic value of each external rectangular area;
s4, classification processing based on SVM: after the image characteristic value is obtained in S3, distinguishing whether the foreground area is environmental interference or not by using the image characteristic value; the SVM is a two-classification model, and the basic model of the SVM is a linear classifier with the maximum interval defined on a feature space; the basic idea of SVM learning is to solve a separation hyperplane which can correctly divide a training data set and has the maximum geometric interval; for linearly separable data sets, there are an infinite number of such hyperplanes, but the most geometrically spaced apart hyperplane is unique; in the method, an SVM is used as a basic classifier for classification, and the training and learning steps of the SVM are as follows:
s41, collecting a sample image: in the method, 500 positive and negative sample images are collected respectively, and the collected sample patterns are subjected to S3 feature extraction to form a sample feature set and labeled, wherein the labeling refers to: the environmental interference image is marked as 0, and the non-interference image is marked as 1;
s42, SVM learning parameters are set: after grid searching and tuning, the SVM learning parameters are set as: kernel method = polynomial kernel, polynomial degree =2, penalty coefficient C =0.01;
s43, importing the sample feature set into a trainer for training to finally obtain a classification model;
s5, removing environmental interference by using the classification model: using the classification model in the calculation result of S3, wherein the classification model outputs a judgment result to each circumscribed rectangular area, if the output result is 0, the circumscribed rectangular area is judged to be environmental interference, and the circumscribed rectangular area is deleted from the result; if the output result is 1, it is retained.
Claims (3)
1. An image analysis anti-interference method under a hydraulic engineering complex environment is characterized in that: the method comprises the following steps:
s1, roughly screening foreground connected regions: firstly, extracting each connected region outline from a foreground binary image processed by a monitored field video stream image, and then calculating the circumscribed rectangular area s of each connected region outline; setting an area threshold t, if s is larger than t, storing and entering the next method for judgment, and if s is smaller than or equal to t, considering that noise information is deleted from the result;
s2, removing the redundant communication area by using a non-maximum suppression method: the result obtained in the step S1 includes a plurality of external rectangles, and the plurality of external rectangles may cause a situation that the external rectangles include or cross a large part of regions, so that redundant rectangular windows should be removed;
s3, image feature extraction: using the HS histogram feature and the HOG gradient feature as image description features; the HS histogram feature only selects one dimension of hue to calculate a histogram, and the hue histogram parameter bin =180; the HOG gradient feature calculation parameters are set as: image size =64 × 128, sliding window size =64 × 128, block size =16 × 16, block_stride = (8,8), cell size =8 × 8, histogram bin =9, total feature dimension 3780; calculating each external rectangular image area in the S2 result by using the two image characteristic operators to obtain an image characteristic value of each external rectangular area;
s4, classification processing based on SVM: after the image characteristic value is obtained in S3, distinguishing whether the foreground area is the environmental interference or not by using the image characteristic value, and classifying by using an SVM as a basic classifier, wherein the SVM training learning step is as follows:
s41, collecting a sample image: in the method, 500 positive and negative sample images are collected respectively, and the collected sample patterns are subjected to S3 feature extraction to form a sample feature set and labeled, wherein the labeling refers to: the environmental interference image is marked as 0, and the non-interference image is marked as 1;
s42, SVM learning parameters are set: after grid searching and tuning, the SVM learning parameters are set as: the kernel method = polynomial kernel, polynomial degree =2, and penalty coefficient C =0.01;
s43, importing the sample feature set into a trainer for training to finally obtain a classification model;
s5, removing environmental interference by using the classification model: using the classification model in the calculation result of S3, wherein the classification model outputs a judgment result to each circumscribed rectangular area, if the output result is 0, the circumscribed rectangular area is judged to be environmental interference, and the circumscribed rectangular area is deleted from the result; if the output result is 1, it is retained.
2. The image analysis anti-interference method under the complex hydraulic engineering environment according to claim 1, characterized in that: in S2, the step of removing redundant rectangular windows comprises the following steps:
s21, taking the area of the external rectangle as a confidence score, wherein the larger the area is, the larger the confidence score of the external rectangle is; calculating S1 to obtain confidence scores of all circumscribed rectangles in the result, and then arranging the confidence scores from large to small;
s22, taking the circumscribed rectangle with the largest confidence score in the result as a reference rectangle, taking the rest circumscribed rectangles as non-reference rectangles, calculating the intersection ratio of each non-reference rectangle and the reference rectangle one by one from large to small, deleting the current circumscribed rectangle from the result if the intersection ratio is more than or equal to 25%, and keeping if the intersection ratio is less than 25%, and calculating in sequence until the complete result is traversed;
s23, selecting the circumscribed rectangle with the maximum confidence score from the non-reference rectangles in the result of S1 as the reference rectangle, and repeating S22 until the result only contains the reference rectangle, thereby removing all redundant rectangle windows in the result.
3. The image analysis anti-interference method under the hydraulic engineering complex environment according to claim 1, is characterized in that: in S1, the area threshold value t range is selected from 50-200 according to the actual engineering requirements.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010385793.9A CN111523613B (en) | 2020-05-09 | 2020-05-09 | Image analysis anti-interference method under complex environment of hydraulic engineering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010385793.9A CN111523613B (en) | 2020-05-09 | 2020-05-09 | Image analysis anti-interference method under complex environment of hydraulic engineering |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111523613A CN111523613A (en) | 2020-08-11 |
CN111523613B true CN111523613B (en) | 2023-03-24 |
Family
ID=71905091
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010385793.9A Active CN111523613B (en) | 2020-05-09 | 2020-05-09 | Image analysis anti-interference method under complex environment of hydraulic engineering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111523613B (en) |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2013091370A1 (en) * | 2011-12-22 | 2013-06-27 | 中国科学院自动化研究所 | Human body part detection method based on parallel statistics learning of 3d depth image information |
CN106096531A (en) * | 2016-05-31 | 2016-11-09 | 安徽省云力信息技术有限公司 | A kind of traffic image polymorphic type vehicle checking method based on degree of depth study |
WO2018086299A1 (en) * | 2016-11-11 | 2018-05-17 | 广东电网有限责任公司清远供电局 | Image processing-based insulator defect detection method and system |
WO2018107939A1 (en) * | 2016-12-14 | 2018-06-21 | 国家海洋局第二海洋研究所 | Edge completeness-based optimal identification method for image segmentation |
CN108335294A (en) * | 2018-02-05 | 2018-07-27 | 贵州电网有限责任公司 | The power distribution room abnormality image-recognizing method of complex condition |
CN108875668A (en) * | 2018-06-28 | 2018-11-23 | 东北大学 | A kind of pedestrian detection method based on multiple features and cascade classifier |
WO2019000653A1 (en) * | 2017-06-30 | 2019-01-03 | 清华大学深圳研究生院 | Image target identification method and apparatus |
CN110363812A (en) * | 2019-07-10 | 2019-10-22 | 国网四川省电力公司电力科学研究院 | A kind of image-recognizing method |
-
2020
- 2020-05-09 CN CN202010385793.9A patent/CN111523613B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2013091370A1 (en) * | 2011-12-22 | 2013-06-27 | 中国科学院自动化研究所 | Human body part detection method based on parallel statistics learning of 3d depth image information |
CN106096531A (en) * | 2016-05-31 | 2016-11-09 | 安徽省云力信息技术有限公司 | A kind of traffic image polymorphic type vehicle checking method based on degree of depth study |
WO2018086299A1 (en) * | 2016-11-11 | 2018-05-17 | 广东电网有限责任公司清远供电局 | Image processing-based insulator defect detection method and system |
WO2018107939A1 (en) * | 2016-12-14 | 2018-06-21 | 国家海洋局第二海洋研究所 | Edge completeness-based optimal identification method for image segmentation |
WO2019000653A1 (en) * | 2017-06-30 | 2019-01-03 | 清华大学深圳研究生院 | Image target identification method and apparatus |
CN108335294A (en) * | 2018-02-05 | 2018-07-27 | 贵州电网有限责任公司 | The power distribution room abnormality image-recognizing method of complex condition |
CN108875668A (en) * | 2018-06-28 | 2018-11-23 | 东北大学 | A kind of pedestrian detection method based on multiple features and cascade classifier |
CN110363812A (en) * | 2019-07-10 | 2019-10-22 | 国网四川省电力公司电力科学研究院 | A kind of image-recognizing method |
Non-Patent Citations (2)
Title |
---|
基于图像高维特征压缩映射的混凝土表面裂缝检测算法;王保宪等;《北京理工大学学报》;20190415(第04期);全文 * |
基于稀疏特征的交通流视频检测算法;张鹏等;《南京大学学报(自然科学)》;20150330(第02期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN111523613A (en) | 2020-08-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Liu et al. | A detection method for apple fruits based on color and shape features | |
Zhao et al. | Cloud shape classification system based on multi-channel cnn and improved fdm | |
CN108446617B (en) | Side face interference resistant rapid human face detection method | |
CN107145889B (en) | Target identification method based on double CNN network with RoI pooling | |
CN108875624B (en) | Face detection method based on multi-scale cascade dense connection neural network | |
US20230289979A1 (en) | A method for video moving object detection based on relative statistical characteristics of image pixels | |
CN107247956B (en) | Rapid target detection method based on grid judgment | |
CN113139521B (en) | Pedestrian boundary crossing monitoring method for electric power monitoring | |
CN112241762B (en) | Fine-grained identification method for pest and disease damage image classification | |
CN109460704B (en) | Fatigue detection method and system based on deep learning and computer equipment | |
CN111738342B (en) | Pantograph foreign matter detection method, storage medium and computer equipment | |
CN104050471A (en) | Natural scene character detection method and system | |
CN105678213B (en) | Dual-mode mask person event automatic detection method based on video feature statistics | |
CN111027475A (en) | Real-time traffic signal lamp identification method based on vision | |
CN110598560A (en) | Night monitoring and identifying method and system based on neural network enhancement | |
CN111126162A (en) | Method, device and storage medium for identifying inflammatory cells in image | |
CN105354547A (en) | Pedestrian detection method in combination of texture and color features | |
Khan et al. | Comparitive study of tree counting algorithms in dense and sparse vegetative regions | |
CN109165592B (en) | Real-time rotatable face detection method based on PICO algorithm | |
CN113139497B (en) | System and method for identifying object on water surface and application based on 5G MEC | |
CN112446417B (en) | Spindle-shaped fruit image segmentation method and system based on multilayer superpixel segmentation | |
CN111523613B (en) | Image analysis anti-interference method under complex environment of hydraulic engineering | |
CN107341456B (en) | Weather sunny and cloudy classification method based on single outdoor color image | |
CN113537173B (en) | Face image authenticity identification method based on face patch mapping | |
CN111985482B (en) | Multi-mode multi-scene automatic license plate recognition method based on deep learning |
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 |