A method for calculating the edge distance of a cleaning robot for a floating photovoltaic power station at sea

Abstract

The present application relates to the technical field of visual detection, and discloses a kind of offshore floating photovoltaic power station cleaning robot edge distance calculation method, comprising the following steps: S1, image and radar azimuth acquisition;S2, image correction and pretreatment;S2-1, based on the aforementioned direction angle, image is sequentially subjected to histogram equalization, rotation and projection processing, and the image horizontal line is 0 ° image;S3, edge and corner point detection are carried out by image recognition system, and point set is obtained;S4, point set is used to remove impurities algorithm, and block straight line fitting is carried out, and intercept data is obtained;S5, in combination with camera inclination, height and straight line intercept data, the actual cleaning robot edge distance is calculated.By in the process of robot normal work, distance photovoltaic panel edge angle and distance are judged at any time, to avoid cleaning robot falling into the sea, to improve the cleaning rate of photovoltaic power station and the safety factor of itself.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of visual detection, in particular to a method for calculating the edge distance of a cleaning robot for a floating photovoltaic power station. BACKGROUND

[0002] The cleaning of the photovoltaic panel surface in a floating power station is an important work for its operation and maintenance. The working path of the cleaning robot is difficult to determine and calculate due to the effects of salt crystallization, biological attachment, strong wave power, etc. on the photovoltaic panel surface at sea.

[0003] Under the condition of strong wind and strong wave, most of the existing cleaning robots cannot determine the distance and angle from the edge of the photovoltaic panel at all times, so that the direction and speed of the robot cannot be corrected in time, which leads to the photovoltaic power station cleaning robot falling into the sea, losing the operation and maintenance ability, and also leading to the decrease of the cleaning rate of the cleaning robot for the photovoltaic power station and the safety factor of the robot itself. Therefore, the research on the edge distance calculation method of the cleaning robot for the floating power station at sea based on image recognition is extremely important. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a method for calculating the edge distance of a cleaning robot for a floating photovoltaic power station at sea, which solves the problem that the cleaning robot in the prior art cannot guarantee the cleaning rate and the safety factor of the robot itself under the condition of strong wind and strong wave.

[0005] To achieve the above purpose, the present application is implemented by the following technical scheme: a method for calculating the edge distance of a cleaning robot for a floating photovoltaic power station at sea, comprising the following steps:

[0006] S1, obtaining an image and a radar azimuth angle;

[0007] S2, correcting and preprocessing the image;

[0008] S3, detecting the edge and the corner point through an image recognition system to obtain a point set;

[0009] S4, using a de-noising algorithm on the point set and performing block straight line fitting to obtain intercept data;

[0010] S5, combining the camera tilt angle, height and straight line intercept data to calculate the actual edge distance of the cleaning robot.

[0011] Preferably, S2 specifically comprises:

[0012] S2-1, based on the aforementioned direction angle, sequentially performing histogram equalization, rotation and projection processing on the image to obtain an image with a horizontal line of 0°;

[0013] S2-2, a photovoltaic panel surface dust removal algorithm based on YCbCr color space, adjusting the pixel color of the dust covered position in the foregoing image to obtain a photovoltaic panel dust removal image;

[0014] S2-3, based on the foregoing photovoltaic panel dust removal image, performing clipping cutting and grayscale processing to obtain a fixed-size grayscale image suitable for image recognition algorithm processing.

[0015] Preferably, the S3 specifically comprises:

[0016] S3-1, performing edge detection and edge expansion processing on the foregoing fixed-size grayscale image to obtain an expanded edge grayscale image;

[0017] S3-2, performing large-angle point detection and binary processing according to the expanded edge grayscale image to obtain a large-angle point binary image;

[0018] S3-3, performing sub-pixel angle point detection according to the large-angle point binary image to obtain a large-angle point connected domain center point pixel set, i.e., an original image angle point accurate pixel coordinate data set;

[0019] S3-4, setting a horizontal and vertical coordinate difference threshold value, respectively;

[0020] S3-5, arranging the point set in order of vertical coordinate size;

[0021] S3-6, judging the adjacent points of each point in the point set one by one to obtain a final fitting point set.

[0022] Preferably, the S4 specifically comprises:

[0023] S4-1, according to the edge angle point coordinate characteristics, i.e., in the image coordinates (i, j), the edge angle point coordinate j value is adjacent, which can be determined as one of the point coordinates of a straight line, and a certain number of points are required to form a straight line. According to the above principle, the coordinate set is removed, and a multi-edge point set coordinate is obtained;

[0024] S4-2, according to the foregoing multi-edge point set coordinate, performing one-dimensional linear fitting of each edge point set coordinate to obtain a group of straight line intercept data.

[0025] The application provides an offshore floating photovoltaic power station cleaning robot edge distance calculation method. The application has the following beneficial effects:

[0026] The application can judge the distance and angle of the robot from the photovoltaic panel at any time during normal operation of the robot, correct the moving direction and speed of the robot in time, avoid the cleaning robot from falling into the sea, and improve the cleaning rate of the photovoltaic power station and the safety factor of the robot. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a flowchart of an image recognition-based edge distance calculation method for a marine floating photovoltaic power station cleaning robot;

[0028] Figure 2 is a flowchart of a point set decontamination algorithm. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the specification of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0030] Embodiments:

[0031] Please refer to the drawings in the specification of the present application Figure 1 - the drawings in the specification of the present application Figure 2 The embodiment of the present application provides an edge distance calculation method for a marine floating photovoltaic power station cleaning robot, comprising the following steps:

[0032] S1, image and radar azimuth angle acquisition;

[0033] In the daily robot cleaning process, images are collected from real-time videos of the camera. However, due to a large number of interference factors in the original image, such as different collection times and robot angles, which will cause different overall brightness and effects in each image; in addition, due to different photovoltaic panel arrays of the robot, the edge angle of the photovoltaic panel is not horizontal, which will cause poor straight line fitting effect. Therefore, the images collected at different times need to be corrected and pretreated.

[0034] S2, image correction and pretreatment;

[0035] S2 specifically includes:

[0036] S2-1, based on the aforementioned direction angle, the image is sequentially subjected to histogram equalization, rotation and projection processing to obtain an image with a horizontal line of 0°. For each collected photo, first histogram equalization is performed to make the brightness better distributed on the histogram, so that the image contrast can be enhanced without affecting the overall contrast. Subsequently, the image is rotated and corrected in combination with the azimuth angle obtained by the radar sensor, so that the horizontal line in the image is 0°, and a robot front image of a specified size is intercepted on this basis, thus obtaining a corrected image.

[0037] S2-2, a photovoltaic panel surface dust removal algorithm based on YCbCr color space, adjusts the pixel color of the dust covered position in the foregoing image to obtain a photovoltaic panel dust removal image; and the foregoing corrected image is subjected to dust removal processing. The photovoltaic panel surface will reduce the accuracy of the corner point detection due to dust accumulation, and due to the action of sea wind, the dust will accumulate into irregular curves, which will also affect the subsequent edge recognition, so further pixel correction is required for the corrected image. First, the collected RGB color space is converted into YCbCr color space, from which the following conclusions can be drawn: the brightness of the dust area of the Y channel image is higher than the average brightness of the non-dust area; the dust area of the Cb channel image is darker than the average brightness of the non-dust area; and the dust area of the Cr channel image is higher than the average brightness of the non-dust area. Therefore, the following rules can be customized to process the image pixels one by one:

[0038]

[0039] In the formula: R(x, y) is the pixel color value at (x, y) after final dust removal; P(x, y) is the pixel color value at (x, y) in the original RGB color space; Pmean is the average value of the original RGB pixel color; Ymean is the average value of the Y channel pixel color in the YCbCr color space; and Cbmean is the average value of the Cb channel color in the YCbCr color space.

[0040] The above method can rewrite all dust-like pixel colors, although there may be false positives, but the lines of the photovoltaic panel in the processed image are easier to be recognized, so that the position of the corner point pixel is more accurate.

[0041] S2-3, based on the foregoing photovoltaic panel dust removal image, the image is cut and cut and subjected to grayscale processing to obtain a fixed-size grayscale image suitable for image recognition algorithm processing. After the foregoing steps are completed, the foregoing dust removal image is subjected to image grayscale processing to obtain a final processed image.

[0042] S3, edge and corner point detection is performed through an image recognition system to obtain a point set;

[0043] S3 specifically includes:

[0044] S3-1, edge detection is performed on the foregoing fixed-size grayscale image and edge expansion processing is performed to obtain an expanded edge grayscale image; first, based on the foregoing processed image, edge recognition processing is performed on the image to obtain an edge binary image.

[0045] S3-2, according to the expanded edge grayscale image, a large range of corner point detection is performed and binary processing is performed to obtain a large corner point binary image; then, according to the edge binary image, the image is sequentially subjected to expansion and corner point area recognition to obtain a group of corner point connected domain data.

[0046] S3-3, according to the large angle point binary image, sub-pixel angle point detection is made, a large angle point connected domain center point pixel set is obtained, that is, an original image angle point accurate pixel coordinate data set; finally, sub-pixel angle point detection is made according to the angle point region, and an accurate angle point coordinate data set is acquired.

[0047] S3-4, a horizontal and vertical coordinate difference threshold is respectively set; and a horizontal and vertical coordinate judgment threshold is set according to the foregoing target coordinate feature.

[0048] S3-5, the point set is arranged in order of vertical coordinate size; the point set is sorted according to the vertical coordinate size, so that the edge angle point coordinates are adjacent in the set, and the difficulty of removing impurities is reduced.

[0049] S3-6, the adjacent points of each point in the point set are judged one by one, and a final to-be-fitted point set is acquired; finally, the foregoing to-be-fitted point set is subjected to one-dimensional linear fitting, and the linear intercept data can be obtained.

[0050] S4, the point set is subjected to a decontamination algorithm, and is subjected to block linear fitting, and intercept data is acquired;

[0051] S4-1, according to the edge angle point coordinate characteristics, that is, in the image coordinates (i, j), the edge angle point coordinate j value is adjacent, which can be determined as one of the point coordinates of a straight line, and a certain number of points are needed to form a straight line. According to the above principle, the coordinate set is decontaminated, and a multi-edge point set coordinate is obtained.

[0052] S4-2, according to the foregoing multi-edge point set coordinate, one-dimensional linear fitting is performed on each edge point set coordinate, and a group of linear intercept data can be obtained.

[0053] S5, the actual edge distance of the cleaning robot is calculated in combination with the camera tilt angle, height and linear intercept data.

[0054] The actual distance calculation method in combination with the camera tilt angle, height and linear intercept includes:

[0055] According to the relative position of the camera and the ground and the camera installation angle, the relationship between the image edge linear intercept and the real distance can be calculated:

[0056] y=kx+b

[0057] In the formula, k is a mapping coefficient, which is calculated from the edge coordinates in the photo collected by the robot camera and the actual distance between the edge and the robot; x is the intercept of the edge fitting straight line in the collected image; and b is an adjustment coefficient, which is mainly used to make the mapping more accurate.

[0058] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A method for calculating the edge distance of a cleaning robot for a floating photovoltaic power plant at sea, characterized in that, The method comprises the following steps: S1, image and radar azimuth angle acquisition; S2, image correction and preprocessing; specifically comprising: S2-1, based on the aforementioned direction angle, sequentially performing histogram equalization, rotation and projection processing on the image to obtain an image with a horizontal line of 0°; S2-2, a photovoltaic panel face dust removal algorithm based on YCbCr color space adjusts the pixel color of the dust covered position in the aforementioned image to obtain a photovoltaic panel dust removal image. First, the collected RGB color space is converted into YCbCr color space, and the image pixels are processed one by one according to the following rules: In the formula: R(x, y) is the pixel color value at (x, y) after final dust removal; P(x, y) is the pixel color value at (x, y) in the original RGB color space; Pmean is the average value of the original RGB pixel color; Ymean is the average value of the Y channel pixel color in the YCbCr color space; Cbmean is the average value of the Cb channel color in the YCbCr color space; S2-3, based on the aforementioned photovoltaic panel dust removal image, performing cutting and gray processing to obtain a fixed-size gray image suitable for image recognition algorithm processing; S3, edge and corner point detection is performed by an image recognition system to obtain a point set; specifically comprising: S3-1, edge detection is performed on the aforementioned fixed-size gray image and edge expansion processing is performed to obtain an expanded edge gray image; S3-2, according to the expanded edge gray image, a large range of corner point detection is performed and binary processing is performed to obtain a large corner point binary image; S3-3, according to the large corner point binary image, sub-pixel corner point detection is performed to obtain a large corner point connected domain center pixel point set, i.e. an original image corner point accurate pixel coordinate data set; S3-4, respectively setting the horizontal and vertical coordinate difference threshold values; S3-5, arranging the point set in order of vertical coordinate size; S3-6, judging the adjacent points of each point in the point set one by one to obtain a final fitting point set; S4, using a decontamination algorithm on the point set and performing block straight line fitting to obtain intercept data; S5, combining the camera inclination angle, height and straight line intercept data to calculate the actual cleaning robot edge distance.

2. The method of claim 1, wherein the edge distance is calculated by the following equation: edge distance = (L - 2 * (R + r)) / 2, where L is the length of the cleaning robot, R is the radius of the cleaning robot, and r is the radius of the photovoltaic panel. The S4 specifically comprises: S4-1, according to the edge corner point coordinate characteristics, i.e. in the image coordinates (i, j), the edge corner point coordinate j value is adjacent to determine that it is one of the point coordinates of a straight line, and a certain number of points are required to form a straight line; according to the above principle, the coordinate set is decontaminated to obtain a multi-edge point set coordinate; S4-2, according to the aforementioned multi-edge point set coordinate, performing one-dimensional linear fitting on each edge point set coordinate to obtain a set of straight line intercept data.

Images