A method for processing images of a water turbine runner

Abstract

The application relates to the technical field of water turbine cavitation recognition, and discloses a water turbine runner image processing method, which comprises the following steps: acquiring a water turbine runner image in real time, and preprocessing the image as a target image to obtain a typical target image binary line graph, wherein the target image binary line graph comprises a plurality of single-line-shaped target curve segments; the direction vectors of each pixel point and the right-side adjacent pixel point on each single-line-shaped target curve segment in the target image binary line graph are extracted respectively, and the direction vectors of each pixel point and the right-side adjacent pixel point on the same target curve segment form a vector chain of the target curve segment. The water turbine runner image collected is processed based on the image processing method, and the water turbine runner image features obtained after processing can improve the cavitation recognition efficiency and the recognition accuracy when the water turbine runner image features are used for cavitation recognition in the later period.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water turbine runner cavitation recognition, and particularly relates to a water turbine runner image processing method. BACKGROUND

[0002] The incipient cavitation of a water turbine refers to a phenomenon that a gas nucleus in a liquid sharply grows when a local pressure in the liquid is reduced to a critical value. The incipient cavitation is usually marked by an incipient cavitation coefficient. The definition of the incipient cavitation coefficient of the water turbine is that the incipient cavitation coefficient is the cavitation coefficient when bubbles appear on a blade. The main reasons for the bubbles on the blade are usually three. The first reason is so-called airfoil cavitation, which is mainly generated on a back surface of a blade at a water outlet side and is greatly affected by a change in a draft tube pressure (i.e., the cavitation coefficient). The second reason is blade inlet cavitation, which is mainly generated on a back surface of a blade at a water inlet side and is mainly caused by flow separation. The third reason is axial flow and cross flow turbine blade end gap flow cavitation bubbles.

[0003] It is difficult to observe the incipient cavitation phenomenon of a water turbine runner at a water outlet side. At present, a method for detecting the incipient cavitation is mainly a manual visual inspection method. The visual inspection method is to manually observe a vortex band of a draft tube and a flow state of the water turbine runner at the water outlet side through a straight conical section of the draft tube made of transparent organic glass. In the test, a strobe light (a light source with adjustable frequency of light and dark alternation) is used to light. The frequency of the strobe light is adjusted to be equal to or close to the frequency of the model water turbine. The runner blade at the water outlet side seems to be static or slowly rotating, so that the cavitation state of the runner blade at the water outlet side can be observed. A camera is used to take a picture for recording. However, the method requires a high requirement for the staff. The staff should have relevant work experience. Generally, at least six years of work experience is required to determine whether the incipient cavitation exists in the picture. Therefore, the method not only has strong observer subjectivity, but also has low cavitation discrimination efficiency. SUMMARY

[0004] The present application provides a water turbine runner image processing method. Based on the image processing method, the real-time collected water turbine runner image is processed. The image features obtained after the processing can improve the cavitation discrimination efficiency and the discrimination accuracy in the later cavitation discrimination.

[0005] In order to achieve the above-mentioned application purposes, the technical scheme of the present application is as follows:

[0006] A water turbine runner image processing method, comprising the following steps:

[0007] The real-time obtained water turbine runner image is taken as a target image. The target image is preprocessed to obtain a typical target image binary line graph. The target image binary line graph includes a plurality of single-line-shaped curve segments. The single-line-shaped curve segments in the target image binary line graph are taken as target curve segments.

[0008] respectively extract the direction vector of each pixel point and the right adjacent pixel point on each single line shape target curve segment in the target image binary line graph, and the direction vector of each pixel point and the right adjacent pixel point on the same target curve segment constitutes the vector chain of the target curve segment.

[0009] Further, the pre-processing of the target image to obtain a typical target image binary line graph comprises:

[0010] After the real-time collected water turbine runner image is processed through cutting, masking, greying and edge enhancement, the image is binarized, and the Y direction repeated pixel points in the image are removed, so that each pixel point has and only has one right pixel point connected thereto, and finally a target image binary line graph comprising a plurality of single line shape curve segments is formed.

[0011] Further, the respective extraction of the direction vector of each pixel point and the right adjacent pixel point on each single line shape target curve segment in the target image binary line graph comprises:

[0012] The leftmost pixel point on the single line shape target curve segment is taken as the starting point, and all pixel points on the curve segment are traversed from the starting point to obtain the direction vector of each pixel point and the right adjacent pixel point.

[0013] Further, the pre-processing of the target image to obtain a typical target image binary line graph further comprises:

[0014] The obtained target image binary line graph is cut in the ROI region, and the ROI region is transformed in angle and scale with reference to the template image.

[0015] Further, the template image refers to an image with typical characteristics selected from a sample image of a water turbine runner in which cavitation has occurred, and a local cavitation bubble pattern of the image is cut as a template image.

[0016] Further, after the template image is obtained, the same binarization processing method as the target image is adopted to finally obtain a typical template image binary line graph.

[0017] Further, in the template image binary line graph, a single line shape curve segment is arbitrarily selected as a reference curve segment, the leftmost pixel point on the reference curve segment is taken as a reference key point, and the parameters of the reference key point are set as (V, M), wherein V is the direction of the vector from the key point to the rightmost pixel point of the reference curve segment, and M is the module.

[0018] Further, the transformation of the ROI region in angle and scale with reference to the template image in the database comprises:

[0019] According to the key point parameter V of the template image, first, the ROI region is rotated to an angle V to obtain a rotated image, and then the rotated image is scaled to M according to the key point parameter M of the template image, and finally the angle and scale transformation of the ROI region is completed.

[0020] The beneficial effects of the present application are:

[0021] The water turbine runner image processing method of the present application is simple and easy to implement, has low requirements on computing resources, can quickly process the real-time collected water turbine runner image, and ultimately can obtain a plurality of vector chains representing image features. The features obtained based on the image processing method can be used to determine whether the water turbine runner has cavitation, and the features can be effectively applied to single-line shape matching pairs and are robust to shape translation, rotation and scaling transformation. The water turbine runner image features obtained after processing can improve the cavitation discrimination efficiency and discrimination accuracy when used for cavitation discrimination in the later stage. BRIEF DESCRIPTION OF DRAWINGS

[0022] The foregoing and subsequent specific description of the present application becomes clearer when read in conjunction with the following drawings, in which:

[0023] Figure 1 The method flowchart of the present application. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the technical solutions in the present application, the following will further illustrate the technical solutions for achieving the purposes of the present application through several specific embodiments. It should be noted that the technical solutions claimed in the present application include but are not limited to the following embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should belong to the scope of protection of the present application.

[0025] The incipient cavitation of the water turbine refers to the phenomenon that the local pressure in the liquid decreases to the critical value, and the contained gas nucleus grows rapidly when cavitation starts. The incipient cavitation is often marked by the incipient cavitation coefficient. The definition of the incipient cavitation coefficient of the water turbine is: the cavitation coefficient when the runner blade starts to appear bubbles.

[0026] The initial cavitation phenomenon of the runner of a hydraulic turbine is difficult to observe. At present, the method for detecting the initial cavitation is artificial visual observation. The visual observation is performed through a straight cone section of a tail water pipe made of transparent organic glass to observe the vortex band of the tail water pipe and the flow pattern of the runner. In the test, a stroboscope (a light source with adjustable frequency of light and dark alternation) is used to provide light. The frequency of the stroboscope is adjusted to be equal to or close to the frequency of the model hydraulic turbine. The runner blades seem to be static or slowly rotating, so that the cavitation condition of the runner blades can be observed. A camera is used to take pictures for recording. However, the method requires a high requirement for the staff. The staff should have relevant work experience. Generally, the staff should have more than six years of work experience to determine whether the initial cavitation exists in the pictures. The method has strong observer subjectivity and low efficiency.

[0027] Based on this, the embodiment provides a hydraulic turbine runner image processing method. The method includes two stages. The first stage is a template image processing stage. The purpose is to establish a basic database. The subsequent template image is used as a reference to process the real-time collected hydraulic turbine runner image. The second stage is a target image processing stage. The process is to process the real-time collected hydraulic turbine runner image. The related image features in the target image are extracted according to a specific method. The extracted image features can be used for cavitation recognition of the hydraulic turbine runner. The efficiency of cavitation recognition can be improved. Since the features are robust to shape translation, rotation and scaling transformation, the accuracy of cavitation recognition can also be improved.

[0028] The embodiment discloses a hydraulic turbine runner image processing method. Referring to the description Figure 1 , the method is as follows:

[0029] Step S1. Template image processing

[0030] Step S101. In the sample image of the hydraulic turbine runner in which cavitation has occurred, an image with typical features is manually selected. The local cavitation bubble pattern of the image is intercepted as a template image.

[0031] In the embodiment, it should be noted that the image with typical features refers to a clear and unblurred runner image, and the cavitation state of the runner is also clear. After the required image is selected, the local image of the image is intercepted as a template image. The template image usually contains a complete cavitation bubble.

[0032] Step S102. The template image is binarized. The Y-direction repeated pixel points in the image are removed by calculation to ensure that each pixel point has only one right pixel point connected to it. Finally, a typical binary line graph is formed.

[0033] In the embodiment, it is to be noted that the template image is subjected to the above-mentioned binaryzation processing, and finally a single-line binaryzation image is formed, and the contour shape of the line in the image is the contour of the cavitation bubble.

[0034] In the embodiment, it is to be further noted that the contour of the cavitation bubble in the single-line binaryzation image formed after the template image is subjected to the binaryzation processing can be composed of a plurality of disconnected single-line-shaped curve segments, or can be composed of a continuous single-line-shaped curve segment; when the contour of the cavitation bubble is composed of a continuous single-line-shaped curve, the continuous curve needs to be segmented into a plurality of curve segments, and the ultimate purpose is to make the contour of the cavitation bubble in the template image composed of a plurality of single-line-shaped curve segments.

[0035] In the embodiment, it is to be further noted that the typical binary line image obtained by the binaryzation processing is a relatively conventional method in the field of image processing, and will not be described herein.

[0036] Step S103. In the template image binary line image, an arbitrary single-line-shaped curve segment is selected as a reference curve segment, and the leftmost pixel point on the reference curve segment is selected as a reference key point, and the parameters of the reference key point are set as (V, M), wherein V is the direction of the vector from the reference key point to the rightmost pixel point of the reference curve segment, and M is the module.

[0037] Step S104. Based on the processed template image binary line image, the directional vectors of each pixel point and the right adjacent pixel point on each single-line-shaped curve segment in the template image binary line image are extracted; specifically, the leftmost pixel point on the single-line-shaped curve segment is taken as a starting point, and all the pixel points on the curve segment are traversed from the starting point to obtain the directional vectors of each pixel point and the right adjacent pixel point, and the directional vectors of each pixel point and the right adjacent pixel point on the same curve segment constitute a vector chain of the curve segment. The template image contains a plurality of curve segments, and therefore a plurality of vector chains of the template image are finally obtained, and the plurality of vector chains can be further combined into a template matrix.

[0038] In the embodiment, the reference key point, the reference key point parameters, the vector chain and other related data of the template image and the image processing method are stored in the database, and finally serve as a template for the target image to be processed for reference.

[0039] Step S2. Target image processing

[0040] Step S201. The real-time obtained water turbine runner image is taken as a target image, the target image is preprocessed, a typical target image binary line image is obtained, and the target image binary line image includes a plurality of single-line-shaped target curve segments.

[0041] In the embodiment, it is to be noted that the pre-processing means of the target image refers to the template image, and finally the binary line graph of the target image can be obtained through the same processing manner.

[0042] In the embodiment, it is to be noted that before the binary processing of the target image, the target image also needs to be cut, masked, grayed, and edge enhanced, and other additional pre-processing means. These processing means are relatively common in the image processing field, and will not be described here.

[0043] Step S202. Based on the processed target image binary line graph, the target curve segment in the image is traversed. When the number of white points contained in the target curve segment is greater than the set threshold value, the curve segment is retained, and the curve segment is taken as a to-be-identified curve segment. When the number of white points contained in the target curve segment is less than the set threshold value, the curve segment is discarded.

[0044] In the embodiment, it is to be noted that after the image binary processing, the image is composed of a plurality of black points and white points. The curve segment of a single line shape is formed by a plurality of continuous white points in the image. The white points and the black points are both pixel points of the image.

[0045] In the embodiment, it is to be noted that the threshold value is a manually set parameter, which can be set by oneself according to the demand.

[0046] In the embodiment, by traversing a plurality of target curve segments in the target image, the target curve segment meeting the requirement is selected as a to-be-identified curve segment, which can greatly reduce the calculation amount and improve the image processing efficiency.

[0047] Step S203. Based on the to-be-identified curve segment, the ROI region is intercepted from the target image binary line graph, and the intercepted ROI region contains a to-be-identified curve segment. Taking the intercepted ROI region as the research object, first, the key point parameters (V1, M1) of the ROI region image are obtained by referring to the processing manner of the template image, and then the ROI region is transformed in angle and scale by referring to the parameters of the reference key points of the template image in the database. Specifically, first, the ROI region is rotated from V1 to angle V according to the parameter V of the reference key points of the template image, to obtain a rotated image, and then the rotated image is scaled from M1 to M according to the parameter M of the reference key points of the template image, to finally complete the angle and scale transformation of the ROI region.

[0048] In the embodiment, it is to be noted that the ROI region is intercepted, which is a relatively common method in the image processing field, and will not be described here.

[0049] Step S204. Based on the ROI region with the angle and scale transformation completed, taking the leftmost pixel point of the curve segment to be recognized in the region as a target key point, traversing all the pixel points on the curve segment to be recognized from the target key point, obtaining the direction vector of each pixel point and the adjacent pixel point on the right, and all the obtained direction vectors constitute the vector chain of the curve segment to be recognized.

[0050] In this embodiment, it should be noted that the vector chain of the ROI region is a real-time obtained feature vector, which is related to the real-time collected turbine runner image.

[0051] In this embodiment, it should be noted that there are multiple ROI regions intercepted by the curve segment to be recognized, and the vector chain of the curve segment to be recognized in each ROI region is obtained by referring to the above steps.

[0052] In this embodiment, the vector chain of the ROI region of the target image is obtained through the above series of image processing means, which can be used as a cavitation recognition feature for cavitation recognition of the turbine runner, and thus can help the technical personnel to quickly judge whether cavitation has occurred in the region of the turbine runner, thereby improving the efficiency and accuracy of cavitation discrimination.

[0053] In this embodiment, there are multiple ways to judge whether the turbine runner has cavitation through the vector chain of the ROI region of the target image, and one of them can be selected when judging cavitation. Some cavitation discrimination methods of the turbine runner will be listed below.

[0054] Method one, Euclidean distance: the straight line distance between two points, the calculation formula is as follows:

[0055]

[0056] When judging the cavitation of the turbine runner, the parameters in the vector chain of the ROI region and the parameters in the vector chain of the template image are substituted into the above formula, and the calculation result is used to measure the similarity between the target image and the template image.

[0057] Method two, Manhattan distance: the sum of the distances of two points in different dimension directions, the calculation formula is as follows:

[0058]

[0059] When judging the cavitation of the turbine runner, the parameters in the vector chain of the ROI region and the parameters in the vector chain of the template image are substituted into the above formula, and the calculation result is used to measure the similarity between the target image and the template image.

[0060] Method three, cosine distance: the concept of cosine of angle is used to measure the similarity between vectors, the calculation formula is as follows:

[0061]

[0062] In the judgment of the water turbine runner cavitation, the parameters in the vector chain of the ROI region and the parameters in the vector chain of the template image are substituted into the above formula, and the calculation result is used to measure the similarity of the target image and the template image. The smaller the cosine distance is, the more similar the ROI region and the template image are.

[0063] It should be noted that if 2-3 ROI regions in a target image can be successfully matched with the template image, it is considered that the water turbine runner has cavitation.

[0064] After obtaining the vector chain of the ROI region, the above-mentioned methods can be used to realize the cavitation discrimination of the water turbine runner. Among them, the Euclidean distance and the cosine distance are applied more, but the application scenarios of the two are slightly different. In the field, the cosine distance can be used to judge whether the water turbine runner has cavitation. By setting an angle threshold, when the calculation result is less than the set angle threshold, it is considered that the two are matched.

[0065] Further, in addition to the above-mentioned cavitation determination methods, a target matrix can be formed by combining the vector chains of multiple ROI regions, and a template matrix can be formed by combining the vector chains of the template image. The two matrices are compared. If the two matrices are equal, it means that cavitation has occurred in the water turbine runner region composed of multiple ROI regions.

[0066] When determining whether the two matrices are equal, the two matrices can be modified by artificial experience parameters, and finally the modified two matrices are used for discrimination.

[0067] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Any simple modification, equivalent change, etc. based on the technical essence of the present application to the above embodiment falls within the protection scope of the present application.

Claims

1. A method for image processing of a water turbine runner, characterized in that, include: The real-time image of the turbine runner is used as the target image. The target image is preprocessed to obtain a typical target image binary line map. The target image binary line map includes multiple single-line shaped curve segments. The single-line shaped curve segments in the target image binary line map are used as target curve segments. Extract the direction vectors of each pixel and its right-side adjacent pixel on each single-line shaped target curve segment in the binary line diagram of the target image. The direction vectors of each pixel and its right-side adjacent pixel on the same target curve segment constitute the vector chain of that target curve segment. The preprocessing of the target image to obtain a typical target image binary line map includes: The real-time acquired images of the turbine runner are processed by cropping, masking, graying, and edge enhancement. Then, the images are binarized. By removing duplicate pixels in the Y direction, each pixel is connected to only one pixel to its right, ultimately forming a target image binary line map that includes multiple single-line shaped curve segments. The step of extracting the direction vectors of each pixel point and its right-side adjacent pixel point on each single-line shaped target curve segment in the binary line image of the target image includes: Starting from the leftmost pixel on the single-line shaped target curve segment, traverse all pixels on the curve segment from the starting point and obtain the direction vector between each pixel and its right-side adjacent pixel.

2. The image processing method for a water turbine runner according to claim 1, characterized in that, The preprocessing of the target image to obtain a typical target image binary line graph also includes: The ROI region is cropped from the obtained binary line graph of the target image, and the angle and scale of the ROI region are transformed with reference to the template image.

3. The image processing method for a water turbine runner according to claim 2, characterized in that, The template image refers to the image with typical characteristics selected from the sample images of the turbine runner that has already undergone cavitation, and the local cavitation bubble pattern of the image is extracted as the template image.

4. The image processing method for a water turbine runner according to claim 3, characterized in that, After obtaining the template image, the same binarization process as the target image is used to finally obtain a typical template image binary line graph.

5. The image processing method for a water turbine runner according to claim 4, characterized in that, In the binary line graph of the template image, arbitrarily select a single-line shaped curve segment as the reference curve segment, and the leftmost pixel on the reference curve segment as the reference key point. Set the parameters of the reference key point as (V, M), where V is the direction of the vector from the key point to the rightmost pixel of the reference curve segment, and M is the magnitude.

6. The image processing method for a water turbine runner according to claim 5, characterized in that, The ROI region is transformed in terms of angle and scale using template images from the database, including: Based on the key point parameters V of the template image, the ROI region is first rotated to angle V to obtain the rotated image. Then, based on the key point parameters M of the template image, the rotated image is scaled to M, thus completing the angle and scale transformation of the ROI region.

Images