CN114392964B - Insulator rinse-system that dirty degree of intellectuality was judged - Google Patents

Abstract

The invention discloses an insulator flushing system for intelligently judging the degree of contamination, which comprises a reference insulator, a high-definition camera group, a searchlight, a two-degree-of-freedom water cannon device, a centrifugal pump, a control box and a water tank, wherein the reference insulator, the high-definition camera group, the searchlight, the two-degree-of-freedom water cannon device, the centrifugal pump, the control box and the water tank are sequentially arranged on a frame from front to back; positioning the insulator by using a binocular positioning camera, and evaluating the dirt degree of the target insulator by using a monocular high-definition camera; and then controlling the double-freedom-degree water cannon device to wash the insulator. The invention can accurately identify the dirt degree of the surface of the insulator, improve the efficiency and the precision of the insulator water washing and avoid the occurrence of pollution flashover accidents.

Description

An intelligent insulator flushing system for judging the degree of dirt

technical field

The invention belongs to the field of cleaning insulators in rail transit, and in particular relates to an insulator cleaning system for intelligently judging the degree of dirt.

Background technique

With the rapid development of my country's rail transit industry, domestic electrified railways are gradually increasing, which play a vital role in the national economy and people's quality of life. The catenary system of the electrified railway is mainly used to supply power to the electric locomotive, which is transformed by the traction substation and then sent to the transmission line. An important part of.

In the rail transit catenary, the insulator is an important part to isolate the current-carrying conductor from the ground and ensure the safety of the railway system. Due to the influence of the environment, the surface of the insulator often has stains attached to it. The composition of the stains varies depending on the geographical environment. Similarly, it is usually dominated by a large amount of dust, coal slag in the mine, metal salt particles, etc. When the accumulation of stains is too serious, due to the potential difference between the two ends of the insulator, it is easy to cause creepage phenomenon, and in severe cases, pollution flashover accidents will occur , which has become one of the hazards in electrified railways. In order to avoid the occurrence of pollution flashover accidents, it needs to be cleaned regularly, but at home and abroad, manual control of flushing is still the main method, and there are still situations in China where manual wiping of insulators is used, which not only reduces cleaning and maintenance efficiency, but also has many problems. major safety hazard.

With the development of modern science and technology, binocular vision, as an important branch of machine vision, has been widely used in aerial surveying and mapping, military applications, medical imaging and industrial inspection after decades of development. Binocular stereo vision is based on the principle of parallax and uses imaging equipment to obtain the left and right images of the measured object from different positions, and then calculates the position deviation of the two-dimensional image of the space point according to the principle of triangulation, so as to use the position deviation to perform three-dimensional reconstruction to obtain the measured object The three-dimensional geometric information of the object. The deep learning neural network can complete the recognition of a specific object by training the characteristic parameters such as the shape, texture, and color of an object. Combining it with binocular vision, the target insulator can be found in the real-time image, thereby completing The purpose of intelligent identification and positioning.

In previous inventions and designs, the original database stored by the computer was used to identify insulators in specific environments. However, for insulators with different backgrounds, environments, and lighting, the color, degree of reflection, and shooting angle will all affect the characteristic parameters of the image. , although the spatial position can be found according to the outline, but the degree of dirt on the surface of the insulator cannot be accurately judged according to the color characteristics.

Contents of the invention

In order to improve the efficiency and precision of water flushing of insulators and avoid pollution flashover accidents, the present invention provides an insulator flushing system with intelligent judgment of pollution degree.

An intelligent insulator flushing system for judging the degree of dirt of the present invention is specifically: a reference insulator, a high-definition camera group, a searchlight, a two-degree-of-freedom water cannon device, a centrifugal pump, a control box, and a water tank are sequentially installed on the vehicle frame from front to back .

The two-degree-of-freedom water cannon device includes a rotary transmission module composed of a rotary motor, a rotary reducer and a slewing support, a pitch transmission module composed of a pitching motor, a pitching reducer and a gun rod, and a gun body and a base. The protective structure and the binocular positioning camera installed on the gun barrel. Limit sensors are arranged on the slewing bearing and gun rod.

The camera base of the high-definition camera group is installed on the vehicle frame through a bracket. A height adjustment motor is set on the bracket, and the height adjustment motor adjusts the height of the monocular high-definition camera through toothed belt transmission; a level adjustment motor is set on the camera base, and the level adjustment motor adjusts the horizontal rotation of the monocular high-definition camera through toothed belt transmission; A waterproof baffle is set on the top of the monocular high-definition camera.

A PLC controller, a computer, a drive controller and a solenoid valve are arranged in the control box. The computer is connected to the binocular positioning camera and the monocular high-definition camera; the PLC controller is connected to the slewing motor and pitch motor, horizontal adjustment motor and height adjustment motor;

The two-degree-of-freedom water cannon device is connected to the centrifugal pump through the water pipe, and then connected to the water tank through the water pipe to absorb water.

The insulator positioning and dirt degree recognition program and the insulator flushing control program are set in the computer.

Further, the insulator positioning and contamination degree identification procedures are as follows:

Step 1: First, grayscale the image collected by the binocular positioning camera, take the grayscale value of each pixel, and mark the average grayscale value inside the insulator. According to the shape and texture characteristics of the collected insulator pictures, different The eigenvalues in the background environment are distinguished, and the recognition of insulators is trained through deep learning algorithms.

Step 2: Due to the one-to-one correspondence relationship between the imaging pixels of the same object in the left and right cameras under the same external environment, according to the transformation matrix and the spatial position difference of the two cameras, restore the spatial coordinates of the central feature point of the insulator, PLC The controller adjusts the gun rod to aim at the target insulator according to the three-dimensional coordinates.

Step 3: According to the spatial coordinate difference between the binocular positioning camera and the monocular high-definition camera, and the existing rotation angle of the gun barrel, use the space rotation matrix to convert the coordinates of the target insulator into the coordinate system with the monocular high-definition camera as the origin, and PLC control According to the converted coordinates, the controller adjusts the corner of the monocular high-definition camera to aim at the target insulator to take high-definition pictures.

Step 4: According to the color space theory, calculate the color mean X ₁ , median X ₂ , variance X ₃ , range X ₄ , skewness X ₅ , and kurtosis of the target insulator in the high-definition image captured by the monocular high-definition camera X ₆ , energy value X ₇ and entropy value X ₈ , calculate the target insulator total color feature function P ₁ =f(X ₁ , X ₂ , X 3 , X ₄ , X ₅ , X ₆ , X ₇ , _{X 8 )} .

Step 5: According to the gray-scale co-occurrence matrix GLCM, calculate the layered distribution map of different gray-scale values i, and generate the target insulator edge texture image feature function P ₂ =f(i ₁ ,i ₂ ,i ₃ ,i ₄ ,i ₅ , _i6 ).

Step 6: Comparing the two characteristic functions of the target insulator and the reference insulator, calculating the deviation value, and evaluating the degree of contamination of the target insulator according to the deviation value.

Further, the degree of dirt is divided into five categories: A: very light, B: light, C: medium, D: heavy, E: very heavy.

Further, the conversion of space coordinates in step 3 is specifically:

Calculate the three-dimensional coordinates (x ₁ , y ₁ , z ₁ ) of the target insulator in the coordinate system 1 of the binocular positioning camera, and calculate the current rotation angle θ of the gun shaft; since the coordinate origin O ₂ of the monocular high-definition camera is in the coordinate system 1 The coordinates are fixed values and conform to the vector calculation of the space rotation matrix E, so through the space rotation displacement matrix equation E, the three-dimensional coordinates (x ₂ , y ₂ , z ₂ ) of the target insulator under the coordinate system 2 are obtained.

Further, the insulator flushing control program is specifically:

After the gun barrel is aimed at the target insulator, compare the two characteristic functions of the target insulator with the clean reference insulator, calculate the deviation value, and evaluate the degree of dirtiness of the target insulator according to the deviation value, if the degree of dirtiness is lower than the set safety value , then skip the insulator and identify the next target; if the degree of dirt is higher than the set safety value, the computer sends a signal to drive the controller to open the solenoid valve to flush the target insulator.

Further, when the target insulator is compared with the clean reference insulator, an auxiliary light source is provided through a searchlight to ensure that the target insulator and the clean reference insulator have the same lighting conditions and the same direction.

The beneficial technical effect of the present invention is:

With the support of the existing deep learning and intelligent recognition technology, the present invention takes the image information of the clean reference insulator at the same time, the same external conditions and the same shooting angle into consideration with the insulator to be washed in consideration of the influence of the external environment and illumination. In contrast, the processing technology of color feature function and texture feature function in high-definition images is used to complete the judgment of the degree of dirt on the surface of the insulator, thereby achieving the purpose of identifying stains and selective flushing, and improving the accuracy and efficiency of insulator water flushing.

Description of drawings

Fig. 1 is an overall structural diagram of an insulator flushing system for intelligently judging the degree of dirt of the present invention.

Fig. 2 is a schematic structural diagram of the dual-degree-of-freedom water cannon device of the insulator flushing system for intelligently judging the degree of dirt of the present invention.

Fig. 3 is a schematic structural diagram of a high-definition camera group of an insulator flushing system for intelligently judging the degree of dirt of the present invention.

Fig. 4 is a schematic diagram of the positioning of the insulator and the identification of the degree of dirt in the present invention.

Fig. 5 is a schematic diagram of coordinate transformation in the present invention.

Fig. 6 is a principle diagram of insulator flushing control in the present invention.

Explanation of symbols in Figure 1-3: 1-frame; 2-water tank; 3-control box; 4-centrifugal pump; 5-double degree of freedom water cannon device; 51-rotary motor; 52-rotary reducer; Motor; 54-pitch reducer; 55-gun rod; 56-gun body; 57-limit sensor; 58-slewing bearing; 59-base; 6-HD camera group; 61-height adjustment motor; 62-camera base; 63-monocular high-definition camera; 64-waterproof baffle; 65-toothed belt; 66-level adjustment motor; 67-bracket; 7-binocular positioning camera; 8-searchlight; 9-reference insulator.

Detailed ways

The present invention will be described in further detail below in conjunction with the accompanying drawings and specific implementation methods.

An intelligent insulator flushing system for judging the degree of dirt of the present invention is shown in Figure 1, specifically: a reference insulator 9, a high-definition camera group 6, a searchlight 8, and a two-degree-of-freedom water cannon are installed on the frame 1 in sequence from front to back Device 5, centrifugal pump 4, control box 3 and water tank 2.

The two-degree-of-freedom water cannon device 5 as shown in Figure 2 specifically includes a rotary transmission module composed of a rotary motor 51, a rotary reducer 52 and a rotary bearing 58, and is formed by a pitching motor 53, a pitching reducer 54 and a gun rod 55. The pitch transmission module that constitutes, the protective structure that is formed by gun body 56 and base 59, and the binocular positioning camera 7 that is installed on the gun rod 55. A limit sensor 57 is arranged on the slewing support 58 and the gun rod 55 to ensure the normal reset of the gun rod 55 and to prevent internal circuit damage caused by excessive rotation angle. The high-pressure jet technology is used inside the gun barrel to process the high-pressure water flow so that it can complete the flushing work with charged water.

As shown in FIG. 3 , the camera base 62 of the high-definition camera group 6 is installed on the vehicle frame 1 through a bracket 67 . The height adjustment motor 61 is set on the support 67, and the height adjustment motor 61 is adjusted the height of the monocular high-definition camera 63 through the transmission of the toothed belt 65; Adjust the horizontal rotation of the monocular high-definition camera 63. The top of the monocular high-definition camera 63 is provided with a waterproof baffle 64.

A PLC controller, a computer, a drive controller and a solenoid valve are arranged in the control box 3 . The computer is connected to the binocular positioning camera 7 and the monocular high-definition camera 63; the PLC controller is connected to the rotary motor 51 and the pitch motor 53, the horizontal adjustment motor 66 and the height adjustment motor 61;

The two-degree-of-freedom water cannon device 5 is connected to the centrifugal pump 4 through a water pipe, and then connected to the water tank 2 through a water pipe to absorb water.

The insulator positioning and dirt degree recognition program and the insulator flushing control program are set in the computer.

Further, the principle of insulator positioning and dirt degree identification is shown in Figure 4, specifically:

Step 1: First, grayscale the image collected by the binocular positioning camera 7, take the grayscale value of each pixel, and mark the average grayscale value inside the insulator, according to the shape and texture features of the collected insulator pictures, The eigenvalues under different background environments are distinguished, and the recognition of insulators is trained through deep learning algorithms.

Step 2: Due to the one-to-one correspondence relationship between the imaging pixels of the same object in the left and right cameras under the same external environment, according to the transformation matrix and the spatial position difference of the two cameras, restore the spatial coordinates of the central feature point of the insulator, PLC The controller adjusts the gun rod 55 to align with the target insulator according to the three-dimensional coordinates.

Step 3: According to the spatial coordinate difference between the binocular positioning camera 7 and the monocular high-definition camera 63, and the existing rotation angle of the gun shaft 55, use the spatial rotation matrix to transform the target insulator coordinates into a coordinate system with the monocular high-definition camera 63 as the origin Inside, the PLC controller adjusts the 63 corners of the monocular high-definition camera according to the converted coordinates, so that it can aim at the target insulator to take high-definition pictures.

Step 4: According to the color space theory, calculate the color mean value X ₁ , median value X ₂ , variance X ₃ , range X ₄ , skewness value X ₅ , and kurtosis of the target insulator in the high-definition image captured by the monocular high-definition camera 63 value X ₆ , energy value X ₇ and entropy value X ₈ , calculate the target insulator total color characteristic function P ₁ =f(X ₁ ,X ₂ ,X ₃ ,X 4 ,X ₅ ,X ₆ ,X ₇ ,X _{8 )} .

Step 5: According to the gray level co-occurrence matrix GLCM, calculate the layered distribution map of different gray values i (here set as 6 layers), and generate the target insulator edge texture image feature function P ₂ =f(i ₁ ,i ₂ , i ₃ , i ₄ , i ₅ , i ₆ ).

Step 6: Comparing the two characteristic functions of the target insulator and the reference insulator 9, calculating the deviation value, and evaluating the degree of contamination of the target insulator according to the deviation value.

Further, the degree of dirt is divided into five categories: A: very light, B: light, C: medium, D: heavy, E: very heavy.

Further, as shown in Figure 5, the conversion of the spatial coordinates in step 3 is specifically:

Calculate the three-dimensional coordinates (x ₁ , y ₁ , z ₁ ) of the target insulator located in the coordinate system 1 of the binocular positioning camera 7, and calculate the current rotation angle θ of the gun shaft 55; since the coordinate origin O ₂ of the monocular high-definition camera 63 is in the coordinate system The coordinates in 1 are fixed values and conform to the vector calculation of the space rotation matrix E, so the three-dimensional coordinates (x ₂ , y ₂ , z ₂ ) of the target insulator in the coordinate system 2 are obtained through the space rotation displacement matrix equation E.

Further, as shown in Figure 6, the insulator flushing control program is specifically:

After the gun rod 55 is aimed at the target insulator, compare the two characteristic functions of the target insulator with the clean reference insulator 9, calculate the deviation value, and evaluate the degree of dirtiness of the target insulator according to the deviation value, if the degree of dirtiness is lower than the set If the level of contamination is higher than the set safety value, the computer will send a signal to drive the controller to open the solenoid valve to flush the target insulator.

Further, when the target insulator is compared with the clean reference insulator 9, the searchlight 8 is used to provide an auxiliary light source to ensure that the target insulator and the clean reference insulator 9 have the same lighting conditions and the same direction, so that accurate contamination recognition can be achieved and the flushing effect is better.

Claims (5)

1. The utility model provides an insulator washing unit that intelligent dirty degree was judged, its characterized in that, the device specifically is: a reference insulator (9), a high-definition camera group (6), a searchlight (8), a two-degree-of-freedom water cannon device (5), a centrifugal pump (4), a control box (3) and a water tank (2) are sequentially arranged on the frame (1) from front to back;

the two-degree-of-freedom water cannon device (5) comprises a rotary transmission module, a pitching transmission module, a protection structure and a binocular positioning camera (7), wherein the rotary transmission module is composed of a rotary motor (51), a rotary speed reducer (52) and a rotary support (58), the pitching transmission module is composed of a pitching motor (53), a pitching speed reducer (54) and a cannon rod (55), the protection structure is composed of a cannon body (56) and a base (59), and the binocular positioning camera (7) is installed on the cannon rod (55); a limit sensor (57) is arranged on the slewing bearing (58) and the gun rod (55);

the camera base (62) of the high-definition camera group (6) is arranged on the frame (1) through a bracket (67); a height adjusting motor (61) is arranged on the bracket (67), and the height adjusting motor (61) adjusts the height of the monocular high-definition camera (63) through the transmission of a toothed belt (65); a horizontal adjusting motor (66) is arranged on the camera base (62), and the horizontal adjusting motor (66) adjusts the horizontal rotation of the monocular high-definition camera (63) through the transmission of a toothed belt (65); a waterproof baffle plate (64) is arranged at the top of the monocular high-definition camera (63);

a PLC controller, a computer, a driving controller and an electromagnetic valve are arranged in the control box (3); the computer is connected with the binocular positioning camera (7) and the monocular high-definition camera (63); the PLC is connected with a rotary motor (51), a pitching motor (53), a horizontal adjusting motor (66) and a height adjusting motor (61);

the two-degree-of-freedom water cannon device (5) is connected with the centrifugal pump (4) through a water pipe and is connected with the water tank (2) through a water pipe to absorb water;

the computer is provided with an insulator positioning and dirt degree identification program and an insulator washing control program;

the operation steps of the insulator positioning and contamination degree identification program are specifically as follows:

step 1: firstly, carrying out graying processing on an image collected by a binocular positioning camera (7), taking a gray value for each pixel point, marking an average value of the gray value inside an insulator, distinguishing characteristic values under different background environments according to the shape characteristic and the texture characteristic of a collected insulator picture, and training the identification of the insulator through a deep learning algorithm;

and 2, step: because the same object is in the same condition of the external environment, the imaging pixel points in the left camera and the right camera are in one-to-one correspondence, the space coordinate of the central characteristic point of the insulator is restored according to the transformation matrix and the space position difference of the two cameras, and the PLC adjusts the gun rod (55) to be aligned to the target insulator according to the three-dimensional coordinate;

and 3, step 3: according to the space coordinate difference between the binocular positioning camera (7) and the monocular high-definition camera (63) and the existing rotation angle of the gun barrel (55), a space rotation matrix is used for converting the coordinates of the target insulator into a coordinate system with the monocular high-definition camera (63) as an original point, and the PLC controller adjusts the rotation angle of the monocular high-definition camera (63) according to the converted coordinates to enable the monocular high-definition camera to be aligned with the target insulator to shoot high-definition pictures;

and 4, step 4: according to the color space theory, the color mean value of the target insulator in the high-definition image shot by the monocular high-definition camera (63) is calculatedX ₁ Median value ofX ₂ Variance, varianceX ₃ Very poor in qualityX ₄ Deviation value ofX ₅ Kurtosis valueX ₆ Energy value ofX ₇ And entropy valueX ₈ Calculating the total color characteristic function of the target insulatorP ₁ =f（X ₁ ,X ₂ ,X ₃ ,X ₄ ,X ₅ ,X ₆ ,X ₇ ,X ₈ ）；

And 5: calculating different gray values according to the gray level co-occurrence matrix GLCMiThe layered distribution diagram generates the characteristic function of the edge texture image of the target insulatorP ₂ =f（i ₁ ,i ₂ ,i ₃ ,i ₄ ,i ₅ ,i ₆ ）；

Step 6: and comparing the two characteristic functions of the target insulator and the reference insulator (9), calculating a deviation value, and evaluating the contamination degree of the target insulator according to the deviation value.

2. The insulator washing device for intelligently judging the degree of contamination according to claim 1, wherein the degree of contamination is divided into five categories: a: very light, B: light weight and C: medium, D: heavy, E: it is heavy.

3. The insulator washing device for intelligently judging the degree of contamination according to claim 1, wherein the transformation of the spatial coordinates in the step 3 is specifically as follows:

calculating the three-dimensional coordinates of the target insulator in the coordinate system 1 of the binocular positioning camera (7) ((x ₁ ,y ₁ ,z ₁ ) And calculating the current rotation angle of the gun barrel (55)θ(ii) a Because the coordinate origin of the monocular high-definition camera (63)O ₂ The coordinates in the coordinate system 1 are fixed values and conform to the vector calculation of the space rotation matrix E, and then the three-dimensional coordinates of the target insulator under the coordinate system 2 are obtained through the space rotation displacement matrix equation E (step (c)) (x ₂ ,y ₂ ,z ₂ ）。

4. The insulator washing device for intelligently judging the degree of contamination according to claim 1, wherein the insulator washing control program specifically comprises the following operation steps:

after a gun barrel (55) aims at a target insulator, comparing the target insulator with two characteristic functions of a clean reference insulator (9), calculating a deviation value, evaluating the contamination degree of the target insulator according to the deviation value, and if the contamination degree is lower than a set safety value, skipping the insulator and identifying the next target; and if the contamination degree is higher than a set safety value, the computer sends a signal, and the drive controller opens the electromagnetic valve to flush the target insulator.

5. The insulator washing device for intelligently judging the degree of contamination is characterized in that when the target insulator is compared with the clean reference insulator (9), an auxiliary light source is provided through a searchlight (8), so that the target insulator and the clean reference insulator (9) have the same lighting condition and the same direction.

Images