Intelligent sensing method for external damage prevention information of power transmission line

Abstract

The invention discloses a power transmission line external damage prevention information intelligent sensing method, and belongs to the technical field of power transmission line intelligent sensing, and the method comprises the following steps: 1, constructing a weather preprocessing model; 2, constructing an external damage prevention information processing model; and step 3, building a tower monitoring hardware system. The construction of the weather preprocessing model comprises construction of an atmospheric light scattering model, a fog region detection model and a defogging model, and the atmospheric light scattering model, the fog region detection model and the defogging model are combined with one another and are used for sensing external damage information of the power transmission line in foggy days. According to the invention, external damage information can be timely transmitted to a control center or a mobile device of an operator during external damage of the power transmission line, so that the operator can timely deal with the external damage information, and the safety of the power transmission line is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent perception of transmission lines, and particularly relates to an intelligent perception method for preventing external damage information of transmission lines. Background Art

[0002] At present, monitoring the tower state and the external intrusion state of transmission lines faces many technical and economic problems. Geographical location, environmental conditions and human factors are all important factors affecting the safety and stability of power grid transmission. Geographically, in North China of our country, there are various terrains including the North China Plain and the Loess Plateau. Among them, the Loess Plateau is the region with the most serious soil erosion and the most fragile ecological environment in the world. Moreover, it is very difficult to maintain long-term flatness in terms of terrain. Most of the land in these areas belongs to thick-layered loess. The biggest feature of this type of land is that the soil is soft and is eroded by flowing water for a long time. The large dispersion and geological problems make the towers built here extremely vulnerable to external damage, including tower inclination, displacement, etc., which further threatens the transmission lines. The situation in East China is different from that in North China. This region usually does not involve line damage caused by terrain problems. However, since most of the regions are economically developed areas, there is a situation where the high-voltage wires erected will pass through urban and rural areas with large population flows. Therefore, it is difficult to supervise and maintain the dangerous areas formed under the lines and below them. Due to the complex environmental conditions where the transmission lines in our country are located, the lines are extremely vulnerable to external damage, including disasters such as line icing, aging, and natural fires in adjacent forest areas. It also includes some human factors. The destruction and theft of power facilities by humans are also relatively frequent. Among them, illegal construction by large machinery has become the main factor for damaging high-voltage lines. Because the overhead high-voltage lines are far from the ground, construction workers on the ground occasionally overlook or fail to control the distance between large machinery and the lines properly, resulting in external damage faults of power lines.

[0003] The large-scale coverage of overhead lines combined with the above various situations has led to an increasing number of cases where transmission lines and towers are damaged externally. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art, and provide an intelligent perception method for preventing external damage information of transmission lines, which solves the problems in the above background art.

[0005] The purpose of the present invention is achieved as follows: An intelligent perception method for preventing external damage information of transmission lines, which includes the following steps: Step 1, construct a weather preprocessing model; Step 2, construct an external damage prevention information processing model; Step 3, build a tower monitoring hardware system.

[0006] The construction of the weather preprocessing model includes constructing an atmospheric light scattering model, a fog area detection model, and a defogging model, and combining the atmospheric light scattering model, the fog area detection model, and the defogging model to be used for the perception of external damage information of transmission lines in foggy weather;

[0007] The atmospheric light scattering model includes dark image enhancement, planetary image enhancement. Natural images, whether foggy or not, are composed of an incident component and a reflection component; the incident component represents the brightness information of the image, and the reflection component represents the internal information of the image. Its model is:

[0008] S(x, y) = L(x, y)·R(x, y)

[0009] Among them, S is the original foggy image, R is the reflection component, L is the incident component, and (x, y) is the image coordinate;

[0010] Based on the Retinex defogging algorithm, R is obtained from S. Taking the logarithm of both sides of the equation gives:

[0011] logS(x, y) = logL(x, y) + logR(x, y)

[0012] To achieve dynamic range compression, color or brightness restoration, in order to improve color fidelity, a single-scale Retinex (SSR) algorithm is constructed, and the convolution operation of the Gaussian function and the foggy image is used to estimate the incident component. The formula is:

[0013] L(x, y) = S(x, y) * G(x, y)

[0014] Among them, * represents the convolution operation, and G(x, y) is the Gaussian function; in order to make up for the shortcomings of SSR, a multi-scale Retinex algorithm and a multi-scale Retinex algorithm with color restoration are set; the MSR algorithm uses multiple Gaussian functions with different scales in the RGB channels to calculate the output of the SSR algorithm and sums them up weighted on this basis.

[0015]

[0016] In the formula, ω j represents the weight of each Gaussian function, all of which are 1 / 3, and the image pixel values are set within the range of [0, 255]; the MSR algorithm has the advantages of small-scale dynamic distance compression and image enhancement. In order to improve the restoration of colors in the image, color restoration is used to control saturation.

[0017] The fog area detection model includes creating a semi-inverse image The semi-inverse image is obtained by replacing the RGB value of each pixel on each channel with the maximum value after the original image is inverted. The formula is:

[0018]

[0019] In the formula, r, g, and b are the three original channel values. In the fog-free area, since at least one channel has the characteristic of a small value, the operation will replace this value with its reciprocal. In the sky or haze area, since all values have the characteristic of high values, the max operation will return the same value; by comparing with the hue H of the original image in the HSV color space, the pixels to be restored are found, and the color appearance similar to the original image is maintained;

[0020] To identify the areas affected by the haze, calculate the difference between the hue H channels of the original image I and the semi-inverse image I si and perform thresholding on it using a predefined value τ; the value of τ is used to select the pixels with a similar appearance in the initial and semi-inverse versions. The formula for judging the influence of the haze degree on the haze image is

[0021]

[0022] In the formula, I H and respectively represent the H channels of I and I si . Generated using τ = 10, and only the pixels τ with a hue difference lower than this threshold are marked as blurred pixels, otherwise it is a clear image.

[0023] The defogging model includes edge-aware image defogging. By the weighted guided filter method, the edge-aware weight is introduced to replace the regularization parameter ε in the GIF. The edge-aware weight formula is

[0024]

[0025] In the formula, p′ is the pixel at the center of the window, ζ is the radius of the window ω, and p is the pixel around p′, is the variance in the window ω with p′ as the center pixel; ε takes the value of (0.001×L) 2 , where L is the dynamic range of the pixel values of the actual image; when an edge appears in the window centered on p′, that is, Γ G (p′) > 1, otherwise Γ G (p′) ≤ 1; used to realize the perception of edges with different degrees of gradient changes.

[0026] The anti-external breakage information processing model includes moving target detection. Using the Gaussian distribution and linearly weighting it to represent the distribution law of these pixels, the adaptive mixture Gaussian model is used to detect the target; the one-dimensional Gaussian distribution probability function is

[0027]

[0028] where μ and σ 2 represent the mean and variance of the Gaussian distribution respectively. The one-dimensional Gaussian distribution probability function is used to expand the joint density function of the multi-dimensional variable X = (x1, x2, … x n ):

[0029]

[0030] where d represents the dimension of the variable, u represents the mean of each dimensional variable, and Σ represents the covariance matrix, which describes the correlation degree of each dimensional variable. For intuitive representation, a two-dimensional Gaussian distribution is taken as an example. Therefore, d = 2, and there is:

[0031]

[0032] The scatter points are distributed according to the single Gaussian probability function, and its Gaussian parameters are ideal. The data is estimated using the single Gaussian probability function. The measured effect is that the sample size in the central region of the scatter points is set to be greater than that at the edge, and the number of samples distributed in the central region is set to be less than that in the edge region;

[0033] The mixture Gaussian background model is adopted to suppress the interference caused by the dynamic background. The mixture Gaussian distribution of the image pixel points is obtained by linearly adding K Gaussian distributions. The formula is

[0034]

[0035] where K is the number of Gaussian distributions, and each Gaussian distribution represents a component; the larger the K value, the more components, and the stronger the adaptability of the model to complex backgrounds. The computational amount is taken as 3 - 5; w i,t represents the weight of the i-th Gaussian distribution component at time t, and u i,t represents the mean vector of the i-th Gaussian distribution component at time t, and I t (x, y) represents the pixel value of the image at the position (x, y) at time t.

[0036] The update steps of the Gaussian distribution model are

[0037] (1) Using the conditions of the above formula, the difference d i,t-1 between the pixel value I(x, y) and the mean u i of any Gaussian distribution component at the corresponding position is i compared with 2.5σ i ; if d i ≤ 2.5σ

[0038] then execute (2), otherwise execute (3); i (2) It is determined that the pixel value I(x, y) matches the component G iThe weights are used and step (4) is executed;

[0039] (3) It is determined that the pixel value I(x, y) does not match the component G i If the match fails, the weight of G i is decreased. If the weight becomes negative, the Gaussian distribution component is deleted, and it returns to (1) and is executed until all distributions have completed the match with I(x, y) and step (5) is executed;

[0040] (4) Update G i according to the above formula for the mean, variance, and weight, and sort the distributions in M p in descending order according to the ratio of the weight w to the variance σ;

[0041] (5) If the number of successful matches of Ip is 0 at this time, a new Gaussian distribution G * is added. If the number of Gaussian distribution components for the current pixel has not reached the upper limit, it is directly added. Otherwise, the last component in the sorted Mp is replaced, and the sorting is redone in descending order;

[0042] (6) The sum of the weights of all Gaussian distribution components in Mp is calculated and all weights are normalized;

[0043] (7) Using the above formula conditions, B Gaussian distribution components are obtained as the background model for the next frame of the image at (x, y).

[0044] The tower monitoring hardware system includes a processor, an inclination detection module, an RGB capacitive touch screen, GPRS, a storage module, and a power supply module;

[0045] After power-on, the system first loads the character library and emWin-related images, and then enters the parameter setting and display interface. The parameters can be set in two ways: through the mobile phone APP or the touch screen. After the parameter setting is completed, the measurement is started. The processor collects and processes the data related to the tower state and displays it on the RGB screen in real time. In terms of data storage, the data is generated into a file and stored in the front-end SD card, and at the same time, the data is transmitted to the mobile phone APP and the cloud server in real time.

[0046] The processor uses the low-power STM32F429 with an ARMCortex-M4 core as the main control MCU. The internal data path, registers, and memory interfaces are all set to 32 bits; an extended LCD-TFT controller and a variable storage controller are set;

[0047] The power supply module uses an XL1509 voltage regulator chip to regulate the voltage to 5V for powering the sensor to simultaneously meet the power supply needs of the sensor and the core processor. At the same time, the output 5V is regulated by AMS1086 to generate 3.3V voltage for powering the core processor, SD card, and other peripheral modules;

[0048] The power supply voltage of the inclination sensor of the inclination detection module is 5V. When communicating with the processor, the judgment of the input high level is V IH > 4V. For the ARM processor with a power supply voltage of 3.3V, a bi-directional level conversion circuit is adopted to establish communication between the processor and the sensor.

[0049] The beneficial effects of the present invention are as follows: Firstly, for the detection of foggy areas, the processing speed is improved. The haze coverage degree of the image is detected by comparing the error between the semi-inverse image and the original image. When the coverage degree reaches the threshold, the image is de-fogged. Then, for the images that need to be de-fogged, a de-fogging model is established. The atmospheric light scattering model is used for image de-fogging, and algorithm analysis is carried out on two key points, namely atmospheric light and transmittance, in the model, and the solution method in this paper is proposed. Among them, for the estimation of atmospheric light, the sky area in the image is first identified and then further estimated. For the transmittance, the initial transmittance of the local area of the image is first estimated by optimizing the contrast, and then the initial transmittance map is refined by weighted guided filtering. After the image is de-fogged, the color deviation that often appears is corrected by color transfer transmission.

[0050] For the detection of moving targets, the mixed Gaussian background model is adopted to complete the detection of moving targets, which greatly improves the anti-interference ability of the system to complex backgrounds. At the same time, the detection of moving targets under foggy videos and non-foggy videos is compared, further verifying that de-fogging can further improve the effect of moving target detection.

[0051] The front-end system construction scheme with STM32 cooperating with the μC / OSIII real-time operating system is adopted. In terms of hardware, the design and construction of circuits including a display screen, a storage module, GPRS, and power supply-related circuits are completed. In terms of software, the transplantation and construction of the μC / OSIII system and task allocation and deployment are completed, the transplantation of the emWin graphics library and the design of the front-end interface are completed, and the establishment of a font library and file access operations are completed. In terms of tower inclination monitoring, by analyzing and comparing two inclination sensors, namely MPU6050 and SCA100T, the selection of the sensor and the acquisition of inclination data are completed. When the transmission line is damaged externally, the external damage information can be transmitted to the control center or the mobile device of the operator in time, which is convenient for the operator to process in time and ensures the safety of the transmission line. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is the overall design block diagram of the tower detection system of the present invention;

[0053] Figure 2 is the circuit connection diagram of the RGB display of the present invention;

[0054] Figure 3 is the circuit connection diagram of the SD card of the present invention;

[0055] Figure 4 It is the hierarchical structure diagram of the emWin graphics library of the present invention;

[0056] Figure 5 It is the schematic diagram of the storage operation of the present invention. Detailed implementation manners

[0057] The following further describes the present invention in detail with reference to the accompanying drawings. It should be noted that this is only for a clearer illustration and explanation of the present invention.

[0058] As Figures 1-5 shown, this embodiment discloses an intelligent perception method for preventing external damage information of transmission lines, which includes the following steps:

[0059] Step 1: Construct a weather preprocessing model;

[0060] According to the Nayar and Narasimhan scattering coefficients being independent of the wavelength of visible light in a homogeneous atmosphere, and the proposed simplified physical model for image restoration, the formula is:

[0061] I(x) = I ∞ ρ(x)e -βd(x) +I ∞ (1 - e -βd(x) )

[0062] where I ∞ is the brightness of the sky, which is the pixel value in the image. p(x) represents the normalized radiation degree of the scene point x, β represents the scattering coefficient of the atmosphere, and d(x) represents the distance between the actual scene corresponding to the image pixel x and the camera. On the right side of the equation, the first term represents the direct transmission model, and the second term represents the atmospheric light model. The equation also shows that the proportion of the direct transmission model will decrease with the increase of the distance, which is also the reason why some distant views in the image still look blurred in the case of no fog.

[0063] The construction of the weather preprocessing model includes constructing an atmospheric light scattering model, a fog area detection model, and a defogging model, and combining the atmospheric light scattering model, the fog area detection model, and the defogging model for perceiving external damage information of transmission lines in foggy weather;

[0064] The atmospheric light scattering model includes dark image enhancement, planetary image enhancement. Natural images, whether foggy or not, are composed of an incident component and a reflection component; the incident component represents the brightness information of the image, and the reflection component represents the internal information of the image. Its model is:

[0065] S(x, y) = L(x, y)·R(x, y)

[0066] Among them, S is the original foggy image, R is the reflection component, L is the incident component, and (x, y) is the image coordinate;

[0067] The Retinex-based defogging algorithm obtains R from S. Taking the logarithm of both sides of the equation gives:

[0068] logS(x, y) = logL(x, y) + logR(x, y)

[0069] For realizing dynamic range compression, restoring color or brightness, and in order to improve color fidelity, a single-scale Retinex (SSR) algorithm is constructed. The convolution operation of the Gaussian function and the foggy image is used to estimate the incident component. The formula is:

[0070] L(x, y) = S(x, y) * G(x, y)

[0071] Among them, * represents the convolution operation, and G(x, y) is the Gaussian function; in order to make up for the shortcomings of SSR, a multi-scale Retinex algorithm and a multi-scale Retinex algorithm with color restoration are set; the MSR algorithm uses multiple Gaussian functions with different scales in the RGB channels to calculate the output of the SSR algorithm and performs weighted summation on this basis.

[0072]

[0073] In the formula, ω j represents the weight of each Gaussian function, all of which are 1 / 3, and the image pixel values are set within the range of [0, 255]; the MSR algorithm has the advantages of small-scale dynamic distance compression and image enhancement. In order to improve the restoration of colors in the image, color restoration is used to control saturation.

[0074] The fog area detection model includes creating a semi-inverse image The semi-inverse image is obtained by replacing the RGB value of each pixel on each channel with the maximum value after the original image is inverted. The formula is:

[0075]

[0076] In the formula, r, g, and b are the three original channel values. In the fog-free area, due to the characteristic that at least one channel has a small value, the operation will replace this value with its reciprocal. In the sky or haze area, due to the characteristic that all values have high values, the max operation will return the same value; by comparing with the hue H of the HSV color space of the original image, the pixels to be restored are found, and the color appearance similar to the original image is maintained;

[0077] In order to identify the areas affected by fogginess, the original image I and the semi-inverse image I are calculated siThe difference between the H channels of the hue, and threshold it using a predefined value τ; the value of τ is used to select pixels with similar appearances in the initial and semi-inverted versions, and this value is affected by the haze degree of the haze image. The judgment formula is,

[0078]

[0079] In the formula, I H and respectively represent the H channels of I and I si Generated using τ = 10, and only pixels τ with a hue difference lower than this threshold are marked as blurred pixels, otherwise it is a clear image.

[0080] The defogging model includes edge-aware image defogging. Through the weighted guided filtering method, an edge-aware weight is introduced to replace the regularization parameter ε in the GIF. The edge-aware weight formula is,

[0081]

[0082] In the formula, p′ is the pixel at the center of the window, ζ is the radius of the window ω, p is the pixel around p′, is the variance in the window ω with p′ as the center pixel; ε takes the value of (0.001×L) 2 , where L is the dynamic range of the pixel values of the actual image; when an edge appears in the window with p′ as the center, That is, Γ G (p′) > 1, otherwise Γ G (p′) ≤ 1; used to achieve the perception of edges with different degrees of gradient changes.

[0083] Step two: Construct an anti-external damage information processing model;

[0084] When actually applied to an image, for the change rule of each pixel, a separate Gaussian distribution should be established, and this Gaussian distribution is updated in real time as time changes. However, due to the existence of a dynamic background, and the changes of these backgrounds follow certain rules, the mixture Gaussian background model can well suppress the interference brought by the dynamic background. The mixture Gaussian distribution applied to image pixels is also obtained by linearly adding K Gaussian distributions.

[0085] The anti-external damage information processing model includes moving target detection. Using a Gaussian distribution and linearly weighting it to represent the distribution rule of these pixels, the adaptive mixture Gaussian model is used to detect the target; the one-dimensional Gaussian distribution probability function is,

[0086]

[0087] In the formula, μ and σ 2respectively represent the mean and variance of the Gaussian distribution. The one-dimensional Gaussian distribution probability function is used to expand the joint density function of the multi-dimensional variable X = (x1, x2, … x n ):

[0088]

[0089] In the formula, d represents the dimension of the variable, u represents the mean of each dimension variable, Σ represents the covariance matrix, which describes the correlation degree of each dimension variable. For intuitive representation, a two-dimensional Gaussian distribution is taken as an example. Therefore, d = 2, and there are:

[0090]

[0091] The scatter points are distributed according to the single Gaussian probability function, and its Gaussian parameters are ideal. The data is estimated by the single Gaussian probability function. The measured effect is that the sample size in the central area is set to be larger than that in the edge, and the number of samples distributed in the central area is set to be less than that in the edge area;

[0092] The mixture Gaussian background model is adopted to suppress the interference caused by the dynamic background. The mixture Gaussian distribution of the image pixel points is obtained by linearly adding K Gaussian distributions. The formula is

[0093]

[0094] In the formula, K is the number of Gaussian distributions, and each Gaussian distribution represents a component; the larger the K value, the more components, and the stronger the adaptability of the model to complex backgrounds. The amount of computation is 3 - 5; w i,t represents the weight of the i-th Gaussian distribution component at time t, and u i,t represents the mean vector of the i-th Gaussian distribution component at time t, and I t (x, y) represents the pixel value of the image at the position (x, y) at time t.

[0095] The update steps of the Gaussian distribution model are

[0096] (1) Using the conditions of the above formula, the difference d i,t-1 between the pixel value I(x, y) and the mean u i of any Gaussian distribution component at the corresponding position is i compared with 2.5σ i ; if d i ≤ 2.5σ

[0097] (2) It is determined that the pixel value I(x, y) matches the component G i successfully, and the weight of G i is increased and (4) is executed;

[0098] (3) Determine the pixel value I(x, y) and the component G i Matching failed, lower G i If the weight becomes negative, delete the Gaussian distribution component and return to (1) until all distributions have completed the match with I(x, y) and execute (5);

[0099] (4) Update G according to the above formula i The mean, variance and weight of M p The distributions in are arranged in descending order according to the ratio of weight w to variance σ;

[0100] (5) If the number of successful matches of IP is 0, a new Gaussian distribution G is added. * , if the Gaussian distribution component of the current pixel does not reach the upper limit, it is added directly, otherwise the last component in Mp is replaced and rearranged in descending order;

[0101] (6) Count the sum of the weights of all Gaussian distribution components in Mp and normalize all weights;

[0102] (7) Using the above formula conditions, obtain B Gaussian distribution components as the background model of the next frame image at (x, y).

[0103] The detection of crane intrusion is mainly based on the characteristics of the crane's raised mechanical boom. When the height of the mechanical boom reaches a certain threshold, it is judged as an intrusion. Since the mechanical arm often presents a straight line state, the recognition process uses the fast and practical Hough line detection. However, Hough line detection operates on edge grayscale images. Therefore, it is necessary to extract the image edge through appropriate edge detection, obtain the image edge image, and then further identify the straight line features of the boom in the edge image.

[0104] Step 3: Build the tower monitoring hardware system.

[0105] In addition to video surveillance of transmission lines and the areas below them, status monitoring of pole towers is also an important part of power transmission security, especially in goaf areas or when pole towers are tilted and collapsed due to soil problems. In some places where the direction of power transmission needs to change, some corner pole towers are easily tilted because they are dragged by two transmission lines at a certain angle. Although there are daily line inspection personnel, it is often easy to overlook the slight tilt of the pole tower in the early stage. Therefore, high-precision sensors are used to monitor the status of the pole towers, and the data is uploaded to the cloud server in real time through wireless technology. When the status of the pole tower is abnormal, an alarm message is sent to the staff as soon as possible, which greatly improves the efficiency and safety reliability of pole tower monitoring and reduces the workload of front-line inspection personnel.

[0106] The tower monitoring hardware system includes a processor, an inclination detection module, an RGB capacitive touch screen, GPRS, a storage module, and a power supply module;

[0107] After being powered on, the system first loads the character library and emWin-related images, and then enters the parameter setting and display interface. The parameters can be set in two ways: through the mobile phone APP or the touch screen. After the parameter setting is completed, the measurement is started. The processor collects and processes the data related to the tower state and displays it on the RGB screen in real time. In terms of data storage, the data is generated into a file and stored in the front-end SD card, and at the same time, the data is transmitted to the mobile phone APP and the cloud server in real time.

[0108] The processor uses the low-power STM32F429 with an ARMCortex-M4 core as the main control MCU. The internal data path, registers, and memory interfaces are all set to 32 bits; an extended LCD-TFT controller and a variable storage controller are set;

[0109] In order to meet the power supply needs of both the sensor and the core processor at the same time, the power supply module uses the XL1509 voltage regulator chip to regulate the voltage to 5V for powering the sensor, and at the same time, the output 5V is regulated by the AMS1086 to generate 3.3V voltage for powering the core processor, SD card, and other peripheral modules;

[0110] The power supply voltage of the inclination sensor in the inclination detection module is 5V. When communicating with the processor, the judgment of the input high level is V IH > 4V. For the ARM processor with a power supply voltage of 3.3V, a bi-directional level conversion circuit is used to establish communication between the processor and the sensor.

[0111] The storage module; the storage module of the system includes three parts: an 8M serial Flash memory W25Q64 for storing various character library data displayed on the RGB screen, a 16M random access memory (SDRAM) W9812G2 for the display quick cache of the display screen, and a 2G SD memory card. Among them, the SD card is used to store the pictures and character libraries required for interface display on the one hand, and to store data on the other hand.

[0112] RGB monitor; To meet the needs of actual tower site commissioning and observation and facilitate the commissioning by front-line workers, the front-end embedded system is connected to the human-machine interaction interface. The touch screen consists of two parts: the ATO70TN92 liquid crystal display screen and the capacitive touch screen. And the capacitive touch screen needs to be used in conjunction with the GT911 touch chip. The display screen is a 7-inch liquid crystal screen with a resolution of 800*480, and the GT911 has 26 drive channels and 14 sensing channels, with high precision and processing speed, fully meeting the system requirements. In terms of the processor, the LCD-TFT controller (LTCD) integrated inside the STM32F4 can control the 24-bit RGB channels of the liquid crystal display screen in parallel and supports 8 pixel formats such as ARGB8888, RGB565, and ARGB1555 and multi-layer display. The development of the system is completed through the ARM development tool MDK-ARM integrated development environment (IDE). The IDE integrates functions such as software program writing, compilation, and debugging, and can display the variables involved in the system in the form of an oscilloscope, bringing great convenience to the debugging of sensors.

[0113] The above is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and its concept of the present invention, making equivalent substitutions or changes should be covered by the protection scope of the present invention.

Claims

1. An intelligent perception method for preventing external damage information of transmission lines, characterized in that, It includes the following steps: Step 1, construct a weather preprocessing model; Step 2, construct an anti-external damage information processing model; Step 3, build a tower monitoring hardware system.

2. The intelligent perception method for preventing external damage information of transmission lines according to claim 1, wherein: The construction of the weather preprocessing model includes constructing an atmospheric light scattering model, a fog area detection model and a defogging model, and combining the atmospheric light scattering model, the fog area detection model and the defogging model to be used for the perception of external damage information of the transmission line in foggy weather; The atmospheric light scattering model includes dark image enhancement, planetary image enhancement, and natural images. Whether there is fog or not, natural images are composed of an incident component and a reflection component; the incident component represents the brightness information of the image, and the reflection component represents the internal information of the image. Its model is: S(x,y) = L(x,y)·R(x,y) Where S is the original foggy image, R is the reflection component, L is the incident component, and (x,y) is the image coordinate; Based on the Retinex defogging algorithm, R is obtained from S. Taking the logarithm of both sides of the equation gives: log S(x,y) = logL(x,y) + logR(x,y) It is used to achieve dynamic range compression, color or brightness restoration. In order to improve color fidelity, a single-scale Retinex (SSR) algorithm is constructed, and the convolution operation of the Gaussian function and the foggy image is used to estimate the incident component. The formula is: L(x,y) = S(x,y)*G(x,y) Where * represents the convolution operation, and G(x,y) is the Gaussian function; in order to make up for the shortcomings of SSR, a multi-scale Retinex algorithm and a multi-scale Retinex algorithm with color restoration are set; the MSR algorithm uses multiple Gaussian functions with different scales in the RGB channels to calculate the output of the SSR algorithm and performs weighted summation on this basis. where ω j represents the weight of each Gaussian function, all of which are 1 / 3, and the image pixel values are set within the range of [0, 255]; the MSR algorithm has the advantages of small-scale dynamic distance compression and image enhancement. To improve the restoration of colors in the image, color restoration is used to control saturation.

3. The intelligent perception method for preventing external damage information of a transmission line according to claim 2, wherein: The fog area detection model includes creating a semi-inverse image The semi-inverse image is obtained by replacing the RGB value of each pixel on each channel with the maximum value after the original image is inverted, and the formula is: In the formula, r, g, and b are the values of the three original channels respectively. In the fog-free area, since at least one channel has the characteristic of a small value, the operation will replace this value with its reciprocal. In the sky or haze area, since all values have the characteristic of a high value, the max operation will return the same value; by comparing with the hue H of the original image in the HSV color space, the pixels to be restored are found, and the color appearance similar to the original image is maintained; To identify the regions affected by haze, the difference between the hue H channels of the original image I and the semi-inverse image I si is calculated and thresholded using a predefined value τ; the value of τ is used to select pixels with a similar appearance in the initial and semi-inverted versions, and the formula for judging the influence of the haze degree of the haze image is where, I H and respectively represent the H channels of I and I si generated using τ = 10, and only pixels τ with a hue difference lower than this threshold are marked as blurred pixels, otherwise it is a clear image.

4. The intelligent perception method for preventing external damage information of transmission lines according to claim 2, characterized in that: The defogging model includes edge-aware image defogging. By the weighted guided filtering method, the edge-aware weight is introduced to replace the regularization parameter ε in the GIF. The edge-aware weight formula is Wherein, p′ is the pixel at the center of the window, ζ is the radius of the window ω, and p is the pixel around p′. is the variance in the window ω with p′ as the central pixel; ε takes a value of (0.001 × L) 2 , where L is the dynamic range of the pixel values of the actual image; when an edge appears in the window centered on p′, i.e., Γ G (p′) > 1, otherwise Γ G (p′) ≤ 1; it is used to realize the perception of edges with different degrees of gradient changes.

5. The intelligent perception method for preventing external damage information of transmission lines according to claim 1, wherein: The anti-external damage information processing model includes moving target detection. The Gaussian distribution is adopted and linearly weighted to represent the distribution law of these pixels. The adaptive mixture Gaussian model is used to detect the target; the one-dimensional Gaussian distribution probability function is where μ and σ 2 represent the mean and variance of the Gaussian distribution respectively. The one-dimensional Gaussian distribution probability function is used to expand the joint density function of the multi-dimensional variable X = (x1, x2, … x n ): In the formula, d represents the dimension of the variable, u represents the mean of each dimension variable, and Σ represents the covariance matrix, which describes the correlation degree of each dimension variable. For intuitive representation, a two-dimensional Gaussian distribution is taken as an example. Therefore, d = 2, and there is: The scatter points are distributed according to the single Gaussian probability function, and its Gaussian parameters are ideal. The data uses the single Gaussian probability function for parameter estimation. The measured effect is that the sample size of the scatter points in the central area is set to be greater than that at the edge, and the number of samples distributed in the central area is set to be less than that in the edge area; The Gaussian mixture background model is adopted to suppress the interference brought by the dynamic background. The Gaussian mixture distribution of image pixels is obtained by linearly adding K Gaussian distributions. The formula is Where K is the number of Gaussian distributions, and each Gaussian distribution represents a component; the larger the value of K, the more components, and the stronger the model's ability to adapt to complex backgrounds. The computational workload is 3-5; w i,t represents the weight of the i-th Gaussian distribution component at time t, and u i,t represents the mean vector of the i-th Gaussian distribution component at time t, and I t (x, y) represents the pixel value of the image at the position (x, y) at time t.

6. The intelligent perception method for preventing external damage information of transmission lines according to claim 5, characterized in that: The update steps of the Gaussian distribution model are (1) Using the above formula, the pixel value I(x, y) and the mean value u of any Gaussian distribution component at the corresponding position i,t-1 The difference d i Then with 2.5σ i Make a comparison; if d i ≤2.5σ i Then execute (2), otherwise execute (3); (2) Determine that the pixel value I(x, y) matches component G i Successfully matched, increase the weight of G i and execute (4); (3) It is determined that the pixel value I(x, y) does not match the component G i The matching fails, and the weight value of G i is decreased. If the weight value becomes negative, the Gaussian distribution component is deleted, and (1) is returned and executed until all distributions have completed the matching with I(x, y) and (5) is executed; (4) Update G according to the above formula i for the mean, variance, and weights, and arrange the distributions in M p in descending order according to the ratio of the weight w to the variance σ; (5) If the number of successful matches of Ip is 0 at this time, a new Gaussian distribution G is added. * If the Gaussian distribution components of the current pixel do not reach the upper limit, they are directly added. Otherwise, the last component in the sorted Mp is replaced, and the components are sorted in descending order again. (6) Statistically sum the weights of all Gaussian distribution components in Mp and normalize all weights; (7) Use the above formula conditions to obtain B Gaussian distribution components as the background model of the next frame of image at (x, y).

7. The intelligent perception method for preventing external damage information of transmission lines according to claim 1, characterized in that: The tower monitoring hardware system includes a processor, an inclination detection module, an RGB capacitive touch screen, GPRS, a storage module, and a power supply module; After being powered on, the system first loads the character library and emWin-related images, and then enters the parameter setting display interface. The parameters can be set in two ways: through the mobile phone APP or the touch screen. After the parameter setting is completed, the measurement is started. The processor collects and processes the data related to the tower state and displays it on the RGB screen in real time. In terms of data storage, the data is generated into a file and stored in the front-end SD card, and at the same time, the data is transmitted to the mobile phone APP and the cloud server in real time.

8. The intelligent perception method for preventing external damage information of transmission lines according to claim 7, wherein: The processor uses the low-power STM32F429 with an ARMCortex-M4 core as the main control MCU. The internal data path, registers, and memory interfaces are all set to 32 bits; an extended LCD-TFT controller and a variable storage controller are set; In order to meet the power supply needs of the sensor and the core processor at the same time, the power supply module uses the XL1509 voltage regulator chip to regulate the voltage to 5V to supply power to the sensor, and at the same time, the output 5V is regulated by AMS1086 to generate a 3.3V voltage to supply power to the core processor, SD card, and other peripheral modules; The power supply voltage of the inclination sensor of the inclination detection module is 5V. When communicating with the processor, the judgment of the input high level is V IH > 4V. For the ARM processor with a power supply voltage of 3.3V, a bidirectional level conversion circuit is adopted to establish communication between the processor and the sensor.

Images