CN111553265A - Method and system for detecting internal defects of drainage pipeline - Google Patents
Method and system for detecting internal defects of drainage pipeline Download PDFInfo
- Publication number
- CN111553265A CN111553265A CN202010343267.6A CN202010343267A CN111553265A CN 111553265 A CN111553265 A CN 111553265A CN 202010343267 A CN202010343267 A CN 202010343267A CN 111553265 A CN111553265 A CN 111553265A
- Authority
- CN
- China
- Prior art keywords
- image
- video
- local
- defect
- sample
- 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.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 197
- 238000000034 method Methods 0.000 title claims abstract description 69
- 238000012549 training Methods 0.000 claims abstract description 108
- 238000001514 detection method Methods 0.000 claims abstract description 95
- 238000009432 framing Methods 0.000 claims abstract description 93
- 238000012545 processing Methods 0.000 claims abstract description 84
- 238000012360 testing method Methods 0.000 claims abstract description 60
- 238000005516 engineering process Methods 0.000 claims abstract description 10
- 230000002159 abnormal effect Effects 0.000 claims description 18
- 238000001914 filtration Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 12
- 230000007797 corrosion Effects 0.000 claims description 8
- 238000005260 corrosion Methods 0.000 claims description 8
- 238000012850 discrimination method Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 7
- 230000033001 locomotion Effects 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 6
- 230000008021 deposition Effects 0.000 claims description 4
- 230000007849 functional defect Effects 0.000 claims description 4
- 239000000463 material Substances 0.000 claims description 4
- 230000007847 structural defect Effects 0.000 claims description 4
- 230000005856 abnormality Effects 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 230000035515 penetration Effects 0.000 claims description 3
- 230000033228 biological regulation Effects 0.000 claims description 2
- 230000009466 transformation Effects 0.000 description 13
- 230000003628 erosive effect Effects 0.000 description 6
- 238000007689 inspection Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 230000000877 morphologic effect Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000009193 crawling Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 230000010339 dilation Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 201000011243 gastrointestinal stromal tumor Diseases 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000010865 sewage Substances 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a method and a system for detecting internal defects of a drainage pipeline. The method comprises the following steps: acquiring detection video data inside the pipeline through field detection, wherein the video data comprises a video of a collected sample and a test video; performing frame processing on a video of a collected sample and collecting a sample library image; performing model training according to the sample library images to obtain a training model, wherein the training model comprises a global defect training model and a local defect training model; performing framing processing on the test video to obtain an ultra-clear image, and processing the ultra-clear image; and detecting the frame processing result image of the test by using the global defect training model and the local defect training model in sequence by using a funnel mode to obtain a defect detection result. By adopting the video automatic framing technology and the double-model precise detection strategy, the detection period of the internal defects of the drainage pipeline can be shortened, and the accuracy of defect interpretation can be improved.
Description
Technical Field
The invention relates to the field of detection of internal defects of drainage pipelines, in particular to a method and a system for detecting internal defects of drainage pipelines.
Background
The drainage system is an important component of urban public facilities, and has the functions of protecting and improving the environment and eliminating the sewage hazard. In order to ensure that the drainage pipeline is smooth, drainage hidden dangers need to be found in time and the pipeline is prevented from leaking and polluting, the drainage pipeline is required to be regularly inspected, and the defects of the pipeline are overcome. The defects of the drainage pipeline comprise structural defects, such as breakage, deformation, dislocation, disjointing, leakage, corrosion, rubber ring falling off, hidden branch pipe connection and foreign matter invasion, and functional defects, such as deposition, scaling, obstacles, tree roots, water accumulation, plugging, scum and the like. The defects are inspected and evaluated, and reliable technical basis is provided for maintenance of the drainage pipeline. At present, due to the development of a pipeline detection technology, engineering construction enterprises and institutions gradually adopt a CCTV detection mode to effectively detect drainage pipelines, and convenience is brought to pipeline maintenance.
The CCTV detection process is characterized in that detection personnel control the advancing speed and direction of a crawler in a pipeline through a main controller and control a camera to transmit images inside the pipeline to a display screen of the main controller through a cable, an operator can monitor the internal conditions of the pipeline in real time, and meanwhile, original image records are stored for further analysis. The detection of QV also needs the operating lever of the detector, and sends the camera to the pipeline opening, and the center of the pipeline is aligned for detection. Under the condition that personnel and construction period are in tension, the detailed condition of the pipeline defect cannot be recorded in real time in the whole field detection process. The pipeline inspection personnel finish the inspection and provide the inspection image data, and the field personnel manually interpret the image, record the detailed condition of the defect and summarize the data. The manual interpretation needs to play the image completely, the time is long, and interpretation results of different field workers also come in and go out to a certain extent.
Disclosure of Invention
The invention aims to provide a method and a system for detecting internal defects of a drainage pipeline, which can shorten the construction period and improve the accuracy of defect interpretation.
In order to achieve the purpose, the invention provides the following scheme:
a method for detecting internal defects of a drainage pipeline comprises the following steps:
acquiring detection video data inside a pipeline through field detection, wherein one part of the video data is randomly selected as a video of a collected sample, the other part of the video data is used as a test video, and the video of the collected sample and the test video are not mutually crossed and overlapped;
performing video framing processing on the video of the collected sample to obtain a framing processing result image of the video of the collected sample;
acquiring a sample library image according to the video framing processing result image of the acquired sample, wherein the sample library image comprises a global defect image sample, a local defect image sample, a normal pipeline image sample and a local normal pipeline image sample;
performing model training according to the sample library images to obtain a training model, wherein the training model comprises a global defect training model and a local defect training model;
performing framing processing on the test video to obtain a framing processing result image of the test video;
and preprocessing a framing result image of the test video by adopting a dual-model accurate detection strategy, and detecting a preprocessed result by adopting the global defect training model and the local defect training model to obtain a defect detection result.
Optionally, the video framing processing is performed on the video of the collected sample to obtain a framing processing result image of the video of the collected sample, specifically including;
adopt FFmpeg image fast framing technique to it carries out framing processing to the video of gathering the sample, the video framing processing's result is many resolutions and the super clear image that the size is the same, super clear image and former video size, resolution ratio are the same, including global map and local map, the global map is the image that the camera lens was taken at the inside direct vision place ahead of pipeline, the local map is the camera lens and looks around the local image that the department was shot at the setting in-process in the pipeline.
Optionally, the acquiring a sample library image according to the video framing processing result image of the acquired sample specifically includes:
judging the video framing processing result of the collected sample according to the definition of internal defects of the drainage pipeline by the national urban drainage pipeline detection and evaluation technical regulation, and selecting ten types of structural defect images of fracture, deformation, corrosion, dislocation, fluctuation, disjunction, interface material shedding, branch pipe hidden connection, foreign matter penetration and leakage and six types of functional defect images of deposition, scaling, obstacles, residual walls, dam roots, tree roots and scum as samples of a training model, wherein the global defect sample image is a large image with the same size as the original size of the video, the local defect image sample is a small image at the position and at the edge of a defect intercepted in a local image, the size is 416 x 416, and the resolution is the same as the original video;
selecting a normal pipeline image and a local normal pipeline image as samples of a training model, wherein the normal pipeline image sample is a high-definition large image with the original size of a video, the local normal pipeline image sample is a small image randomly intercepted from the local image, the size of the local normal pipeline image sample is 416 x 416, and the resolution is the same as that of the original video;
and combining the global defect image sample, the local defect image sample, the normal pipeline image sample and the local normal image sample to obtain a sample library image.
Optionally, the performing model training according to the sample library image to obtain a training model specifically includes:
and training a model by adopting the global defect image sample, the local defect image sample, the normal pipeline image sample and the local normal image sample to obtain a global defect training model and a local defect training model.
Optionally, the framing the test video to obtain a framing result image of the test video specifically includes:
adopt FFmpeg image fast framing technique right the test video carries out framing processing, and the video framing processing's result is many resolutions and the super clear image that the size is the same, super clear image and former video size, resolution ratio are the same, including global image and local image, global image is the image that the camera lens was taken in the inside direct vision place ahead of pipeline, local image is the camera lens and looks around the local image that the in-process was looked around in the pipeline and sets for the department shooting in the process.
Optionally, a dual-model accurate detection strategy is adopted, the framing result image of the test video is preprocessed, the global defect training model and the local defect training model are adopted to detect the preprocessed result, and a defect detection result is obtained, and the method specifically includes the following steps:
preliminarily screening the framing processing result of the test video by adopting a multi-domain joint discrimination method to obtain a screened effective image;
extracting key video frame images from the screened effective images by using an image definition judging technology;
and detecting the key video frame image by using the global defect training model and the local defect training model in sequence in a funnel mode to obtain a defect detection result.
Optionally, the preliminary screening is performed on the frame division processing result of the test video by using a multi-domain joint discrimination method to obtain a screened image, and the method specifically includes:
and carrying out image foreground and background distinguishing and enhancing on the frame processing result of the test video, filtering a video frame image before the robot crawls at the pipe orifice and filtering a video frame image of the motion state of the robot, and obtaining a screened effective image.
Optionally, extracting a key video frame image from the screened effective image by using an image sharpness discrimination technique, specifically including:
processing the screened effective images by adopting a definition discrimination algorithm based on no reference image to obtain processed images;
determining a variance according to the processed image;
and distinguishing a clear image and a blurred image according to the size of the variance, wherein the clear image is a key video frame image.
Optionally, the detecting of the key video frame image by using the global defect training model and the local defect training model in a funnel mode to obtain a defect detection result specifically includes:
detecting the key video frame images by adopting a global defect training model, if the key video frame images are globally normal, judging that the pipeline is normal, and if the key video frame images are abnormal, judging the type and other related information of the pipeline defect according to the probability distribution of the global defect type; if the key video frame image is a local image, dividing the local image into nine local sub-images, and further detecting by adopting a local defect training model;
dividing the local image into 9 local sub-images, wherein the 9 local sub-images can basically cover the whole local image and are overlapped with each other, and the set of the local sub-images is a local image; the local sub-images divided by the same local image are mutually associated; detecting local subimages by adopting a local defect training model, and if 9 local subimages corresponding to the local subimages are not abnormal, judging that no abnormality exists in the local subimages; if any sub-image of the 9 local sub-images corresponding to the local image is abnormal, judging that the local image is abnormal, and judging the local image as the abnormal type with the most occurrence times in the sub-image;
and after the detection is finished, outputting the corresponding pipeline starting point number, the corresponding pipeline ending point number, the corresponding defect type and the corresponding defect picture in the mode of result, storing and naming to obtain a defect detection result.
A system for detecting internal defects of a drain pipeline, comprising:
the video data acquisition module is used for acquiring detection video data inside the pipeline through field detection, wherein one part of the video data is randomly selected as a video of a collected sample, the other part of the video data is used as a test video, and the video of the collected sample and the test video are not mutually crossed and overlapped;
the first framing processing module is used for performing video framing processing on the video of the collected sample to obtain a framing processing result image of the video of the collected sample;
the sample library image acquisition module is used for acquiring sample library images according to the video framing processing result images of the acquired samples, wherein the sample library images comprise global defect image samples, local defect image samples, normal pipeline image samples and local normal pipeline image samples;
the training module is used for carrying out model training according to the sample library images to obtain a training model, and the training model comprises a global defect training model and a local defect training model;
the second framing processing module is used for framing the test video to obtain a framing processing result image of the test video;
and the detection module is used for preprocessing a framing result image of the test video by adopting a dual-model accurate detection strategy, and detecting a preprocessed result by adopting the global defect training model and the local defect training model to obtain a defect detection result.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a method and a system for detecting internal defects of a drainage pipeline, wherein video data for detecting the internal defects of the pipeline are obtained through field detection, one part of the video data is randomly selected as a video for collecting a sample, the other part of the video data is used as a test video, and the video for collecting the sample and the test video are not overlapped in a cross mode; performing video framing processing on the video of the collected sample to obtain a framing processing result of the video of the collected sample; acquiring a sample library image according to a video framing processing result of an acquired sample, wherein the sample library image comprises a global defect image sample, a local defect image sample, a normal pipeline image sample and a local normal pipeline image sample; training the model according to the sample library image to obtain a training model, wherein the training model comprises a global defect training model and a local defect training model; and automatically framing the test video to obtain a framing processing result of the test video. By adopting a dual-model accurate detection strategy and detecting the framing result of the test video by using the global defect training model and the local defect training model, the defect detection result is obtained.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of the method for detecting internal defects of a drainage pipeline according to the present invention;
FIG. 2 is a first illustrative representation of the erosion transformation of a target area according to the present invention;
FIG. 3 is a graphical representation of the erosion transformation of the target area of the present invention;
FIG. 4 is a first expanded transformation rendering of the target area of the present invention;
FIG. 5 is a graphical representation of the target region dilation transformation of the present invention;
FIG. 6 is a schematic diagram of distribution of to-be-detected image subgraphs according to the present invention;
FIG. 7 is a view showing the structure of the inspection system for internal defects of a drainpipe according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a method and a system for detecting internal defects of a drainage pipeline, which can shorten the construction period and improve the accuracy of defect interpretation.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
FIG. 1 is a flow chart of the method for detecting internal defects of a drainage pipeline according to the present invention. As shown in FIG. 1, a method for detecting internal defects of a drainage pipeline comprises the following steps:
step 101: the method comprises the steps that detection video data inside a pipeline are obtained through field detection, one part of the video data is randomly selected to be used as a video of a collected sample, the other part of the video data is used as a test video, and the video of the collected sample and the test video are not overlapped in a cross mode.
The drainage pipeline detection adopts a CCTV manual field detection method, and detection objects are drainage pipelines made of concrete and PVC materials. The field inspection personnel send the robot to the underground, and the robot is operated to enable the robot to crawl towards the interior of the pipeline at the pipe orifice. When the robot reaches the pipe orifice, field detection personnel operate the equipment to enable the crawling distance read by the equipment to return to zero (the equipment judges the distance from the inside of the pipeline to the pipe orifice of the robot according to the crawling distance in the shooting process). In the operation process, if a defect in the pipeline is found, the robot needs to be operated to stop and the shooting lens is made to be static to clearly shoot the global defect (such as integrally embodied defects of stagger, deformation, fluctuation and the like) in the pipeline, if a local defect (such as locally embodied defects of fracture, corrosion and the like) is found, the detector enables the robot to be static, and meanwhile, the lens looks around the interface of the pipeline, aligns the defect and gazes for observation. The step of shooting image data comprises two parts of a sample image and a test image, wherein the shot images are all avi format super-clear images, the resolution is 1920 x 1080, the frame rate is 25 frames/second, and the images have no difference in the aspects of resolution, quality and the like.
Step 102: performing video framing processing on the video of the collected sample to obtain a framing processing result image of the video of the collected sample, wherein the video framing processing result image specifically comprises;
adopt FFmpeg image fast framing technique to it carries out framing processing to the video of gathering the sample, the video framing processing's result is many resolutions and the super clear image that the size is the same, super clear image and former video size, resolution ratio are the same, including global map and local map, the global map is the image that the camera lens was taken at the inside direct vision place ahead of pipeline, the local map is the camera lens and looks around the local image that the department was shot at the setting in-process in the pipeline.
Step 103: acquiring a sample library image according to the video framing processing result image of the acquired sample, wherein the sample library image comprises a global defect image sample, a local defect image sample, a normal pipeline image sample and a local normal pipeline image sample, and the method specifically comprises the following steps:
according to the definition of internal defects of drainage pipelines in China 'town drainage pipeline detection and evaluation technical rules', the video framing processing result of the collected sample is interpreted, and ten types of structural defect images of fracture, deformation, corrosion, dislocation, fluctuation, disjointing, interface material falling, branch pipe hidden joint, foreign matter penetration and leakage, and six types of functional defect images of deposition, scaling, obstacles, residual walls, dam roots, tree roots and scum are selected as samples of a training model, wherein the global defect sample image is a large image with the same size as the original size of the video, the local defect image sample is a small image at the position and at the edge of a defect intercepted in a local image, the size is 416 x 416, and the resolution is the same as the original video.
The method comprises the steps of selecting a normal pipeline image and a local normal pipeline image as samples of a training model, wherein the normal pipeline image sample is a high-definition large image with the original size of a video, the local normal pipeline image sample is a small image randomly captured from the local image, the size of the small image is 416 x 416, and the resolution is the same as that of the original video.
And combining the global defect image sample, the local defect image sample, the normal pipeline image sample and the local normal image sample to obtain a sample library image.
Step 104: performing model training according to the sample library image to obtain a training model, wherein the training model comprises a global defect training model and a local defect training model, and specifically comprises the following steps:
and training a model by adopting the global defect image sample, the local defect image sample, the normal pipeline image sample and the local normal image sample to obtain a global defect training model and a local defect training model.
Aiming at the fact that scenes and characteristics of various defects in a pipeline detection task are complex and changeable, the defects of shooting in the front view of a lens and the defects of shooting in the interior of a pipeline are different, if all the defects are subjected to model training through one-time feature extraction, the same type of input (large images or small images) can not be used for identifying the corresponding defects particularly accurately. Therefore, a dual-model precise detection strategy is provided for the two major types of characteristics, the global defect and the local defect are distinguished according to respective characteristics, and two sample libraries are manufactured, namely a global defect sample library and a local defect sample library. That is, a global defect-applicable model and a local defect-applicable model are trained separately.
Step 105: performing frame processing on the test video to obtain a frame processing result image of the test video, which specifically comprises:
adopt FFmpeg image fast framing technique right the test video carries out framing processing, and the video framing processing's result is many resolutions and the super clear image that the size is the same, super clear image and former video size, resolution ratio are the same, including global image and local image, global image is the image that the camera lens was taken in the inside direct vision place ahead of pipeline, local image is the camera lens and looks around the local image that the in-process was looked around in the pipeline and sets for the department shooting in the process.
Because the scenes inside the pipeline are changeable and very complex, the background information of the scenes with various defects is complex and changeable, the types of the defects are complex and various, the scenes with the same defect are also different, and even the form of the same defect is greatly changed along with the scenes. If the corresponding defect is detected effectively in such a complex background or even in the foreground itself, a large prediction error is formed by simply performing the abnormal detection from the perspective of the video stream, and even the abnormal detection cannot be predicted at all, so that the defect can be detected only from the perspective of the abnormal detection based on the picture. The invention divides the video into one frame and one frame of pictures and detects the pictures.
And performing framing processing on the test video by adopting an FFmpeg image fast framing technology to obtain a plurality of ultra-clear images with the same resolution.
FFmpeg includes a very integrated image stream manipulation interface, such as may define the quality of the framed pictures, the resolution of the framed pictures, the framing interval, the output format, etc., that can effectively extract each frame of the video stream. The result of the framing process is used for defect detection, and thus the picture quality and resolution of the original video are required to be preserved. The result of the framing process is to obtain an ultra-clear image of the same resolution from the test video.
Step 106: adopting a double-model accurate detection strategy, preprocessing a framing result image of the test video, adopting the global defect training model and the local defect training model to detect a preprocessed result, and obtaining a defect detection result, wherein the method specifically comprises the following steps:
step 1061: preliminarily screening the frame processing result of the test video by adopting a multi-domain joint discrimination method to obtain a screened effective image, which specifically comprises the following steps:
and carrying out image foreground and background distinguishing and enhancing on the frame processing result of the test video, filtering a video frame image before the robot crawls at the pipe orifice and filtering a video frame image of the motion state of the robot, and obtaining a screened effective image.
And performing preliminary screening on the regions where the defects possibly appear on the ultra-clear image after the test video is framed by using a multi-domain joint discrimination method. The pipeline detection image is often high in frame rate, so tens of thousands of pictures are available after framing. If the method of frame-by-frame detection is adopted, the method is not satisfactory from either time consumption or final detection accuracy. Aiming at the situation, a multi-domain joint discrimination method is provided to eliminate the regions which are impossible to have defects in the video stream, the super-clear image is subjected to image foreground and background discrimination and enhancement, the video frame image before the robot crawls at the pipe orifice is filtered, and the video frame image in the motion state of the robot is filtered to obtain the screened effective image, so that a large amount of invalid data can be filtered.
(1) Image foreground background discrimination enhancement
Morphological filtering: morphological filtering is a common and effective image preprocessing method, and in the application process of the method, structural elements need to be specified first, and then the structural elements and an input image are subjected to convolution-like operation to obtain an output image. The difference between the calculation and the actual convolution operation is that the multiplication operation in the convolution operation is replaced by logic operation, namely, through the logic operation of the structural elements and the corresponding image pixels, the background clutter in the image is suppressed, the contrast of the image is improved, and the signal-to-noise ratio is improved.
The structural elements in morphological filtering refer to basic structures with a certain specific shape, such as: the rectangle, the circle or the rhombus with a certain size has an important influence on the filtering effect. The larger the template of the structural element, the larger the affected neighborhood in the image, and the larger the corresponding amount of computation. In addition, the shape, content, and operational properties of the structural elements also have an effect on the final output result. The morphological filtering applies both erosion transform and dilation transform algorithms. The erosion transform may "narrow" the target area, which essentially causes the image boundaries to shrink, and may be used to eliminate small and meaningless objects.
And (3) expansion transformation: roughly speaking, the expansion "enlarges" the target region range, incorporates background points in contact with the target region into the object, and expands the target boundary to the outside. The effect is to fill some holes in the target area and to eliminate small particle noise contained in the target area.
Besides expansion and corrosion, opening and closing operations are defined according to the sequence of combination of the expansion and the corrosion. Background suppression is achieved by adopting zafhat transformation, which is an important transformation that is an essential combination of opening and closing operations in mathematical morphology and a background difference technique. The specific application idea is that firstly, a background difference technology is utilized to enlarge a clustering center of a background and a foreground, then a proper structural element is selected to carry out open operation on a target area, and a small discrete target is removed; and then, completing the target edge by using closed operation, and adding a self-adaptive iteration number to finally obtain a target area image with large foreground background distinguishing degree. FIG. 2 is a first illustration of the erosion transform for the target area of the present invention. FIG. 3 is a graphical representation of the erosion transformation of the target area of the present invention. FIG. 4 is a diagram illustrating the expansion transformation of the target region according to the present invention. FIG. 5 is a diagram illustrating the expansion transformation of the target region according to the present invention.
(2) Filtering video frame images before the robot crawls at the pipe orifice
If the video frame images before the robot starts to crawl at the nozzle on the images can be filtered, a large amount of data processing time can be saved. Therefore, the distance information of the robot crawling inside the pipeline, which is displayed on the image, needs to be read so as to filter some image information which is irrelevant to detection.
After the images are distinguished and enhanced by the foreground and the background, a target area image with large distinguishing degree of the foreground and the background is obtained after a series of image processing, and characters in the target area image can be extracted to obtain shot characteristic information. Unlike traditional character recognition, the numbers in the pipeline detection image often have decimal points and are difficult to recognize. The existing character database is abandoned, a supervised learning method based on an improved SVM is used, a unique character database is retrained for detecting images, wherein the important point is the identification of decimal points, the decimal points are distinguished from other small targets which possibly appear by using the characteristics of a spatial domain, and the character database with better effect is obtained by training. The method comprises the following specific steps:
a. selecting a sample set of an area to be identified, wherein each number is equivalent to the decimal point selection number;
b. marking a real label on each sample data set, and extracting various characteristic information;
c. training a labeled sample data set by using an improved SVM algorithm as a classifier;
d. randomly selecting a test set to verify the training precision;
e. if the precision does not reach the standard, increasing training samples and repeating the operation of the step c and the step d;
f. and training to obtain a character detection model meeting the recognition precision.
(3) Video frame image for filtering motion state of robot
When a detector finds the internal defect of the pipeline, the detector needs to operate the robot to make the robot still and make the shooting lens still, if a local defect (such as fracture, corrosion and the like) is found, the detector makes the robot still, and simultaneously the lens looks around the interface of the pipeline to align the defect and observe the defect still, and if the detector filters out the video frame image of the lens motion state, the operation time of a large amount of image detection can be saved. Therefore, the video jitter detection technology based on image similarity is adopted, and the filtering principle is as follows:
if the similarity between a plurality of adjacent video frames is very high, the lens can be approximately considered not to move, and once the image similarity is low or even completely dissimilar, the camera can be judged to start moving. Therefore, in this way, we can judge whether the camera is moving or not. The traditional image similarity detection method mainly comprises a key point matching method, such as: SIFT, ORB, SURF, GIST extract the key point information, the accuracy of this kind of method is high, but at the same time, it is inevitable that it is comparatively complicated in computational complexity, the running speed is comparatively slow. In addition, algorithms such as histogram statistics, key point + decision tree and the like are provided. However, the operation speed is relatively slow, and the similarity detection under a large number of pictures cannot be met. Therefore, a perceptual hash algorithm is to be adopted for similarity detection. The perception hash algorithm generates a 'fingerprint' for each picture, and the fingerprints of the two pictures are compared to judge whether the similarity of the fingerprints belongs to the same picture. The method comprises three types of aHash, pHash and dHash, which have the characteristics, but the relative operation speed is faster than that of a key point feature extraction algorithm, so that the method is very suitable for the practical situation of the invention.
Step 1062: by using an image definition judging technology, extracting a key video frame image from the screened effective image, specifically comprising the following steps:
processing the screened effective image by adopting a definition discrimination algorithm based on a non-reference image, firstly carrying out Laplacian transformation processing on the effective image, solving the variance of the image after transformation processing, and distinguishing a clear image and a fuzzy image according to the variance, wherein the clear image is a key video frame image.
Because most of the lenses are moved in the pipeline detection process, a large number of video frames are distorted after framing, and the detection accuracy is greatly reduced if the distorted images are subjected to defect detection. In this case, after the regions where defects may exist are screened out and before defect detection, the candidate image must be subjected to sharpness detection to replace the distorted image. Aiming at the characteristics that the background is complex and changeable and the scene is switched at any time in the pipeline detection process, a definition discrimination algorithm based on no reference image is provided for extracting the key frame.
Fuzzy detection based on Laplacian operator
For a two-dimensional image f (x, y), the simplest definition of the second order differential — the laplacian is defined as:
for any order differential operator, a linear operator is used, so the second order differential operator and the following first order differential operator can be used to generate templates and then convolve the results.
According to the previous definition of the second order differential:
according to the above definition, in combination with the definition of the laplacian, we obtain:
that is, the calculation result of the laplacian operator for one point is the sum of the upper, lower, left, and right grays minus four times the gray itself. Likewise, all the gray values of the above formula are all added with negative signs, i.e., -1, -1, -1, 4, according to different definitions of the second order differential. Note, however, that the sign changes, and the addition or subtraction to the original image should be changed when sharpening. The four adjacent laplacian operators are arranged above, and the operators are rotated by 45 degrees and then added with the original operators to become eight-neighborhood operators, namely the difference between the sum of 8 pixels around one pixel and 8 times of the middle pixel is used as the laplacian calculation result.
Since abrupt changes (details) in the image are to be emphasized, the area of smooth gray scale, no response, i.e. the sum of the template coefficients is 0, is also a second order differential requirement.
The final sharpening formula:
g is the output, f is the original image, c is the coefficient, i.e. how much detail is to be added.
After the laplace transform, the edge portion of the image is emphasized and the rest is weakened. It is easy to think that a sharp picture representation is visually edge-salient, while a blurred image representation is visually edge-hard to recognize. The variance is thus determined for the laplace-transformed image. A picture with a large variance means that the edge is sharp, i.e. a sharp image, whereas a blurred image is obtained.
Calculating a variance formula: d (X) { [ X-E (X) ]]2}＝E{X2-2XE(X)+[E(X)]2}
Step 1063: and detecting the key video frame image by adopting a funnel mode and successively utilizing the global defect training model and the local defect training model to obtain a defect detection result, wherein the method specifically comprises the following steps:
detecting the key video frame images by adopting a global defect training model, if the key video frame images are globally normal, judging that the pipeline is normal, and if the key video frame images are abnormal, judging the type and other related information of the pipeline defect according to the probability distribution of the global defect type; and if the key video frame image is a local image, dividing the local image into nine local sub-images, and further detecting by adopting a local defect training model.
Dividing the local image into 9 local sub-images, wherein the 9 local sub-images can basically cover the whole local image and are overlapped with each other, and the set of the local sub-images is a local image; the local sub-images divided by the same local image are mutually associated; detecting local subimages by adopting a local defect training model, and if 9 local subimages corresponding to the local subimages are not abnormal, judging that no abnormality exists in the local subimages; and if any sub-image of the 9 local sub-images corresponding to the local image is abnormal, judging that the local image is abnormal, and judging the local image as the abnormal type with the largest occurrence frequency in the sub-image.
And after the detection is finished, outputting the corresponding pipeline starting point number, the corresponding pipeline ending point number, the corresponding defect type and the corresponding defect picture in the mode of result, storing and naming to obtain a defect detection result.
The method integrates the traditional image processing technology by combining the theoretical research basis of the aspects of target identification, image processing, image classification and the like, and realizes the algorithms and functions of metadata input, framing, labeling, training, detection, output and the like of the detected image. In addition, the invention integrates the deep learning algorithm, improves the theoretical value of the defect detection accuracy from the algorithm capability, and reflects the industrial application prospect of the deep learning algorithm from the side. The research can be applied to dredging detection and repair engineering of the drainage pipeline, the implementation period is shortened, the engineering cost is reduced, and the detection accuracy is improved.
FIG. 7 is a view showing the structure of the inspection system for internal defects of a drainpipe according to the present invention. A detection system for internal defects of drainage pipelines comprises:
the video data acquisition module 201 is configured to acquire video data for internal detection of a pipeline through field detection, where a part of the video data is randomly selected as a video for acquiring a sample, and another part of the video data is used as a test video, and the video for acquiring the sample and the test video are not overlapped with each other in a cross manner.
The first framing processing module 202 is configured to perform video framing processing on the video of the collected sample to obtain a framing processing result image of the video of the collected sample.
The sample library image acquisition module 203 is configured to acquire a sample library image according to a video framing processing result image of the acquired sample, where the sample library image includes a global defect image sample, a local defect image sample, a normal pipeline image sample, and a local normal pipeline image sample.
And the training module 204 is configured to perform model training according to the sample library image to obtain a training model, where the training model includes a global defect training model and a local defect training model.
And the second framing processing module 205 is configured to perform framing processing on the test video to obtain a framing processing result image of the test video.
And the detection module 206 is configured to adopt a dual-model accurate detection strategy to preprocess a framing result image of the test video, and adopt the global defect training model and the local defect training model to detect a preprocessed result to obtain a defect detection result.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (10)
1. A method for detecting internal defects of a drainage pipeline is characterized by comprising the following steps:
acquiring detection video data inside a pipeline through field detection, wherein one part of the video data is randomly selected as a video of a collected sample, the other part of the video data is used as a test video, and the video of the collected sample and the test video are not mutually crossed and overlapped;
performing video framing processing on the video of the collected sample to obtain a framing processing result image of the video of the collected sample;
acquiring a sample library image according to the video framing processing result image of the acquired sample, wherein the sample library image comprises a global defect image sample, a local defect image sample, a normal pipeline image sample and a local normal pipeline image sample;
performing model training according to the sample library images to obtain a training model, wherein the training model comprises a global defect training model and a local defect training model;
performing framing processing on the test video to obtain a framing processing result image of the test video;
and preprocessing a framing result image of the test video by adopting a dual-model accurate detection strategy, and detecting a preprocessed result by adopting the global defect training model and the local defect training model to obtain a defect detection result.
2. The method for detecting the internal defect of the drainage pipeline according to claim 1, wherein the step of performing video framing processing on the collected sample video to obtain a framing processing result image of the collected sample video specifically comprises;
adopt FFmpeg image fast framing technique to it carries out framing processing to the video of gathering the sample, the video framing processing's result is many resolutions and the super clear image that the size is the same, super clear image and former video size, resolution ratio are the same, including global map and local map, the global map is the image that the camera lens was taken at the inside direct vision place ahead of pipeline, the local map is the camera lens and looks around the local image that the department was shot at the setting in-process in the pipeline.
3. The method for detecting the internal defect of the drainage pipeline according to claim 1, wherein the step of collecting a sample library image according to the video framing processing result image of the collected sample specifically comprises the following steps:
judging the video framing processing result of the collected sample according to the definition of internal defects of the drainage pipeline by the national urban drainage pipeline detection and evaluation technical regulation, and selecting ten types of structural defect images of fracture, deformation, corrosion, dislocation, fluctuation, disjunction, interface material shedding, branch pipe hidden connection, foreign matter penetration and leakage and six types of functional defect images of deposition, scaling, obstacles, residual walls, dam roots, tree roots and scum as samples of a training model, wherein the global defect sample image is a large image with the same size as the original size of the video, the local defect image sample is a small image at the position and at the edge of a defect intercepted in a local image, the size is 416 x 416, and the resolution is the same as the original video;
selecting a normal pipeline image and a local normal pipeline image as samples of a training model, wherein the normal pipeline image sample is a high-definition large image with the original size of a video, the local normal pipeline image sample is a small image randomly intercepted from the local image, the size of the local normal pipeline image sample is 416 x 416, and the resolution is the same as that of the original video;
and combining the global defect image sample, the local defect image sample, the normal pipeline image sample and the local normal image sample to obtain a sample library image.
4. The method for detecting the internal defect of the drainage pipeline according to claim 1, wherein model training is performed according to the sample library image to obtain a training model, and specifically comprises the following steps:
and training a model by adopting the global defect image sample, the local defect image sample, the normal pipeline image sample and the local normal image sample to obtain a global defect training model and a local defect training model.
5. The method for detecting the internal defect of the drainage pipeline according to claim 1, wherein the step of performing the framing processing on the test video to obtain the framing processing result image of the test video specifically comprises the following steps:
adopt FFmpeg image fast framing technique right the test video carries out framing processing, and the video framing processing's result is many resolutions and the super clear image that the size is the same, super clear image and former video size, resolution ratio are the same, including global image and local image, global image is the image that the camera lens was taken in the inside direct vision place ahead of pipeline, local image is the camera lens and looks around the local image that the in-process was looked around in the pipeline and sets for the department shooting in the process.
6. The method for detecting the internal defect of the drainage pipeline according to claim 1, wherein the step of preprocessing the framing result image of the test video by using a dual-model precise detection strategy, and the step of detecting the preprocessed result by using the global defect training model and the local defect training model to obtain the defect detection result specifically comprises the steps of:
preliminarily screening the framing processing result of the test video by adopting a multi-domain joint discrimination method to obtain a screened effective image;
extracting key video frame images from the screened effective images by using an image definition judging technology;
and detecting the key video frame image by using the global defect training model and the local defect training model in sequence in a funnel mode to obtain a defect detection result.
7. The method for detecting the internal defects of the drainage pipeline according to claim 6, wherein the preliminary screening is performed on the framing processing result of the test video by adopting a multi-domain joint discrimination method to obtain a screened image, and specifically comprises the following steps:
and carrying out image foreground and background distinguishing and enhancing on the frame processing result of the test video, filtering a video frame image before the robot crawls at the pipe orifice and filtering a video frame image of the motion state of the robot, and obtaining a screened effective image.
8. The method for detecting the internal defects of the drainage pipeline according to claim 6, wherein a key video frame image is extracted from the screened effective image by using an image definition distinguishing technology, and the method specifically comprises the following steps:
processing the screened effective images by adopting a definition discrimination algorithm based on no reference image to obtain processed images;
determining a variance according to the processed image;
and distinguishing a clear image and a blurred image according to the size of the variance, wherein the clear image is a key video frame image.
9. The method for detecting the internal defect of the drainage pipeline according to claim 6, wherein the detecting is performed by using the global defect training model and the local defect training model in a funnel mode on the key video frame image to obtain a defect detection result, and specifically comprises:
detecting the key video frame images by adopting a global defect training model, if the key video frame images are globally normal, judging that the pipeline is normal, and if the key video frame images are abnormal, judging the type and other related information of the pipeline defect according to the probability distribution of the global defect type; if the key video frame image is a local image, dividing the local image into nine local sub-images, and further detecting by adopting a local defect training model;
dividing the local image into 9 local sub-images, wherein the 9 local sub-images can basically cover the whole local image and are overlapped with each other, and the set of the local sub-images is a local image; the local sub-images divided by the same local image are mutually associated; detecting local subimages by adopting a local defect training model, and if 9 local subimages corresponding to the local subimages are not abnormal, judging that no abnormality exists in the local subimages; if any sub-image of the 9 local sub-images corresponding to the local image is abnormal, judging that the local image is abnormal, and judging the local image as the abnormal type with the most occurrence times in the sub-image;
and after the detection is finished, outputting the corresponding pipeline starting point number, the corresponding pipeline ending point number, the corresponding defect type and the corresponding defect picture in the mode of result, storing and naming to obtain a defect detection result.
10. A detection system for internal defects of a drainage pipeline, comprising:
the video data acquisition module is used for acquiring detection video data inside the pipeline through field detection, wherein one part of the video data is randomly selected as a video of a collected sample, the other part of the video data is used as a test video, and the video of the collected sample and the test video are not mutually crossed and overlapped;
the first framing processing module is used for performing video framing processing on the video of the collected sample to obtain a framing processing result image of the video of the collected sample;
the sample library image acquisition module is used for acquiring sample library images according to the video framing processing result images of the acquired samples, wherein the sample library images comprise global defect image samples, local defect image samples, normal pipeline image samples and local normal pipeline image samples;
the training module is used for carrying out model training according to the sample library images to obtain a training model, and the training model comprises a global defect training model and a local defect training model;
the second framing processing module is used for framing the test video to obtain a framing processing result image of the test video;
and the detection module is used for preprocessing a framing result image of the test video by adopting a dual-model accurate detection strategy, and detecting a preprocessed result by adopting the global defect training model and the local defect training model to obtain a defect detection result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010343267.6A CN111553265B (en) | 2020-04-27 | 2020-04-27 | Method and system for detecting internal defects of drainage pipeline |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010343267.6A CN111553265B (en) | 2020-04-27 | 2020-04-27 | Method and system for detecting internal defects of drainage pipeline |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111553265A true CN111553265A (en) | 2020-08-18 |
| CN111553265B CN111553265B (en) | 2021-10-29 |
Family
ID=72004037
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010343267.6A Active CN111553265B (en) | 2020-04-27 | 2020-04-27 | Method and system for detecting internal defects of drainage pipeline |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111553265B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112233071A (en) * | 2020-09-28 | 2021-01-15 | 国网浙江省电力有限公司杭州供电公司 | Multi-granularity hidden danger detection method and system based on power transmission network picture in complex environment |
| CN113139507A (en) * | 2021-05-12 | 2021-07-20 | 保定金迪地下管线探测工程有限公司 | Automatic capturing method and system for drainage pipeline defect photos |
| CN113160210A (en) * | 2021-05-10 | 2021-07-23 | 深圳市水务工程检测有限公司 | Drainage pipeline defect detection method and device based on depth camera |
| CN113160176A (en) * | 2021-04-23 | 2021-07-23 | 歌尔股份有限公司 | Defect detection method and device |
| CN113606502A (en) * | 2021-07-16 | 2021-11-05 | 青岛新奥燃气设施开发有限公司 | Method for judging whether operator performs pipeline air leakage detection based on machine vision |
| CN113758662A (en) * | 2021-09-24 | 2021-12-07 | 长江三峡通航管理局 | Hydraulic hoist pipe connection leakproofness detecting system |
| CN114120209A (en) * | 2022-01-27 | 2022-03-01 | 深圳市博铭维技术股份有限公司 | Pipeline defect detection method, system, equipment and storage medium |
| CN114363499A (en) * | 2022-03-21 | 2022-04-15 | 深圳百胜扬工业电子商务平台发展有限公司 | Image processing method and device, computer equipment and storage medium |
| CN114511557A (en) * | 2022-04-02 | 2022-05-17 | 深圳市君合环保水务科技有限公司 | Image processing-based underdrain structure defect detection method |
| CN115115611A (en) * | 2022-07-21 | 2022-09-27 | 明觉科技（北京）有限公司 | Vehicle damage identification method and device, electronic equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107833220A (en) * | 2017-11-28 | 2018-03-23 | 河海大学常州校区 | Fabric defect detection method based on depth convolutional neural networks and vision significance |
| CN109767422A (en) * | 2018-12-08 | 2019-05-17 | 深圳市勘察研究院有限公司 | Pipe detection recognition methods, storage medium and robot based on deep learning |
| CN110598792A (en) * | 2019-09-16 | 2019-12-20 | 福州大学 | Drainage pipeline defect detection training data generation method based on PGGAN transfer learning |
-
2020
- 2020-04-27 CN CN202010343267.6A patent/CN111553265B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107833220A (en) * | 2017-11-28 | 2018-03-23 | 河海大学常州校区 | Fabric defect detection method based on depth convolutional neural networks and vision significance |
| CN109767422A (en) * | 2018-12-08 | 2019-05-17 | 深圳市勘察研究院有限公司 | Pipe detection recognition methods, storage medium and robot based on deep learning |
| CN110598792A (en) * | 2019-09-16 | 2019-12-20 | 福州大学 | Drainage pipeline defect detection training data generation method based on PGGAN transfer learning |
Non-Patent Citations (5)
| Title |
|---|
| YIXUAN LI等: "A Quantitative Analysis Method for Pipeline Defect Inspection Based on Infrared Thermal Imaging", 《IEEE》 * |
| 何志勇等: "一种微小表面缺陷的机器视觉检测方法", 《应用科学学报》 * |
| 吕兵等: "基于卷积神经网络的CCTV视频中排水管道缺陷的智能检测", 《测绘通报》 * |
| 李波锋: "基于机器视觉的排水管道缺陷检测算法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 韩继云: "《土木工程质量与性能检测鉴定加固》", 31 August 2010 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112233071A (en) * | 2020-09-28 | 2021-01-15 | 国网浙江省电力有限公司杭州供电公司 | Multi-granularity hidden danger detection method and system based on power transmission network picture in complex environment |
| CN113160176A (en) * | 2021-04-23 | 2021-07-23 | 歌尔股份有限公司 | Defect detection method and device |
| CN113160210A (en) * | 2021-05-10 | 2021-07-23 | 深圳市水务工程检测有限公司 | Drainage pipeline defect detection method and device based on depth camera |
| CN113139507A (en) * | 2021-05-12 | 2021-07-20 | 保定金迪地下管线探测工程有限公司 | Automatic capturing method and system for drainage pipeline defect photos |
| CN113606502A (en) * | 2021-07-16 | 2021-11-05 | 青岛新奥燃气设施开发有限公司 | Method for judging whether operator performs pipeline air leakage detection based on machine vision |
| CN113758662A (en) * | 2021-09-24 | 2021-12-07 | 长江三峡通航管理局 | Hydraulic hoist pipe connection leakproofness detecting system |
| CN113758662B (en) * | 2021-09-24 | 2024-03-12 | 长江三峡通航管理局 | Pipeline connection tightness detection system of hydraulic hoist |
| CN114120209A (en) * | 2022-01-27 | 2022-03-01 | 深圳市博铭维技术股份有限公司 | Pipeline defect detection method, system, equipment and storage medium |
| CN114363499A (en) * | 2022-03-21 | 2022-04-15 | 深圳百胜扬工业电子商务平台发展有限公司 | Image processing method and device, computer equipment and storage medium |
| CN114511557A (en) * | 2022-04-02 | 2022-05-17 | 深圳市君合环保水务科技有限公司 | Image processing-based underdrain structure defect detection method |
| CN114511557B (en) * | 2022-04-02 | 2022-09-13 | 深圳市君合环保水务科技有限公司 | Image processing-based underdrain structure defect detection method |
| CN115115611A (en) * | 2022-07-21 | 2022-09-27 | 明觉科技（北京）有限公司 | Vehicle damage identification method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111553265B (en) | 2021-10-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111553265B (en) | Method and system for detecting internal defects of drainage pipeline | |
| CN108460764B (en) | Ultrasonic image intelligent segmentation method based on automatic context and data enhancement | |
| Iyer et al. | Segmentation of pipe images for crack detection in buried sewers | |
| CN103093198B (en) | A kind of crowd density monitoring method and device | |
| Kim et al. | A fast and robust moving object segmentation in video sequences | |
| CN101179713A (en) | Method of detecting single moving target under complex background | |
| Ma et al. | Automatic defogging, deblurring, and real-time segmentation system for sewer pipeline defects | |
| Lian et al. | A novel method on moving-objects detection based on background subtraction and three frames differencing | |
| CN113780110A (en) | Method and device for detecting weak and small targets in image sequence in real time | |
| CN107346547A (en) | Real-time foreground extracting method and device based on monocular platform | |
| Arya et al. | Real-time vehicle detection and tracking | |
| Zuo et al. | Classifying cracks at sub-class level in closed circuit television sewer inspection videos | |
| KR101243294B1 (en) | Method and apparatus for extracting and tracking moving objects | |
| CN111597992B (en) | Scene object abnormity identification method based on video monitoring | |
| Selvaraju et al. | Support Vector Machine based Remote Sensing using Satellite Data Image | |
| CN116883893A (en) | Tunnel face underground water intelligent identification method and system based on infrared thermal imaging | |
| Babu et al. | A feature based approach for license plate-recognition of Indian number plates | |
| KR20010035100A (en) | An Effective Object Tracking Method and for Apparatus for Interactive HyperLink Video | |
| Das et al. | Automatic License Plate Recognition Technique using Convolutional Neural Network | |
| Chaiyawatana et al. | Robust object detection on video surveillance | |
| Lyu et al. | Scene-Adaptive Real-Time Fast Dehazing and Detection in Driving Environment | |
| CN116416270B (en) | Method and device for monitoring underground working surface leakage water | |
| CN114821795B (en) | Personnel running detection and early warning method and system based on ReiD technology | |
| CN116994160A (en) | Slope disease detection method, device and system based on unmanned aerial vehicle acquired image | |
| Liu et al. | Research on visual background extractor to identify the vehicles based on edge similarity |
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 |