CN105321173B

CN105321173B - A kind of train crossing cable cleat automatic defect detection method based on machine vision

Info

Publication number: CN105321173B
Application number: CN201510611588.9A
Authority: CN
Inventors: 张静; 曾振; 杜晓辉; 倪光明; 刘娟秀; 刘霖; 刘永; 叶玉堂
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2015-09-23
Filing date: 2015-09-23
Publication date: 2017-11-17
Anticipated expiration: 2035-09-23
Also published as: CN105321173A

Abstract

A kind of train crossing cable cleat automatic defect detection method based on machine vision of the disclosure of the invention, is related to the processing of cable cleat image in image processing field, particularly train crossing.Fixture image is gathered first；Vertical projection is carried out to image, and judges whether there is pipeline in picture according to drop shadow curve, pipeline location is determined if having pipeline, and intercept pipeline image；Statistics with histogram is carried out, is carried out negating binaryzation according to Low threshold, then judges whether there is fixture on the pipeline in piping drawing picture, if there is fixture to determine chucking position；Statistics with histogram is carried out to interception image, is carried out negating binaryzation according to high threshold, judges whether the fixture on the pipeline in piping drawing picture comes off.With it is easy, quick, efficiently find the fixture that comes off, to change in time.

Description

A kind of train crossing cable cleat automatic defect detection method based on machine vision

Technical field

The present invention relates to the processing of cable cleat image in image processing field, particularly train crossing.

Background technology

Cable cleat is the elimination vortex phenomenon in order to prevent from producing movement because external force suffix is conducted oneself with dignity after cable laying, and Prevent cable due to electrodynamic generation and bounce and the offset phenomena of generation, thus it is solid to need to use fixture to be segmented cable It is fixed.Because train crossing cable cleat arrangement environment is complicated, the card long-time aging on cable cleat comes off, it is necessary to coming off Fixture carry out replacing processing.Train mileage in China's is long, it is necessary to which the tunnel road of detection is more and complicated at present, and China uses at present Or artificial visual the fixture that comes off of method detection, not only Detection accuracy is low, speed is slow, cost is high, and in railway On detected and dangerous by people, such as how most lower cost, accuracy rate detection quick, higher faster have come off fixture The fixture of upper clamp slice is into a major issue in maintaining cables.

The content of the invention

In order to overcome the test problems that inner cable fixture upper clamp slice in tunnel comes off on rail link, the invention provides based on The simplicity of machine vision, fast and efficiently detection method, so as to find the fixture to come off rapidly, to change in time.

The technical scheme is that analyzed by the shape special efficacy of the dash area to fixture so as to confirm fixture It whether there is and judge whether fixture is intact, therefore a kind of train crossing cable cleat automatic defect detection based on machine vision Method, comprise the following steps that：

Step 1：High speed acquisition video camera is loaded ON TRAINS, tunnel cable clamp is gathered during railway operation Greyscale video, and position and the velocity information of collection point are recorded simultaneously；

Step 2：Video is resolved into single image, between-line spacing selection is entered to image, then located as follows to choosing image Reason；

Step 3：Vertical projection is carried out to image, and judges whether there is pipeline in image according to drop shadow curve, if there is pipeline Pipeline location is then determined, and intercepts pipeline image；

Step 4：Statistics with histogram is carried out to the image that step 3 obtains, is carried out negating binaryzation according to Low threshold, then judge Whether there is fixture on pipeline in piping drawing picture, if there is fixture to determine chucking position；

Step 5：Statistics with histogram is carried out to the image that step 3 obtains, is carried out negating binaryzation according to high threshold, judges to manage Whether the fixture on pipeline in road image comes off.

Further, the image obtained in the step 4 to step 3 carries out negating binaryzation according to Low threshold, due to folder Tool and pipeline threshold value are variant, can leave a small amount of connected region after Low threshold processing, judge piping drawing picture according to connected region In pipeline on whether have fixture.

Further, the image obtained in the step 5 to step 3 carries out negating binaryzation according to high threshold, by two-value Change the maximum transversal distance of connected region in image compared with the width of pipeline, if being shorter than duct width, judge fixture Intermediate plate come off.

The determination method of Low threshold and high threshold is in the step 4,5：

Step 3 is obtained into image and carries out statistics with histogram, histogram is divided into by left and right or upper and lower two according to fixture direction Part, two-part crest value being recorded respectively, relatively low crest value being set as Low threshold, higher crest value is set as height Threshold value.

The beneficial effect of the system is, it is only necessary to image capturing system can is installed on normal train to train usually Detected along wire holder, it is simple efficient.By the image procossing to collection, this method can rapidly detect Along Railway tunnel Cable clamp surface fixture card is opened and come off in road.

Brief description of the drawings

Fig. 1 is system flow chart

Fig. 2 is the OK sample images that embodiment 1 gathers

Fig. 3 is the drop shadow curve of embodiment 1

Fig. 4 is region of interesting extraction image

Fig. 5 is Low threshold segmentation figure

Fig. 6 is high threshold segmentation figure

Fig. 7 is the fixture image navigated to

Fig. 8 is the NG sample images that embodiment 2 gathers

Fig. 9 is the region of interest area image extracted

Figure 10 is Low threshold segmentation figure picture

Figure 11 is high threshold segmentation figure picture

Figure 12 is the fixture image that embodiment 2 navigates to

1. pipeline in figure, 2. fixtures, 3. fixture cards

Embodiment

Fig. 1 is system algorithm implementing procedure figure, including the algorithm stream of video image acquisition and ensuing image procossing Journey.

The fixture example images of Fig. 2 preferably, wherein have pipeline 1, fixture 2, fixture card 3.

Video image (being gathered by previous step 1) is gathered by CCD camera first, then the video image collected is divided Then solution carries out X-axis projection, drop shadow curve is as shown in Figure 3 into single-frame images, and by frame input algorithm to image.(due to this reality When camera is placed in example, in order that the image of cable in the picture is as long as possible, so camera is laterally disposed, i.e., on image Vertical direction is horizontal direction in practice, and the horizontal direction in image is vertical direction in practice).

Next pipeline location is determined according to image projection curve, as shown in Fig. 3 figures, huge depression is electricity among curve Cable pipeline position；Then region of interesting extraction is carried out to image, i.e., cable is extracted according to cable channel position Image, the image extracted are as shown in Figure 4.

And then the segmentation of Low threshold image is carried out to the image extracted and high threshold image is split to obtain two segmentation figures As (this two segmentation figure pictures have all carried out inversion operation, i.e., when threshold value is x, the pixel less than x is white, and what it is higher than x is Black), the step is the core of this algorithm, determines whether there is fixture on pipeline using the segmentation figure picture of Low threshold, If fixture can then determine the position of fixture by this segmentation image；Using high threshold image judge fixture whether UNICOM. This example is that Low threshold will using the fixed threshold (this threshold value can be changed) set manually, the key of threshold value setting The pedestal of fixture can be embodied, high threshold wants the shade of energy display pipes upper clip, is found by surveying when polishing condition is identical, It is stable and reliable using fixed threshold.

Next area filtering is carried out to the image of Low threshold segmentation, wherein Low threshold image is split and filters out small area For image as shown in figure 5, equally being operated to high threshold segmentation figure picture, high threshold segmentation is as shown in Figure 6.

And then the connected domain coordinate setting fixture coordinate obtained according to Low threshold image segmentation (Fig. 5), and in high threshold Interception holder part segmentation figure picture in segmentation figure picture (Fig. 6), as shown in Figure 7.Finally in the figure 7 according to connected domain widthwise size come Judge whether the intermediate plate on fixture comes off.(set if there is the lateral length of connected domain more than given duct length by oneself It is fixture (in embodiment as shown in Figure 8) that is intact, otherwise being come off for fixture card then to judge its fixture calmly), and exports knot Fruit, that is, the picture position for the fixture card that comes off.

When the image to be come off for fixture card of input, example image is as shown in figure 8, Fig. 8 is the NG figures detected Picture.Wherein there was only pipeline 1, fixture 2.Wherein Figure 10 is that Low threshold image is split and filters out the image after small area, and Figure 11 is height Threshold Image Segmentation.Then the UNICOM domain coordinate setting chucking position obtained after being split according to Low threshold, and in high threshold image Middle to intercept the image navigated to, image is as shown in figure 12 after interception.

The mileage information and the result of fixture detection output recorded during according to collection image, being calculated needs to change fixture Fixture particular location.

Claims

1. a kind of train crossing cable cleat automatic defect detection method based on machine vision, is comprised the following steps that：

Step 1：High speed acquisition video camera is loaded to the gray scale for gathering tunnel cable fixture during railway operation ON TRAINS Video, and position and the velocity information of collection point are recorded simultaneously；

Step 2：Video is resolved into single image, between-line spacing selection is entered to image, then be handled as follows to choosing image；

Step 3：Vertical projection is carried out to image, and judges whether there is pipeline in image according to drop shadow curve, it is true if having pipeline Determine pipeline location, and intercept pipeline image；

Step 4：Statistics with histogram is carried out to the image that step 3 obtains, is carried out negating binaryzation according to Low threshold, then judge pipeline Whether there is fixture on pipeline in image, if there is fixture to determine chucking position；

Step 5：Statistics with histogram is carried out to the image that step 3 obtains, is carried out negating binaryzation according to high threshold, judges piping drawing Whether the fixture on pipeline as in comes off；Determination methods are：By the maximum transversal of connected region in obtained binary image Distance is compared with the width of pipeline, if being shorter than duct width, judges that the intermediate plate of fixture comes off；

Step 3 is obtained into image and carries out statistics with histogram, histogram is divided into by left and right or upper and lower two parts according to fixture direction, Two-part crest value is recorded respectively, relatively low crest value is set as Low threshold, and higher crest value is set as high threshold.

2. a kind of train crossing cable cleat automatic defect detection method based on machine vision as claimed in claim 1, its The image for being characterised by obtaining step 3 in the step 4 carries out negating binaryzation according to Low threshold, due to fixture and pipeline threshold Be worth it is variant, Low threshold processing after can leave a small amount of connected region, judged according to connected region on the pipeline in piping drawing picture Whether fixture is had.