CN111738342A - Pantograph foreign matter detection method, storage medium and computer equipment - Google Patents
Pantograph foreign matter detection method, storage medium and computer equipment Download PDFInfo
- Publication number
- CN111738342A CN111738342A CN202010589402.5A CN202010589402A CN111738342A CN 111738342 A CN111738342 A CN 111738342A CN 202010589402 A CN202010589402 A CN 202010589402A CN 111738342 A CN111738342 A CN 111738342A
- Authority
- CN
- China
- Prior art keywords
- contour
- image
- threshold
- foreign matter
- outline
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 56
- 238000012545 processing Methods 0.000 claims abstract description 113
- 238000000034 method Methods 0.000 claims abstract description 34
- 230000000877 morphologic effect Effects 0.000 claims abstract description 30
- 238000012549 training Methods 0.000 claims abstract description 14
- 230000006870 function Effects 0.000 claims description 40
- 230000009467 reduction Effects 0.000 claims description 26
- 238000004364 calculation method Methods 0.000 claims description 14
- 238000001914 filtration Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 12
- 238000009432 framing Methods 0.000 claims description 9
- 230000007797 corrosion Effects 0.000 claims description 5
- 238000005260 corrosion Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 238000005530 etching Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 238000012847 principal component analysis method Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000012216 screening Methods 0.000 abstract description 15
- 238000003672 processing method Methods 0.000 abstract description 2
- 239000011159 matrix material Substances 0.000 description 25
- 206010070245 Foreign body Diseases 0.000 description 23
- 239000013598 vector Substances 0.000 description 12
- 238000000513 principal component analysis Methods 0.000 description 7
- 238000012544 monitoring process Methods 0.000 description 5
- 230000002159 abnormal effect Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000000354 decomposition reaction Methods 0.000 description 3
- 230000009977 dual effect Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 238000013145 classification model Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 241001270131 Agaricus moelleri Species 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 235000012149 noodles Nutrition 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
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
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G06T5/70—
-
- 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
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Abstract
The invention discloses a pantograph foreign matter detection method, a storage medium and computer equipment, and relates to the technical field of pantograph-catenary detection. The method comprises the steps of carrying out binarization processing on an input image, carrying out morphological processing on the image after the binarization processing, carrying out denoising processing on the image after the morphological processing, extracting a contour from the image after the denoising processing, calculating contour parameters, carrying out multi-dimensional threshold value screening on the contour parameters and carrying out gray parameter threshold value screening on a contour original gray image, so as to remove a part of non-foreign-matter images, then carrying out SVM classification processing, and finally identifying the pantograph foreign-matter image. The pantograph foreign matter detection method can adapt to the diversity of foreign matters and the complexity of an image background, determines the foreign matter candidate frame by an image processing method before training the SVM model, and filters partial candidate frames by morphological screening and gray feature screening, so that the missing detection rate is low, and the classification efficiency and precision of the SVM are improved.
Description
Technical Field
The invention relates to the technical field of pantograph and catenary detection, in particular to the technical field of pantograph and pantograph detection, and more particularly relates to a pantograph and foreign matter detection method, a storage medium and computer equipment.
Background
All rail transit trains acquire electric energy from the outside through sliding contact between a train pantograph and a contact network, and the contact relation between the pantograph and the contact network is important for safe and normal operation of the trains. Due to the special contact relation and the structural characteristics of the pantograph and the contact network, the pantograph or the contact network of the operation train is not convenient to install test equipment, and once the operation train is in serious abnormal contact with the pantograph network or is attached with foreign matters, the operation train is also not convenient to check the train roof in a line section, and the train roof needs to be checked by means of special attachment equipment or the train is maintained to operate by changing the pantograph, so that the train is directly influenced to operate at a normal point.
In the prior art, the maintenance and monitoring of the running state of the pantograph are mainly performed by manually checking and registering by special technical personnel after a train is put in storage every day, and whether the overall structure size and the lifting state of the pantograph are normal or not is judged. However, this maintenance method can only ensure whether the pantograph itself is significantly abnormal, and cannot ensure that a normal pantograph-catenary contact relationship can be achieved after the pantograph is put into operation. In addition, a high-definition camera can be arranged on the roof of the train to record the pantograph state in the running process of the train, but the abnormal state can be checked only in a manual video playback mode, and the pantograph-catenary contact relation state and whether foreign matters are impacted or attached cannot be evaluated in real time.
At present, foreign matter detection of a pantograph is mainly performed by means of image recognition. The current contact net and the regional foreign matter detection of pantograph slide discernment mainly have following shortcoming: (1) the recognition rate is low, and different foreign matters are difficult to recognize by singly adopting the traditional image processing mode due to the variety of the types of the foreign matters; (2) the false alarm rate is high, and because the foreign matter form is various, there are infinite possible foreign matters to appear in theory, and the foreign matters are very likely to be very similar to the background of the detection image, and the image background is relatively complicated simultaneously, if want to use traditional mode to discern as many foreign matters as possible, will cause a large amount of false recognitions.
Disclosure of Invention
In order to overcome the defects and shortcomings in the prior art, the application provides a pantograph foreign matter detection method, and aims to solve the problems of low foreign matter identification rate, high false alarm rate and large image data processing amount in the prior art. According to the pantograph foreign matter detection method, binarization processing is carried out on an input image, morphological processing is carried out on the image after binarization processing, denoising processing is carried out on the image after morphological processing, the outline is extracted from the image after denoising processing, outline parameters are calculated, multi-dimensional threshold value screening is carried out on the outline parameters, gray parameter threshold value screening is carried out on an original gray image of the outline, therefore, a part of non-foreign matter images are removed, then SVM classification processing is carried out, and finally a pantograph foreign matter image is identified. The pantograph foreign matter detection method can adapt to the diversity of foreign matters and the complexity of the image background, partial images are judged and filtered through two threshold values, the calculation amount of SVM model classification is greatly reduced, the missing detection rate is lowered, and the classification efficiency and precision of the SVM are improved.
In order to solve the problems in the prior art, the invention is realized by the following technical scheme:
a pantograph foreign matter detection method comprises the following steps:
an image processing step: collecting and reading a gray image of a pantograph slide bar area, performing binarization processing on the image of the pantograph slide bar area by adopting a local self-adaptive binarization method, performing morphological processing on the image obtained after the binarization processing, performing denoising processing on the image after the morphological processing, searching a connected domain in the image after the denoising processing, and outputting the outer boundary of the connected domain as a contour; calculating outline parameters of the output outline, wherein the outline parameters comprise an outline area, an outline area-to-outline frame area ratio, an outline width, an outline height and an outline width-to-height ratio;
a threshold I judging step, namely comparing and judging the contour parameters calculated in the image processing step with a preset contour parameter threshold, if all the contour parameters of the output contour are within the contour parameter threshold, judging the contour as a suspected foreign body contour and outputting the contour to the next step, and if any contour parameter is not within the corresponding contour parameter threshold, judging the contour as a non-foreign body contour; if the contours in the input image are all non-foreign body contours, updating data and inputting the next image, and repeating the image processing step and the step for detection;
a threshold II judging step, namely corresponding the suspected foreign matter contour judged in the threshold I judging step to an original input image corresponding to the suspected foreign matter contour, and calculating a pixel gray level mean value and a variance in the corresponding contour in the original input image; respectively comparing the calculated gray level mean value and the calculated variance with a preset gray level mean value threshold value and a preset variance threshold value, and if the gray level mean value and the variance of the suspected foreign matter outline are both located in the gray level mean value threshold value and the variance threshold value, judging that the suspected foreign matter outline is an approximate foreign matter outline; if any element of the gray mean value and the variance does not meet the gray mean value threshold and the variance threshold, judging that the suspected foreign matter outline is a non-foreign matter outline; if the suspected foreign body contours in the input image are all non-foreign body contours, updating data and inputting the next image, and repeating the image processing step, the threshold I judging step and the step for detection;
an SVM classification step, namely performing dimensionality reduction on the sample image by using a principal component analysis method, and training by using data subjected to dimensionality reduction to obtain an SVM classification decision function; and (3) corresponding the approximate foreign matter outline judged in the threshold II judging step to the original image, setting an interested area according to the position of the approximate foreign matter outline frame in the original image, and substituting the interested area into an SVM classification decision function to perform secondary classification of foreign matters or non-foreign matters.
In the threshold value I judging step, the preset contour parameter threshold values comprise a contour area threshold value, a contour area-to-contour frame area ratio threshold value, a contour width threshold value, a contour height threshold value and a contour width-to-height ratio threshold value; the contour parameter threshold, the gray mean threshold and the variance threshold are obtained by manually framing a foreign object area in each image in a plurality of sample images, then obtaining the contour parameter, the gray mean and the variance of the foreign object contour through an image processing step and carrying out statistics by using statistics.
The preset contour parameter threshold, the preset gray level mean threshold and the preset variance threshold in the threshold I judging step and the preset contour parameter threshold in the threshold II judging step are obtained through the following processes:
manually framing a foreign matter area in each image in a plurality of sample images, then carrying out binarization processing, morphological processing, denoising processing and contour extraction processing to obtain a plurality of small foreign matter images, and then carrying out contour parameter calculation on the contours of the small foreign matter images to calculate a contour area AckThe ratio AR of the area of the outline to the area of the outline frameckWide profile RWckHigh profile RHckAnd profile aspect ratio RWck/RHckCalculating the pixel gray level mean value and variance of the corresponding contour of the foreign matter small image contour in the original image through a gray level parameter calculation step, carrying out statistics by utilizing statistics to obtain a maximum statistic value MaxSto and a minimum statistic value MinSto of each parameter, and adding redundancy, namely ThrMax ═ MaxSto (1+ Ratio1) and ThrMinMinMin ═ MinSto (1-Ratio2) into the maximum statistic value MaxSto and the minimum statistic value MinSto obtained by statistics, wherein Ratio1 ∈ (0,1) and Ratio2 ∈ (0,1) are obtained, so that the contour area and the contour surface are obtainedThe threshold values of product to outline box area ratio, outline width, outline height, outline width height ratio, gray mean and gray variance.
In the parameter threshold values in the threshold value I judging step and the threshold value II judging step, the Ratio1 and the Ratio2 are 0.05, 0.1 or 0.2. In the parameter threshold values in the threshold value I judging step and the threshold value II judging step,wherein StofThe parameter values satisfying the maximum frequency are distributed statistically.
The specific training process of the SVM classification decision function is as follows:
(a) manually framing a foreign matter area and a non-foreign matter area on an existing pantograph image sample, generating positive and negative samples to be trained, and re-adjusting the size of each sample to be M x N, wherein the set of samples to be trained is D ═ D0,D1,...,Di,...,DI-1]The sample label is y ═ y0,y1,...,yi,...,yI-1]Total of I samples, and yi∈ { -1, +1}, -1 denotes a non-foreign matter sample, +1 denotes a foreign matter sample;
(b) will DiReconstructing into head-connected M x N dimensional column vectors x by rowsiSo that the set of samples to be trained can be written as X ═ X1,x2,...,xI-1];
(c) The PCA dimensionality reduction matrix is obtained by utilizing the data set X, the central vector u of the data set is firstly calculated,then, each sample is subtracted by the central vector to obtain a de-centered sampleWhereinThen the covariance matrix of the center-removed sample is subjected to eigenvalue decomposition,wherein λ ═ λ0,λ1,...,λM*N-1]Is an eigenvalue of the covariance matrix, and0>λ1>...>λM*N-1,Q=[q1,q2,...,qM*N-1]is an eigenvector matrix of a covariance matrix, andicorresponds to qiThen setting a threshold value Tλ=λ0Rat, Rat ∈ (0,1), then selecting greater than TλForming a dimensionality reduction matrix by the eigenvectors corresponding to the eigenvalues;
(d) using a dimensionality reduction matrix to pair the depocenter sample sets according to the following formulaAnd (3) performing dimensionality reduction:whereinThe dimension of the data matrix after dimension reduction is K x I, K<<M*N;
s.t.1-yi(wTφ(xi)+b)-ξi≤0,ξi0, I ≧ 0,1,2, 1, I-1, where w, b are parameters for classifying hyperplanes, ξiThe soft interval parameter is C, the penalty coefficient is C, phi (-) is a mapping relation, and phi (-) is operated through a kernel function; by constructing the lagrange function and the dual transformation, the above model is transformed as follows:
s.t.λi≥0,μi≥0;
by calculating the partial derivatives for each parameter and making them 0, we can obtain:
C=λi+μi,
and substituting the above relation into Lagrange function to obtain optimized model,
wherein phi (x)i)·φ(xj) Calculating by a kernel function, solving by using an SMO algorithm to obtain lambda, and selecting lambda satisfying that C is more than or equal to lambdajA component λ of 0 or morejAnd calculating:
in the image processing step, binarization processing is carried out on the pantograph slide bar area image, specifically: dividing an input image into a plurality of non-overlapping small blocks, and carrying out binarization on each small block according to the corresponding gray scale change characteristic of each small block; the specific process is as follows:
the input image I is divided into 4 non-overlapping small blocks, denoted as IA、IB、ICAnd IDThen E isA=f(IA,thrA)
i=0,1,2,...,IAH-1,j=i=0,1,2,...,IAW-1; wherein E isAIs IAImage after binarization, thrAFor threshold values used in binarization, the threshold value being in accordance with IAObtaining a gray level histogram of (1)AHIs IAHigh of (I)AWIs IAIs wide; in the same way, E can be obtainedB、EC、EDAnd then the binary images E of the input image I are obtained by splicing the images together in a manner of segmentation. The adoption of the local self-adaptive binarization method to obtain the binarized image E overcomes the influence of the brightness and contrast difference of different areas of the image on the binarized extraction target caused by different illumination conditions at different time on the image obtained by on-line pantograph monitoring, and provides a foundation for effectively screening foreign body images by subsequent morphological processing.
In the image processing step, morphological processing comprises image corrosion, image expansion, opening operation and closing operation; in the image processing step, the image obtained after the binarization processing is subjected to morphological processing, specifically, the image obtained after the binarization processing is subjected to two opening operations, and is subjected to one etching to obtain a processed image M.
In the image processing step, the mode of denoising the image after morphological processing is median filtering or Gaussian filtering.
In the SVM classification step, phi (-) is operated by a kernel function, and phi (x)i)·φ(xj) The kernel function in the operation by the kernel function includes one of a linear kernel, a polynomial kernel, and a gaussian kernel.
In the SVM classification step, phi (x)i)·φ(xj) The operation is carried out through a kernel function, and the kernel function adopts a Gaussian kernel function:
a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor performing the steps of the above method when executing the computer program.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method.
Compared with the prior art, the beneficial technical effects brought by the application are shown in that:
1. the invention combines the traditional image processing and machine learning methods, and has the advantages of high detection accuracy, low false detection rate, low missing detection rate and high running speed. Before training an SVM model, a foreign matter candidate frame is determined in an image processing mode, and then part of candidate frames are filtered by morphological screening and gray feature screening, wherein the part of candidate frames are processed on the premise of ensuring that non-foreign matter candidate frames are filtered as far as possible, so that the missing detection rate is low, and the classification efficiency and precision of the SVM are improved.
2. According to the method, a comprehensive threshold judgment method is adopted to filter out part of non-foreign matter candidate frames, the processing data volume is reduced for the subsequent SVM classification model processing, and the SVM classification efficiency and precision can be effectively improved. Meanwhile, the parameter threshold value is obtained by data mining and statistics of a large amount of sample data, and in order to avoid filtering foreign matters, a certain redundancy is increased when the threshold value is set, so that the risk of mistakenly filtering the foreign matters is reduced as much as possible. Meanwhile, the threshold value can be adaptively modified according to the monitoring images of different scenes, so that the method, the system or the computer equipment or the computer program in the storage medium can be well adapted to the scene switching, and the foreign matter detection accuracy and detection efficiency, such as day and night scene switching, can be ensured. In addition, the threshold expansion of the redundancy is combined with the statistical distribution of the sample parameters, so that the effects of considering screening efficiency and retaining effective abnormal data are achieved.
3. According to the method, Principal Component Analysis (PCA) is combined in the process of SVM classification establishment to train the data to be classified, the SVM model is trained by using the data after dimensionality reduction, and the training efficiency of the SVM model is improved on the basis of ensuring the recognition accuracy of the SVM model.
4. The invention sets the judgment threshold value by adopting a training statistic mode, can adapt to the diversity of foreign matters and the complexity of an image background, is convenient to modify algorithm parameters after being set into a computer program, and can better adapt to the detection of the current environment by only modifying the judgment threshold value and a corresponding SVM classification model. The modification of the relevant parameters can be carried out by other non-specialized persons without specialized persons.
5. According to the method, threshold judgment is combined with SVM classification, preliminary rough selection is carried out in the threshold judgment step, SVM classification is accurately identified, and if SVM classification identification is directly carried out, the large amount of SVM classification processing data is caused, the processing efficiency is low, and pantograph foreign matter identification in real-time online monitoring cannot be realized; and the foreign matter cannot be accurately identified by adopting the image characteristic threshold value alone for judgment, so that the condition of false detection is easy to occur. According to the method and the device, threshold judgment and SVM classification are combined, the detection accuracy is high, the false detection rate is low, the missing detection rate is low, and the running speed is high.
6. Because the background of the detection image containing the pantograph is relatively complex, the foreign object image is likely to have similarity with the background image, and the foreign object image can be effectively highlighted by adopting a mode of firstly carrying out opening operation twice and then carrying out corrosion once.
Drawings
Fig. 1 is a flowchart of a pantograph foreign object detection method according to the present invention.
Detailed Description
The technical scheme of the application is further elaborated in the following by combining the drawings in the specification.
Referring to the accompanying drawing 1 of the specification, the embodiment discloses a pantograph foreign matter detection method, and the specific flow is described as follows:
image input: and inputting a gray image of the pantograph slider area, and marking as I.
Image binarization: the local self-adaptive binarization method is adopted, namely, an input image is divided into a plurality of non-overlapping small blocks, and each small block is binarized according to the corresponding gray scale change characteristic of the small block.
For example, the input image I is divided into 4 non-overlapping small blocks, denoted as IA、IB、IC、IDThen E isA=f(IA,thrA)
i=0,1,2,...,IAH-1,j=0,1,2,...,IAW-1, wherein EAIs IAImage after binarization, thrAFor the threshold used in binarization, the threshold may be according to IAObtaining a gray level histogram of (1)AHIs IAHigh of (I)AWIs IAIs wide. In the same way, E can be obtainedB,EC,EDThen, the input image I is pieced together in the manner of segmentation to obtain a binarized image E of the input image I.
Morphological treatment: and performing morphological processing on the binary image E, wherein the morphological processing mainly comprises image corrosion, expansion, opening operation, closing operation and the like. The above operations can be used in combination according to actual conditions, for example, the binary image is first subjected to two opening operations and then subjected to one etching, and the output image is recorded as M.
Median filtering: and further performing drying treatment on the morphologically processed image. Median filtering is used herein, and the filtered image is recorded as
Extracting the contour: in thatSearching for connected domain, and outputting the outer boundary of the connected domain as contour, and recording a certain contour as ck。
Calculating profile parameters: including profile area, ratio of profile area to profile frame area, profile width, profile height, and profile width to height ratio. In detail, the outline area calculation mode is to count the number of pixel points in the outline, and is recorded as Ack。
The outline frame is an external rectangle of the outline, theoretically, the external rectangles of the outline have infinite number, the rectangle sides are simply selected as horizontal and vertical external rectangles respectively to carry out subsequent calculation, the area of the external rectangles is equal to the product of the width and the height of the outline frame, therefore, the calculation mode of the area ratio of the outline area to the outline frame is that the outline area is divided by the outline frame area and is marked as ARck。
Width of the profile, i.e. the width of the profile frame, denoted RWckI.e. the horizontal distance of the rightmost and leftmost pixels of the contour; height of the profile, i.e. the height of the profile frame, denoted RHckI.e. the vertical distance between the uppermost and lowermost pixel points of the profile; the aspect ratio is calculated by dividing the aspect ratio by the aspect ratio.
Whether the parameter satisfies a threshold i: the step is to primarily screen the extracted contours and filter out the contours which are obviously not foreign matters. If a certain contour area is between (area min, area max), the ratio of the contour area to the contour frame area is between (area ratio min, area ratio max), the contour width is between (width min, width max), the contour height is between (HeiMin, HeiMax), and the contour width-height ratio is between (width heiratimin, width HeiMax), then it is determined that the contour is a foreign object contour, if any one of the above conditions is not satisfied, it is determined that the contour is a non-foreign object contour, and the foreign object contour is directly removed, as a statistical determination interval herein: (area min, area max) ═ (387, 20471), (area ratio min, area ratio max) ═ (0.34, 0.83), (HeiMin, HeiMax) ═ (18, 163), (widheiratio min, widhei ratio max) ═ 0.27, 3.91.
If all the contours extracted from the input image are removed, the updated data is input into the next image for detection. If the contour passes the primary screening, the next calculation is carried out. The threshold values are obtained through statistics, for example, a large number of images are processed to the step of extracting the contour, then the foreign body contour and the non-foreign body contour are manually distinguished, then the parameters are respectively calculated, and the corresponding optimal threshold values are set according to the calculated parameter values, wherein the threshold values can allow the non-foreign bodies to be judged as foreign bodies, but real foreign bodies cannot be missed as far as possible, the method is just primary screening, and the subsequent steps are further screened and finally judged.
Calculating a gray level parameter: and corresponding the preliminarily screened contours to an original input image, and then calculating the mean value and the variance of the pixel gray scales in the corresponding contours in the original input image.
Whether the parameter satisfies a threshold value II: the step is to further filter out non-foreign body contours according to the gray scale parameters. If the mean value of the gray levels in a certain contour is between (AveIm, AveMax) and the variance of the gray levels is between (VarMin, VarMax), the contour is further judged to be a foreign body contour, if any condition of the two conditions is not met, the contour is judged to be a non-foreign body contour, and the contour is directly removed. If all the contours input in the step are completely removed, judging that no foreign object exists in the input image, and then updating data and inputting the updated data into the next image for detection. If there are contours that pass this screening, the next step is performed. Similarly, the threshold of the filtering is obtained by statistical means, and the specific manner can refer to the threshold statistics of the first filtering. The threshold setting of this step also allows some non-foreign object profiles to pass the screening while trying to ensure that the true foreign object profile is not missed, as in this document (AveMin, AveMax) ═ 116, 203), (VarMin, VarMax) ═ 3.66, 13.42.
The specific statistical modes of the parameter threshold I and the parameter threshold II are as follows:
according to 36925 pantograph images obtained by monitoring at a certain time, manually framing a foreign matter area in each image to obtain 2189 foreign matter small images, then respectively calculating the outline area, the outline area to outline frame area ratio, the outline width, the outline height, the outline width height ratio, the gray level mean value in the outline and the gray level variance in the outline of each small image according to the calculation mode, further obtaining the maximum statistic MaxSto and the minimum statistic MinSto of each parameter, and adding partial redundancy in order to reduce the risk of filtering foreign matters as much as possible, so that the maximum and minimum filtering threshold values of the parameter interval are set as follows:
ThrMax=MaxSto*(1+Ratio1)
ThrMin=MinSto*(1–Ratio2)
wherein Ratio1 ∈ (0,1)Ratio2 ∈ (0,1), and is usually set to be small, such as 0.05, 0.1, 0.2, etc., in the parameter threshold values in the threshold value I judging step and the threshold value II judging step,wherein StofThe parameter values satisfying the maximum frequency are distributed statistically.
SVM classification: and corresponding the contours after the two-time screening to the original image, setting a region of interest (ROI) in the original image according to a contour frame, and then carrying out secondary classification on foreign matters or non-foreign matters on the ROI.
In the SVM classification process, Principal Component Analysis (PCA) is combined to firstly perform dimensionality reduction on classification data, and the specific training process is as follows:
1) according to the existing pantograph image samples, a foreign matter area and a non-foreign matter area are manually framed respectively, positive and negative samples to be trained are generated, the size of each sample is adjusted to be M x N again, and the set of samples to be trained is recorded as D ═ D0,D1,...,Di,...,DI-1]And the sample label is marked as y ═ y0,y1,...,yi-1,...,yI-1]Total of I samples, and yi∈ { -1, +1}, -1 denotes a non-foreign matter sample, +1 denotes a foreign matter sample;
2) will DiRow-wise reconstruction as an end-to-end M x N dimensional column vector xiSo that the set of samples to be trained can be written as X ═ X1,x2,…,xI-1];
3) The PCA dimensionality reduction matrix is obtained by utilizing the data set X, the central vector u of the data set is firstly calculated,
Then the covariance matrix of the center-removed sample is subjected to eigenvalue decomposition,
wherein λ ═ λ0,λ1,...,λM*N-1]Is an eigenvalue of the covariance matrix, and0>λ1>...>λM*N-1,Q=[q1,q2,...,qM*N-1]is an eigenvector matrix of a covariance matrix, andicorresponds to qiThen setting a threshold value Tλ=λ0Rat, Rat ∈ (0,1), then selecting greater than TλThe eigenvectors corresponding to the eigenvalues of (a) form a dimension reduction matrix. In addition, the eigenvectors corresponding to eigenvalues greater than 0 may also be directly selected or selected according to a certain proportion of the original dimension M × N, for example, the eigenvectors corresponding to the first int (M × N × Rat) eigenvalues are selected as the dimension reduction matrix, which is marked as P, and K eigenvectors selected according to the above-mentioned manner are added, so that P ═ q is added0,q1,...,qK],K<<M*N;
4) Using a dimensionality reduction matrix to pair the depocenter sample sets according to the following formulaAnd (3) performing dimensionality reduction:whereinThe dimension of the data matrix after dimension reduction is K x I, K<<M*N;
s.t.1-yi(wTφ(xi)+b)-ξi≤0,ξi≥0,i=0,1,2,...,I-1
where w, b are parameters for classifying hyperplanes, ξiThe soft interval parameter is C, the penalty coefficient is C, phi (-) is a certain mapping relation, and phi (-) can be directly operated subsequently through a kernel function;
by constructing the lagrangian function and the dual transformation, the above model can be transformed as follows:
s.t.λi≥0,μi≥0
by calculating the partial derivatives for each parameter and making them 0, we can obtain:
C=λi+μi,
and the above-mentioned relation is substituted into Lagrange function, then the optimized model can be obtained
Wherein phi (x)i)·φ(xj) The operation can be performed through kernel functions, which are commonly used as linear kernels, polynomial kernels and gaussian kernels, and the gaussian kernel function is used here:
finally, solving by using an SMO algorithm to obtain lambda, and then selecting lambda satisfying 0 & ltlambdajComponent lambda of < CjAnd calculating:wherein SV is a set of support vectors. Finally, a decision function can be constructed:the decision function is brought in for the sample x to be measured, if f (x) is more than 0, the sample x is foreign matter, otherwise, the sample x is non-foreign matter.
Example 2
Referring to the attached fig. 1, this embodiment discloses:
a pantograph foreign matter detection method comprises the following steps:
an image processing step: collecting and reading a gray level image I of a pantograph slide bar area, and performing binarization processing on the image of the pantograph slide bar area by adopting a local self-adaptive binarization method to obtain an image E after binarization processing; performing morphological processing on the image E obtained after the binarization processing to obtain an image M; denoising the image M after the morphological processing to obtain an imageImage after de-noising processingSearching a connected domain, outputting the outer boundary of the connected domain as a contour, and outputting a certain contour ck(ii) a Calculating the contour parameters of the output contour, wherein the contour parameters comprise a contour area AckThe ratio AR of the area of the outline to the area of the outline frameckWide profile RWckHigh profile RHckAnd profile aspect ratio RWck/RHck。
A threshold I judging step of judging the contour parameters calculated in the image processing stepComparing and judging with a preset contour parameter threshold, if all contour parameters of the output contour are within the contour parameter threshold, judging the contour as a suspected foreign body contour and outputting the suspected foreign body contour to the next step, and if any contour parameter is not within the corresponding contour parameter threshold, judging the contour as a non-foreign body contour; if the contours in the input image are all non-foreign body contours, updating data and inputting the next image, and repeating the image processing step and the step for detection; namely, the contour area A calculated in the contour parameter calculation stepckThe ratio AR of the area of the outline to the area of the outline frameckWide profile RWckHigh profile RHckAnd profile aspect ratio RWck/RHckMaking a judgment if a certain contour ckArea A of the contour ofckBetween (AreaMin, AreaMax), the ratio AR of the outline area to the outline box areackBetween (area ratio min, area ratio max), the profile width RWckBetween (WidthMin, WidthMax), the profile is high RHckBetween (HeiMin, HeiMax), the profile aspect ratio RWck/RHckAt (WidHeiRatioMin, WidHeiRatioMax), the profile c is judgedkIs a foreign matter candidate outline; if the condition is not met, judging that the contour is a non-foreign-body contour, and directly rejecting the contour; if all the contours extracted from the input image are removed, updating data and inputting the next image for detection; and if the contour extracted from the input image meets the determined foreign matter candidate contour, performing the next step.
A threshold II judging step, wherein the foreign matter candidate contour judged in the threshold I judging step is corresponding to the corresponding original input image, and the pixel gray level mean value and the variance in the corresponding contour are calculated in the original input image; respectively comparing the calculated gray level mean value and the calculated variance with a preset gray level mean value threshold value and a preset variance threshold value, and if the gray level mean value and the variance of the suspected foreign matter outline are both located in the gray level mean value threshold value and the variance threshold value, judging that the suspected foreign matter outline is an approximate foreign matter outline; if any element of the gray mean value and the variance does not meet the gray mean value threshold and the variance threshold, judging that the suspected foreign matter outline is a non-foreign matter outline; if the suspected foreign body contours in the input image are all non-foreign body contours, updating data and inputting the next image, and repeating the image processing step, the threshold I judging step and the step for detection;
judging the pixel gray mean value and variance of the foreign object candidate contour in the original input image corresponding to the contour calculated in the gray parameter calculation step, and if the gray mean value is between (AveIm, AveMax) and the gray variance is between (VarMin, VarMax), further judging that the contour is the foreign object contour; if any one of the two conditions is not met, the contour is judged to be a non-foreign-body contour, and the foreign-body contour is directly removed; if all the contours input in the step are completely removed, judging that the input image has no foreign objects, and then updating data and inputting the next image for detection; if the contour passes the judgment of the step, the next step is carried out.
An SVM classification step, namely performing dimensionality reduction on the sample image by using a principal component analysis method, and training by using data subjected to dimensionality reduction to obtain an SVM classification decision function; and (3) corresponding the approximate foreign matter outline judged in the threshold II judging step to the original image, setting an interested area according to the position of the approximate foreign matter outline frame in the original image, and substituting the interested area into an SVM classification decision function to perform secondary classification of foreign matters or non-foreign matters.
Example 3
Referring to the attached fig. 1, this embodiment discloses:
a pantograph foreign matter detection method comprises the following steps:
an image processing step: collecting and reading a gray image of a pantograph slide bar area, performing binarization processing on the image of the pantograph slide bar area by adopting a local self-adaptive binarization method, performing morphological processing on the image obtained after the binarization processing, performing denoising processing on the image after the morphological processing, searching a connected domain in the image after the denoising processing, and outputting the outer boundary of the connected domain as a contour; calculating outline parameters of the output outline, wherein the outline parameters comprise an outline area, an outline area-to-outline frame area ratio, an outline width, an outline height and an outline width-to-height ratio;
a threshold I judging step, namely comparing and judging the contour parameters calculated in the image processing step with a preset contour parameter threshold, if all the contour parameters of the output contour are within the contour parameter threshold, judging the contour as a suspected foreign body contour and outputting the contour to the next step, and if any contour parameter is not within the corresponding contour parameter threshold, judging the contour as a non-foreign body contour; if the contours in the input image are all non-foreign body contours, updating data and inputting the next image, and repeating the image processing step and the step for detection;
a threshold II judging step, namely corresponding the suspected foreign matter contour judged in the threshold I judging step to an original input image corresponding to the suspected foreign matter contour, and calculating a pixel gray level mean value and a variance in the corresponding contour in the original input image; respectively comparing the calculated gray level mean value and the calculated variance with a preset gray level mean value threshold value and a preset variance threshold value, and if the gray level mean value and the variance of the suspected foreign matter outline are both located in the gray level mean value threshold value and the variance threshold value, judging that the suspected foreign matter outline is an approximate foreign matter outline; if any element of the gray mean value and the variance does not meet the gray mean value threshold and the variance threshold, judging that the suspected foreign matter outline is a non-foreign matter outline; if the suspected foreign body contours in the input image are all non-foreign body contours, updating data and inputting the next image, and repeating the image processing step, the threshold I judging step and the step for detection;
an SVM classification step, namely performing dimensionality reduction on the sample image by using a principal component analysis method, and training by using data subjected to dimensionality reduction to obtain an SVM classification decision function; and (3) corresponding the approximate foreign matter outline judged in the threshold II judging step to the original image, setting an interested area according to the position of the approximate foreign matter outline frame in the original image, and substituting the interested area into an SVM classification decision function to perform secondary classification of foreign matters or non-foreign matters.
In the threshold value I judging step, the preset contour parameter threshold values comprise a contour area threshold value, a contour area-to-contour frame area ratio threshold value, a contour width threshold value, a contour height threshold value and a contour width-to-height ratio threshold value; the contour parameter threshold, the gray mean threshold and the variance threshold are obtained by manually framing a foreign object area in each image in a plurality of sample images, then obtaining the contour parameter, the gray mean and the variance of the foreign object contour through an image processing step and carrying out statistics by using statistics.
The preset contour parameter threshold, the preset gray level mean threshold and the preset variance threshold in the threshold I judging step and the preset contour parameter threshold in the threshold II judging step are obtained through the following processes:
manually framing a foreign matter area in each image in a plurality of sample images, then carrying out binarization processing, morphological processing, denoising processing and contour extraction processing to obtain a plurality of small foreign matter images, and then carrying out contour parameter calculation on the contours of the small foreign matter images to calculate a contour area AckThe ratio AR of the area of the outline to the area of the outline frameckWide profile RWckHigh profile RHckAnd profile aspect ratio RWck/RHckCalculating the pixel gray level mean value and the variance of the corresponding contour of the foreign matter small image contour in the original image through a gray level parameter calculation step, carrying out statistics by utilizing statistics to obtain a maximum statistic value MaxSto and a minimum statistic value MinSto of each parameter, and adding redundancy, namely ThrMaxSto ═ MaxSto (1+ Ratio1) and ThrMinMinMin ═ MinSto (1-Ratio2) into the maximum statistic value MaxSto and the minimum statistic value MinSto obtained by statistics, wherein Ratio1 ∈ (0,1) and Ratio2 ∈ (0,1) are added, so that the threshold values of the contour area, the Ratio of the contour area to the contour frame area, the contour width, the contour height, the contour width height Ratio, the gray level mean value and the gray level variance are obtained.
In the parameter threshold values in the threshold value I judging step and the threshold value II judging step, the Ratio1 and the Ratio2 are 0.05, 0.1 or 0.2.
The specific training process of the SVM classification decision function is as follows:
(a) manually framing a foreign matter area and a non-foreign matter area on an existing pantograph image sample, generating positive and negative samples to be trained, and re-adjusting the size of each sample to be M x N, wherein the set of samples to be trained is D ═ D0,D1,...,Di,...,DI-1]The sample label is y ═[y0,y1,...,yi,...,yI-1]Total of I samples, and yi∈ { -1, +1}, -1 denotes a non-foreign matter sample, +1 denotes a foreign matter sample;
(b) will DiReconstructing into head-connected M x N dimensional column vectors x by rowsiSo that the set of samples to be trained can be written as X ═ X1,x2,...,xI-1];
(c) The PCA dimensionality reduction matrix is obtained by utilizing the data set X, the central vector u of the data set is firstly calculated,then, each sample is subtracted by the central vector to obtain a de-centered sampleWhereinThen the covariance matrix of the center-removed sample is subjected to eigenvalue decomposition,wherein λ ═ λ0,λ1,...,λM*N-1]Is an eigenvalue of the covariance matrix, and0>λ1>...>λM*N-1,Q=[q1,q2,...,qM*N-1]is an eigenvector matrix of a covariance matrix, andicorresponds to qiThen setting a threshold value Tλ=λ0Rat, Rat ∈ (0,1), then selecting greater than TλForming a dimensionality reduction matrix by the eigenvectors corresponding to the eigenvalues;
(d) using a dimensionality reduction matrix to pair the depocenter sample sets according to the following formulaAnd (3) performing dimensionality reduction:whereinThe dimension of the data matrix after dimension reduction is K x I, K<<M*N;
s.t.1-yi(wTφ(xi)+b)-ξi≤0,ξimore than or equal to 0, I ═ 0,1,2
Parameters of noodles, ξiThe soft interval parameter is C, the penalty coefficient is C, phi (-) is a mapping relation, and phi (-) is operated through a kernel function; by constructing the lagrange function and the dual transformation, the above model is transformed as follows:
s.t.λi≥0,μi≥0;
by calculating the partial derivatives for each parameter and making them 0, we can obtain:
C=λi+μi,
and substituting the above relation into Lagrange function to obtain optimized model,
wherein phi (x)i)·φ(xj) Calculating by a kernel function, solving by using an SMO algorithm to obtain lambda, and selecting lambda satisfying that C is more than or equal to lambdajA component λ of 0 or morejAnd calculating:
in the image processing step, binarization processing is carried out on the pantograph slide bar area image, specifically: dividing an input image into a plurality of non-overlapping small blocks, and carrying out binarization on each small block according to the corresponding gray scale change characteristic of each small block; the specific process is as follows:
the input image I is divided into 4 non-overlapping small blocks, denoted as IA、IB、ICAnd IDThen E isA=f(IA,thrA)
i=0,1,2,...,IAH-1,j=i=0,1,2,...,IAW-1; wherein E isAIs IAImage after binarization, thrAFor threshold values used in binarization, the threshold value being in accordance with IAObtaining a gray level histogram of (1)AHIs IAHigh of (I)AWIs IAIs wide; in the same way, E can be obtainedB、EC、EDAnd then the binary images E of the input image I are obtained by splicing the images together in a manner of segmentation.
In the image processing step, morphological processing comprises image corrosion, image expansion, opening operation and closing operation; in the image processing step, the image obtained after the binarization processing is subjected to morphological processing, specifically, the image obtained after the binarization processing is subjected to two opening operations, and is subjected to one etching to obtain a processed image M.
In the image processing step, the mode of denoising the image after morphological processing is median filtering or Gaussian filtering.
In the SVM classification step, phi (-) is operated by a kernel function, and phi (x)i)·φ(xj) The kernel function in the operation by the kernel function includes one of a linear kernel, a polynomial kernel, and a gaussian kernel.
example 4
In order to achieve the above object, according to another aspect of the present application, there is also provided a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the above pantograph foreign object detection method when executing the computer program.
The processor may be a Central Processing Unit (CPU) in this embodiment. The Processor may also be other general purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, or a combination thereof.
The memory, which is a non-transitory computer readable storage medium, may be used to store non-transitory software programs, non-transitory computer executable programs, and units, such as the corresponding program units in the above-described method embodiments of the present invention. The processor executes various functional applications of the processor and the processing of the work data by executing the non-transitory software programs, instructions and modules stored in the memory, that is, the method in the above method embodiment is realized.
The memory may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created by the processor, and the like. Further, the memory may include high speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory optionally includes memory located remotely from the processor, and such remote memory may be coupled to the processor via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The one or more units are stored in the memory and, when executed by the processor, perform the methods of embodiments 1-3 above.
Example 5
As another preferred embodiment of the present invention, the present embodiment discloses a computer-readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the steps of embodiment 1 or embodiment 2 or embodiment 3 above:
the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A pantograph foreign matter detection method is characterized by comprising the following steps: the method comprises the following steps:
an image processing step: collecting and reading a gray image of a pantograph slide bar area, performing binarization processing on the image of the pantograph slide bar area by adopting a local self-adaptive binarization method, performing morphological processing on the image obtained after the binarization processing, performing denoising processing on the image after the morphological processing, searching a connected domain in the image after the denoising processing, and outputting the outer boundary of the connected domain as a contour; calculating outline parameters of the output outline, wherein the outline parameters comprise an outline area, an outline area-to-outline frame area ratio, an outline width, an outline height and an outline width-to-height ratio;
a threshold I judging step, namely comparing and judging the contour parameters calculated in the image processing step with a preset contour parameter threshold, if all the contour parameters of the output contour are within the contour parameter threshold, judging the contour as a suspected foreign body contour and outputting the contour to the next step, and if any contour parameter is not within the corresponding contour parameter threshold, judging the contour as a non-foreign body contour; if the contours in the input image are all non-foreign body contours, updating data and inputting the next image, and repeating the image processing step and the step for detection;
a threshold II judging step, namely corresponding the suspected foreign matter contour judged in the threshold I judging step to an original input image corresponding to the suspected foreign matter contour, and calculating a pixel gray level mean value and a variance in the corresponding contour in the original input image; respectively comparing the calculated gray level mean value and the calculated variance with a preset gray level mean value threshold value and a preset variance threshold value, and if the gray level mean value and the variance of the suspected foreign matter outline are both located in the gray level mean value threshold value and the variance threshold value, judging that the suspected foreign matter outline is an approximate foreign matter outline; if any element of the gray mean value and the variance does not meet the gray mean value threshold and the variance threshold, judging that the suspected foreign matter outline is a non-foreign matter outline; if the suspected foreign body contours in the input image are all non-foreign body contours, updating data and inputting the next image, and repeating the image processing step, the threshold I judging step and the step for detection;
an SVM classification step, namely performing dimensionality reduction on the sample image by using a principal component analysis method, and training by using data subjected to dimensionality reduction to obtain an SVM classification decision function; and (3) corresponding the approximate foreign matter outline judged in the threshold II judging step to the original image, setting an interested area according to the position of the approximate foreign matter outline frame in the original image, and substituting the interested area into an SVM classification decision function to perform secondary classification of foreign matters or non-foreign matters.
2. The pantograph foreign matter detection method according to claim 1, wherein: in the threshold value I judging step, the preset contour parameter threshold values comprise a contour area threshold value, a contour area-to-contour frame area ratio threshold value, a contour width threshold value, a contour height threshold value and a contour width height ratio threshold value; the contour parameter threshold, the gray mean threshold and the variance threshold are obtained by manually framing a foreign object area in each image in a plurality of sample images, then obtaining the contour parameter, the gray mean and the variance of the foreign object contour through an image processing step and carrying out statistics by using statistics.
3. The pantograph foreign matter detection method according to claim 2, wherein: the preset contour parameter threshold, the preset gray level mean threshold and the preset variance threshold in the threshold I judging step and the preset contour parameter threshold in the threshold II judging step are obtained through the following processes:
manually framing a foreign matter area in each image in a plurality of sample images, then carrying out binarization processing, morphological processing, denoising processing and contour extraction processing to obtain a plurality of small foreign matter images, and then carrying out contour parameter calculation on the contours of the small foreign matter images to calculate a contour area AckThe ratio AR of the area of the outline to the area of the outline frameckWide profile RWckHigh profile RHckAnd profile aspect ratio RWck/RHckCalculating the pixel gray level mean value and the variance of the corresponding contour of the foreign matter small image contour in the original image through a gray level parameter calculation step, carrying out statistics by utilizing statistics to obtain a maximum statistic value MaxSto and a minimum statistic value MinSto of each parameter, and adding redundancy, namely ThrMaxSto ═ MaxSto (1+ Ratio1) and ThrMinMinMin ═ MinSto (1-Ratio2) into the maximum statistic value MaxSto and the minimum statistic value MinSto obtained by statistics, wherein Ratio1 ∈ (0,1) and Ratio2 ∈ (0,1) are added, so that the threshold values of the contour area, the Ratio of the contour area to the contour frame area, the contour width, the contour height, the contour width height Ratio, the gray level mean value and the gray level variance are obtained.
4. As in claimThe pantograph foreign matter detection method according to claim 3, characterized by: in the parameter threshold values in the threshold value I judging step and the threshold value II judging step,wherein StofThe parameter values satisfying the maximum frequency are distributed statistically.
5. The pantograph foreign matter detection method according to claim 1, wherein: in the image processing step, binarization processing is carried out on the pantograph slide bar area image, specifically: dividing an input image into a plurality of non-overlapping small blocks, and carrying out binarization on each small block according to the corresponding gray scale change characteristic of each small block; the specific process is as follows:
the input image I is divided into 4 non-overlapping small blocks, denoted as IA、IB、ICAnd IDThen, then (ii) a Wherein E isAIs IAImage after binarization, thrAFor threshold values used in binarization, the threshold value being in accordance with IAObtaining a gray level histogram of (1)AHIs IAHigh of (I)AWIs IAIs wide; in the same way, E can be obtainedB、EC、EDAnd then the binary images E of the input image I are obtained by splicing the images together in a manner of segmentation.
6. The pantograph foreign matter detection method according to claim 1, wherein: in the image processing step, morphological processing is performed on the image obtained after binarization processing, specifically: and carrying out corrosion, expansion, opening operation or closing operation on the image after the binarization processing for many times.
7. The pantograph foreign matter detection method according to claim 5, wherein: in the image processing step, morphological processing is performed on the image obtained after binarization processing, specifically: and performing two opening operations on the image after the binarization processing, and performing one etching to obtain a processed image M.
8. The pantograph foreign matter detection method according to claim 1, wherein: in the image processing step, the mode of denoising the image after morphological processing is median filtering or Gaussian filtering.
9. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein: the processor, when executing the computer program, performs the steps of a pantograph foreign object detection method according to any one of claims 1 to 7.
10. A computer-readable storage medium characterized by: stored thereon, a computer program which, when being executed by a processor, carries out the steps of a pantograph foreign object detection method according to any one of the preceding claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010589402.5A CN111738342B (en) | 2020-06-24 | 2020-06-24 | Pantograph foreign matter detection method, storage medium and computer equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010589402.5A CN111738342B (en) | 2020-06-24 | 2020-06-24 | Pantograph foreign matter detection method, storage medium and computer equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111738342A true CN111738342A (en) | 2020-10-02 |
CN111738342B CN111738342B (en) | 2022-06-28 |
Family
ID=72651002
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010589402.5A Active CN111738342B (en) | 2020-06-24 | 2020-06-24 | Pantograph foreign matter detection method, storage medium and computer equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111738342B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113298059A (en) * | 2021-07-27 | 2021-08-24 | 昆山高新轨道交通智能装备有限公司 | Pantograph foreign matter intrusion detection method, device, computer equipment, system and storage medium |
CN113487561A (en) * | 2021-07-02 | 2021-10-08 | 成都唐源电气股份有限公司 | Pantograph foreign matter detection method and device based on gray gradient abnormal voting |
CN113763491A (en) * | 2021-08-26 | 2021-12-07 | 浙江中烟工业有限责任公司 | Visual detection method for tobacco shred barrel residues |
CN114140731A (en) * | 2021-12-08 | 2022-03-04 | 西南交通大学 | Traction substation abnormity detection method |
CN115100580A (en) * | 2022-08-23 | 2022-09-23 | 浙江大华技术股份有限公司 | Foreign matter detection method, device, terminal and computer readable storage medium |
CN115147770A (en) * | 2022-08-30 | 2022-10-04 | 山东千颐科技有限公司 | Belt foreign matter vision recognition system based on image processing |
CN116468729A (en) * | 2023-06-20 | 2023-07-21 | 南昌江铃华翔汽车零部件有限公司 | Automobile chassis foreign matter detection method, system and computer |
CN117315493A (en) * | 2023-11-29 | 2023-12-29 | 浙江天演维真网络科技股份有限公司 | Identification and resolution method, device, equipment and medium for field weeds |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012015698A (en) * | 2010-06-30 | 2012-01-19 | Meidensha Corp | Pantograph detection device and method thereof |
CN103984961A (en) * | 2014-05-30 | 2014-08-13 | 成都西物信安智能系统有限公司 | Image detection method for detecting foreign matter at bottom of vehicle |
JP2016192863A (en) * | 2015-03-31 | 2016-11-10 | 公益財団法人鉄道総合技術研究所 | Device and method for detecting abnormality of pantograph and method for coping with detected abnormality of pantograph |
CN110378897A (en) * | 2019-07-25 | 2019-10-25 | 中车青岛四方机车车辆股份有限公司 | A kind of pantograph running state real-time monitoring method and device based on video |
CN110458126A (en) * | 2019-08-16 | 2019-11-15 | 上海仁童电子科技有限公司 | A kind of pantograph state monitoring method and device |
CN111260629A (en) * | 2020-01-16 | 2020-06-09 | 成都地铁运营有限公司 | Pantograph structure abnormity detection algorithm based on image processing |
-
2020
- 2020-06-24 CN CN202010589402.5A patent/CN111738342B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012015698A (en) * | 2010-06-30 | 2012-01-19 | Meidensha Corp | Pantograph detection device and method thereof |
CN103984961A (en) * | 2014-05-30 | 2014-08-13 | 成都西物信安智能系统有限公司 | Image detection method for detecting foreign matter at bottom of vehicle |
JP2016192863A (en) * | 2015-03-31 | 2016-11-10 | 公益財団法人鉄道総合技術研究所 | Device and method for detecting abnormality of pantograph and method for coping with detected abnormality of pantograph |
CN110378897A (en) * | 2019-07-25 | 2019-10-25 | 中车青岛四方机车车辆股份有限公司 | A kind of pantograph running state real-time monitoring method and device based on video |
CN110458126A (en) * | 2019-08-16 | 2019-11-15 | 上海仁童电子科技有限公司 | A kind of pantograph state monitoring method and device |
CN111260629A (en) * | 2020-01-16 | 2020-06-09 | 成都地铁运营有限公司 | Pantograph structure abnormity detection algorithm based on image processing |
Non-Patent Citations (4)
Title |
---|
DONGKAI ZHANG等: "A Robust Pantograph–Catenary Interaction", 《IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》 * |
YUAN SHEN等: "A Pantograph Horn Detection Method Based on Deep Learning Network", 《2018 THE 3RD OPTOELECTRONICS GLOBAL CONFERENCE》 * |
刘建宏: "高速列车接触网悬挂系统缺陷图像识别技术研究", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 * |
曹蔡旻: "机车受电弓及其车顶状态检测系统", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113487561B (en) * | 2021-07-02 | 2023-06-30 | 成都唐源电气股份有限公司 | Pantograph foreign matter detection method and device based on gray gradient abnormal voting |
CN113487561A (en) * | 2021-07-02 | 2021-10-08 | 成都唐源电气股份有限公司 | Pantograph foreign matter detection method and device based on gray gradient abnormal voting |
CN113298059A (en) * | 2021-07-27 | 2021-08-24 | 昆山高新轨道交通智能装备有限公司 | Pantograph foreign matter intrusion detection method, device, computer equipment, system and storage medium |
CN113763491A (en) * | 2021-08-26 | 2021-12-07 | 浙江中烟工业有限责任公司 | Visual detection method for tobacco shred barrel residues |
CN113763491B (en) * | 2021-08-26 | 2024-03-12 | 浙江中烟工业有限责任公司 | Visual detection method for tobacco shred barrel residues |
CN114140731A (en) * | 2021-12-08 | 2022-03-04 | 西南交通大学 | Traction substation abnormity detection method |
CN115100580A (en) * | 2022-08-23 | 2022-09-23 | 浙江大华技术股份有限公司 | Foreign matter detection method, device, terminal and computer readable storage medium |
CN115147770B (en) * | 2022-08-30 | 2022-12-02 | 山东千颐科技有限公司 | Belt foreign matter vision recognition system based on image processing |
CN115147770A (en) * | 2022-08-30 | 2022-10-04 | 山东千颐科技有限公司 | Belt foreign matter vision recognition system based on image processing |
CN116468729A (en) * | 2023-06-20 | 2023-07-21 | 南昌江铃华翔汽车零部件有限公司 | Automobile chassis foreign matter detection method, system and computer |
CN116468729B (en) * | 2023-06-20 | 2023-09-12 | 南昌江铃华翔汽车零部件有限公司 | Automobile chassis foreign matter detection method, system and computer |
CN117315493A (en) * | 2023-11-29 | 2023-12-29 | 浙江天演维真网络科技股份有限公司 | Identification and resolution method, device, equipment and medium for field weeds |
CN117315493B (en) * | 2023-11-29 | 2024-02-20 | 浙江天演维真网络科技股份有限公司 | Identification and resolution method, device, equipment and medium for field weeds |
Also Published As
Publication number | Publication date |
---|---|
CN111738342B (en) | 2022-06-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111738342B (en) | Pantograph foreign matter detection method, storage medium and computer equipment | |
Bautista et al. | Convolutional neural network for vehicle detection in low resolution traffic videos | |
CN111080620B (en) | Road disease detection method based on deep learning | |
Sen-Ching et al. | Robust techniques for background subtraction in urban traffic video | |
CN110232380A (en) | Fire night scenes restored method based on Mask R-CNN neural network | |
CN103839085B (en) | A kind of detection method of compartment exception crowd density | |
CN109255326B (en) | Traffic scene smoke intelligent detection method based on multi-dimensional information feature fusion | |
CN110826389B (en) | Gait recognition method based on attention 3D frequency convolution neural network | |
Khalifa et al. | Malaysian Vehicle License Plate Recognition. | |
CN101482923A (en) | Human body target detection and sexuality recognition method in video monitoring | |
CN108280409B (en) | Large-space video smoke detection method based on multi-feature fusion | |
Bedruz et al. | Real-time vehicle detection and tracking using a mean-shift based blob analysis and tracking approach | |
CN110929593A (en) | Real-time significance pedestrian detection method based on detail distinguishing and distinguishing | |
CN111753794B (en) | Fruit quality classification method, device, electronic equipment and readable storage medium | |
CN113344475B (en) | Transformer bushing defect identification method and system based on sequence modal decomposition | |
Shoaib et al. | Shadow detection for moving humans using gradient-based background subtraction | |
CN111062273A (en) | Tracing detection and alarm method for left-over articles | |
Lian et al. | A novel method on moving-objects detection based on background subtraction and three frames differencing | |
CN114298948A (en) | Ball machine monitoring abnormity detection method based on PSPNet-RCNN | |
Daniel et al. | Automatic road distress detection and analysis | |
Nejati et al. | License plate recognition based on edge histogram analysis and classifier ensemble | |
CN111368742B (en) | Reconstruction and identification method and system of double yellow traffic marking lines based on video analysis | |
CN115171218A (en) | Material sample feeding abnormal behavior recognition system based on image recognition technology | |
CN112084957B (en) | Mobile target retention detection method and system | |
Kamal et al. | Human detection based on HOG-LBP for monitoring sterile zone |
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 |