Artificial intelligence-based cable defect image recognition system and method

Abstract

This invention discloses an intelligent image recognition system and method for cable defects based on artificial intelligence, belonging to the field of image recognition technology. The invention analyzes cable images captured by a high-definition camera using sensors mounted on a robot to collect data on the thickness of the cable overlay. It calculates and judges the cable sag angle, marks the target cable, generates a predicted curve for the overlay thickness based on meteorological data, calculates a predicted curve for the deformation degree of the target cable based on the predicted curve, classifies risk levels and sets warning coefficients based on the deformation rate extracted from the predicted curve, and calculates the priority of the robot for cable defect detection based on the warning coefficient and the deformation degree, driving the robot to plan the detection path.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, specifically to an artificial intelligence-based image recognition system and method for cable defects. Background Technology

[0002] Currently, defect detection of tunnel cables primarily relies on regular manual inspections. Under normal operating conditions, manual inspections can detect some obvious cable defects. However, when cables are exposed to the dark environment of tunnels for extended periods, especially in extremely cold winter weather, a layer of deposits may form on the cable surface near tunnel leaks. The pressure from this deposits can cause tensile deformation or even "bow-shaped" sagging, resulting in micro-cracks on the cable surface. Traditional manual inspections are not only inefficient in detecting these cracks, but also susceptible to interference from environmental factors, leading to missed defects. Furthermore, manual inspections lack real-time monitoring; some cables break due to excessive deformation caused by untimely inspections, seriously threatening the safe and stable operation of high-voltage transmission lines. Summary of the Invention

[0003] The purpose of this invention is to provide an artificial intelligence-based cable defect image recognition system and method to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an artificial intelligence-based method for cable defect image recognition, the method comprising:

[0005] Step S100: Acquire cable images; obtain the thickness of the covering layer on the cable surface; analyze the cable images based on the acquired covering layer thickness, calculate the cable sag angle, and capture the target cable based on the cable sag angle.

[0006] Step S200: Obtain the predicted temperature and humidity values, and generate a predicted curve for the cover layer thickness. Based on the predicted cover layer thickness values ​​corresponding to each time point in the generated predicted curve, calculate the degree of deformation of the target cable at each time point, and generate a predicted curve for the degree of deformation of the target cable.

[0007] Step S300: Based on the deformation degree prediction curve of the target cable, calculate the deformation development rate of the target cable, classify the risk level of the target cable, and calculate the early warning coefficient corresponding to each risk level.

[0008] Step S400: Calculate the priority of defect detection for the target cable based on the warning coefficient of the risk level corresponding to the target cable and the current degree of deformation of the target cable. The robot automatically plans the detection path according to the calculated priority.

[0009] Furthermore, step S100 includes:

[0010] Step S101: Extract the collected cable images at and below zero degrees Celsius, identify the cable contour in the cable image using image recognition technology, select the cable along the cable contour using a detection box, and establish a coordinate system in the cable image with the lowest point of the detection box as the origin.

[0011] Step S102: Extract the thickness d of the cable surface covering layer collected by the sensor on the robot. In the coordinate system, translate the lower edge of the detection frame upward by d units and translate the upper edge of the detection frame downward by d units. The upper and lower edge lines after translation by d units form the target detection frame. Read the coordinates of the lowest point (x0, y0), the left endpoint (x1, y1), and the right endpoint (x2, y2) of the target detection frame from the coordinate system.

[0012] Step S103: Based on the read coordinates of the lowest point (x0, y0), the left endpoint (x1, y1), and the right endpoint (x2, y2), calculate the cable sag angle using the following formula:

[0013] ;

[0014] in, Indicates the sag angle of the cable;

[0015] Step S104: Identify the cable model in the target detection box using image recognition technology. Retrieve the sag angle threshold of the cable corresponding to the cable model from the database. Compare the calculated sag angle of the cable with the sag angle threshold. If the sag angle is greater than the sag angle threshold, mark the cable as the target cable. Obtain the geographical location of the target cable and store the cable model, geographical location, and sag angle of the target cable.

[0016] The above steps involve using image recognition technology to identify the cable outline and establish a coordinate system. The range of the detection frame is adjusted according to the thickness of the overlay collected by the sensor to eliminate interference. The overlay includes icicles that may form at tunnel leaks when the temperature is extremely low, causing the cable surface to freeze. It may also contain dust, dirt, oil, and chemical liquids. When the overlay reaches a certain thickness, it can cause the cable to bend and sag.

[0017] Furthermore, step S200 includes:

[0018] Step S201: Extract the geographical location of the target cable, retrieve the temperature and humidity forecasts for the geographical location from the meteorological department, sort the temperature and humidity forecasts according to time series, mark the time periods in which the temperature and humidity forecasts remain unchanged, and divide the time into several time periods. The predicted thickness of the cover layer of the target cable for a certain time period is calculated using the following formula:

[0019] ;

[0020] Wherein, D represents the predicted value of the cover layer thickness of the target cable corresponding to the time period, d represents the cover layer thickness of the target cable at the current time point, k1 represents the growth coefficient, T represents the predicted value of the temperature during the time period, β represents the humidity influence factor, and H represents the predicted value of the humidity during the time period.

[0021] Step S202: Sort the calculated predicted values ​​of the target cable's sheath thickness according to the time series to generate a sheath thickness prediction curve; extract the sheath thickness corresponding to a certain time point from the sheath thickness prediction curve, and calculate the degree of deformation of the target cable at that time point using the following formula:

[0022] ;

[0023] Wherein, G represents the degree of deformation of the target cable at the time point, and D represents the thickness of the covering layer on the surface of the target cable at the time point. y1 is the x-coordinate of the left endpoint of the target detection box, y2 is the x-coordinate of the right endpoint of the target detection box, and y3 is the y-coordinate of the right endpoint of the target detection box.

[0024] Step S203: Obtain the calculated deformation degree of the target cable and the corresponding time point, and plot the deformation degree prediction curve of the target cable;

[0025] In the above steps, when the cable deformation reaches a certain level, cracks may appear on the cable surface. Therefore, by comprehensively analyzing the impact of environmental factors on the growth of the covering layer, the accuracy of the prediction can be improved. The generated prediction curve can intuitively show the trend of cable change over time, and realize early warning of risks.

[0026] Furthermore, step S300 includes:

[0027] Step S301: Extract the deformation degrees G1 and G2 of the target cable corresponding to two adjacent time points t1 and t2 from the deformation degree prediction curve of the target cable, and calculate the deformation development rate of the target cable. The calculation formula is as follows:

[0028] ;

[0029] Where v represents the deformation development rate of the target cable at two adjacent time points;

[0030] Step S302: Calculate the weighted average of the calculated deformation development rates of the target cable to obtain the average deformation development rate of the target cable; extract the average deformation development rates of all target cables, sort the average deformation development rates in ascending order to obtain an average deformation development rate sequence; and obtain the average value of the average deformation development rate sequence according to the preset average value calculation formula and standard deviation calculation formula. Sum and standard deviation α; set the first risk threshold as Set the second risk threshold as ;

[0031] Step S303: Classify the target cable with average deformation development rate V≤V1 into the first risk level, classify the target cable with average deformation development rate V1<V<V2 into the second risk level, and classify the target cable with average deformation development rate V≥V2 into the third risk level; set the early warning coefficient corresponding to the first risk level as Q1, set the early warning coefficient corresponding to the second risk level as Q2, set the early warning coefficient corresponding to the third risk level as Q3, and Q1<Q2<Q3;

[0032] In the above steps, due to the water leakage points in the tunnel, a covering layer appears on the surface of some cables. However, the thickness of the covering layer on the cable surface varies depending on the location of the water leakage point, resulting in different deformation development rates of the cable. By calculating the average value of the deformation development rate of the target cable, the interference of data fluctuations is eliminated, and dynamic adaptive grading is achieved, avoiding the limitations of a single fixed threshold.

[0033] Furthermore, step S400 includes:

[0034] Step S401: Calculate the priority of the target cable defect detection according to the preset priority calculation formula. The calculation formula is as follows:

[0035] ;

[0036] Where P represents the priority of the target cable, Q i G represents the warning coefficient corresponding to the i-th risk level of the target cable, and G represents the degree of deformation of the current target cable.

[0037] Step S402: Sort all the calculated priority of the target cables in descending order, and at the same time, the robot extracts the geographical location of the target cables according to the priority order, and performs defect detection on the target cables according to the geographical location of each target cable.

[0038] Furthermore, to better implement the above method, an artificial intelligence-based cable defect image recognition system is also provided. The system includes: an acquisition module, a prediction module, an evaluation module, and an operation module.

[0039] The acquisition module is used to acquire cable images and overlay thickness data, analyze the acquired cable images based on the overlay thickness data, calculate the cable sag angle, and mark the target cable based on the calculated sag angle.

[0040] The prediction module obtains temperature and humidity prediction values ​​from the meteorological department, generates a cover layer thickness prediction curve, calculates the degree of deformation of the target cable at each time point based on the cover layer thickness prediction value corresponding to each time point in the generated cover layer thickness prediction curve, and generates a target cable deformation degree prediction curve.

[0041] The assessment module calculates the deformation development rate of the target cable based on the deformation degree prediction curve of the target cable, classifies the risk level of the target cable, and calculates the early warning coefficient corresponding to each risk level.

[0042] The task module calculates the priority of defect detection for the target cable based on the warning coefficient of the risk level corresponding to the target cable and the current degree of deformation of the target cable. The robot automatically plans the detection path according to the calculated priority.

[0043] Furthermore, the acquisition module includes: an image acquisition unit, an overlay thickness acquisition unit, a coordinate establishment unit, and a target marking unit;

[0044] The image acquisition unit uses a high-definition camera to capture images of cables below zero degrees Celsius in winter.

[0045] The cover layer thickness acquisition unit obtains the cover layer thickness on the cable surface through sensors;

[0046] The coordinate establishment unit extracts the collected cable images below zero degrees Celsius in winter, identifies the cable outline in the cable image using image recognition technology, selects the cable along the cable outline using a detection box, and establishes a coordinate system in the cable image with the lowest point of the detection box as the origin.

[0047] The target marking unit extracts the coating thickness of the cable surface collected by sensors mounted on the robot. It translates the detection frame in the coordinate system according to the coating thickness to generate a target detection frame. It reads the coordinates of the lowest point, left endpoint, and right endpoint of the target detection frame from the coordinate system and calculates the cable sag angle. Using image recognition technology, it identifies the cable model in the target detection frame. Based on the cable model, it retrieves the corresponding cable sag angle threshold from the database. It compares the calculated cable sag angle with the sag angle threshold. If the sag angle is greater than the threshold, the cable is marked as a target cable. The unit obtains the geographical location of the target cable and stores the cable model, geographical location, and sag angle of the target cable.

[0048] Furthermore, the prediction module includes: a meteorological processing unit, a cover layer prediction unit, and a deformation prediction unit;

[0049] The meteorological processing unit extracts the geographical location of the target cable, retrieves the temperature and humidity forecasts for the geographical location from the meteorological department, and sorts the temperature and humidity forecasts according to the time series.

[0050] The cover layer prediction unit marks the time period in which both the temperature prediction value and the humidity prediction value remain unchanged, divides it into several time periods, and calculates the predicted value of the cover layer thickness of the target cable for each time period.

[0051] The deformation prediction unit sorts the calculated predicted values ​​of the target cable's cover layer thickness according to the time series and generates a cover layer thickness prediction curve; it extracts the cover layer thickness corresponding to each time point from the cover layer thickness prediction curve, calculates the degree of deformation of the target cable at the time point, and plots a deformation degree prediction curve of the target cable.

[0052] Furthermore, the evaluation module includes: a rate calculation unit, a risk classification unit, and a coefficient setting unit;

[0053] The rate calculation unit extracts the deformation degree of the target cable corresponding to two adjacent time points from the deformation degree prediction curve of the target cable, and calculates the deformation development rate of the target cable.

[0054] The risk classification unit calculates the weighted average of the deformation development rates of the target cables to obtain the average deformation development rate of the target cables; it extracts the average deformation development rates of all target cables, sorts the average deformation development rates in ascending order to obtain an average deformation development rate sequence, and obtains the average value and standard deviation of the average deformation development rate sequence according to the preset average value calculation formula and standard deviation calculation formula; and it sets a first risk threshold and a second risk threshold.

[0055] The coefficient setting unit classifies target cables whose average deformation development rate is less than or equal to a first risk threshold into a first risk level, target cables whose average deformation development rate is greater than the first risk threshold but less than a second risk threshold into a second risk level, and target cables whose average deformation development rate is greater than or equal to the second risk threshold into a third risk level; and sets the warning coefficient corresponding to each risk level.

[0056] Furthermore, the task module includes: a priority calculation unit and a path planning unit;

[0057] The priority calculation unit calculates the priority of the target cable defect detection according to the preset priority calculation formula;

[0058] The path planning unit sorts the calculated priorities of all target cables in descending order. At the same time, the robot extracts the geographical locations of the target cables according to the priority order and performs defect detection on the target cables based on their geographical locations.

[0059] Compared with the prior art, the beneficial effects of the present invention are:

[0060] 1. Automatic Defect Detection: The robot automatically collects data using high-definition cameras and sensors, and combines image recognition and algorithm analysis to automatically identify cable defects. This avoids the problems of missed detection and low real-time performance that occur during manual inspections, and significantly improves the efficiency of cable defect detection.

[0061] 2. Dynamic prediction and risk assessment: Based on the generated prediction curves of the cover layer thickness and the degree of deformation, the deformation development rate is calculated and the risk level is divided according to the deformation development rate, realizing dynamic monitoring and graded early warning of cable defects.

[0062] 3. Intelligent path planning: Based on the calculated priority, the robot is guided to automatically plan the cable defect detection path, which optimizes the allocation of resources, performs targeted cable defect detection, and ensures the safe transmission of electricity. Attached Figure Description

[0063] Figure 1This is a schematic diagram of the process flow of the cable defect image recognition system and method based on artificial intelligence according to the present invention;

[0064] Figure 2 This is a schematic diagram of the system structure of the cable defect image recognition system and method based on artificial intelligence of the present invention. Detailed Implementation

[0065] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Example 1: As Figure 1 As shown, the present invention provides a technical solution: a cable defect image recognition method based on artificial intelligence, the method comprising:

[0067] Step S100: Acquire cable images; obtain the thickness of the covering layer on the cable surface; analyze the cable images based on the acquired covering layer thickness, calculate the cable sag angle, and capture the target cable based on the cable sag angle.

[0068] Step S100 includes:

[0069] Step S101: Extract the collected cable images at and below zero degrees Celsius, identify the cable contour in the cable image using image recognition technology, select the cable along the cable contour using a detection box, and establish a coordinate system in the cable image with the lowest point of the detection box as the origin.

[0070] Step S102: Extract the thickness d of the cable surface covering layer collected by the sensor on the robot. In the coordinate system, translate the lower edge of the detection frame upward by d units and translate the upper edge of the detection frame downward by d units. The upper and lower edge lines after translation by d units form the target detection frame. Read the coordinates of the lowest point (x0, y0), the left endpoint (x1, y1), and the right endpoint (x2, y2) of the target detection frame from the coordinate system.

[0071] Step S103: Based on the read coordinates of the lowest point (x0, y0), the left endpoint (x1, y1), and the right endpoint (x2, y2), calculate the cable sag angle using the following formula:

[0072] ;

[0073] in, Indicates the sag angle of the cable;

[0074] Step S104: Identify the cable model in the target detection box using image recognition technology. Retrieve the sag angle threshold of the cable corresponding to the cable model from the database. Compare the calculated sag angle of the cable with the sag angle threshold. If the sag angle is greater than the sag angle threshold, mark the cable as the target cable. Obtain the geographical location of the target cable and store the cable model, geographical location, and sag angle of the target cable.

[0075] In an embodiment of the invention, a high-definition camera mounted on a robot is used to capture real-time images of cables at and below zero degrees Celsius, and to draw the cable outline including the ice layer. Sensors are used to collect the thickness of the covering layer on the cable surface in real time (e.g., if the thickness is measured to be d=5mm). Image recognition technology is used to generate a detection box, and a coordinate system is established by setting the lowest point of the detection box as the origin. A target detection box is generated based on the covering layer thickness, and the coordinates of the lowest point, left endpoint, and right endpoint of the target detection box are read. The cable sag angle is calculated based on the acquired coordinates. The droop angle threshold retrieved from the database is 10°. The cable is marked as the target cable, and the geographical location of the target cable (e.g., longitude 120.1°, latitude 30.2°) is recorded.

[0076] Step S200: Obtain the predicted temperature and humidity values, and generate a predicted curve for the cover layer thickness. Based on the predicted cover layer thickness values ​​corresponding to each time point in the generated predicted curve, calculate the degree of deformation of the target cable at each time point, and generate a predicted curve for the degree of deformation of the target cable.

[0077] Step S200 includes:

[0078] Step S201: Extract the geographical location of the target cable, retrieve the temperature and humidity forecasts for the geographical location from the meteorological department, sort the temperature and humidity forecasts according to time series, mark the time periods in which the temperature and humidity forecasts remain unchanged, and divide the time into several time periods. The predicted thickness of the cover layer of the target cable for a certain time period is calculated using the following formula:

[0079] ;

[0080] Wherein, D represents the predicted value of the cover layer thickness of the target cable corresponding to the time period, d represents the cover layer thickness of the target cable at the current time point, k1 represents the growth coefficient, T represents the predicted value of the temperature during the time period, β represents the humidity influence factor, and H represents the predicted value of the humidity during the time period.

[0081] Step S202: Sort the calculated predicted values ​​of the target cable's sheath thickness according to the time series to generate a sheath thickness prediction curve; extract the sheath thickness corresponding to a certain time point from the sheath thickness prediction curve, and calculate the degree of deformation of the target cable at that time point using the following formula:

[0082] ;

[0083] Wherein, G represents the degree of deformation of the target cable at the time point, and D represents the thickness of the covering layer on the surface of the target cable at the time point. y1 is the x-coordinate of the left endpoint of the target detection box, y2 is the x-coordinate of the right endpoint of the target detection box, and y3 is the y-coordinate of the right endpoint of the target detection box.

[0084] Step S203: Obtain the calculated deformation degree of the target cable and the corresponding time point, and plot the deformation degree prediction curve of the target cable;

[0085] In embodiments of the present invention, the geographical location of the target cable is extracted, and the temperature and humidity forecasts for the next 7 days are obtained from the meteorological department. The data is then divided into several time periods, taking one time period as an example (e.g., ...). (T=-3℃, H=70%, k1=0.02, β=0.1), the predicted value of the covering layer thickness is calculated as D=5+0.02×|-3|×(1+0.1×70%)×12=5.935mm, and a covering layer thickness prediction curve is generated. The predicted value for the 24th hour is extracted from the covering layer thickness prediction curve as 6.5mm. The deformation degree G of the target cable is calculated using the formula (assuming a proportionality coefficient). (Unit conversion factor C=1), when D=6.6mm, G=32.5mm, and a deformation degree prediction curve is generated;

[0086] Step S300: Based on the deformation degree prediction curve of the target cable, calculate the deformation development rate of the target cable, classify the risk level of the target cable, and calculate the early warning coefficient corresponding to each risk level.

[0087] Step S300 includes:

[0088] Step S301: Extract the deformation degrees G1 and G2 of the target cable corresponding to two adjacent time points t1 and t2 from the deformation degree prediction curve of the target cable, and calculate the deformation development rate of the target cable. The calculation formula is as follows:

[0089] ;

[0090] Where v represents the deformation development rate of the target cable at two adjacent time points;

[0091] In an embodiment of the present invention, the deformation degree of the target cable at adjacent time points (t1=0h, t2=12h) (G1=5mm, G2=5.93mm) is extracted, and the deformation development rate is calculated as v=(5.93-5) / (12-0)=0.0775mm / h;

[0092] Step S302: Calculate the weighted average of the calculated deformation development rates of the target cable to obtain the average deformation development rate of the target cable; extract the average deformation development rates of all target cables, sort the average deformation development rates in ascending order to obtain an average deformation development rate sequence; and obtain the average value of the average deformation development rate sequence according to the preset average value calculation formula and standard deviation calculation formula. Sum and standard deviation α; set the first risk threshold as Set the second risk threshold as ;

[0093] Step S303: Classify the target cable with average deformation development rate V≤V1 into the first risk level, classify the target cable with average deformation development rate V1<V<V2 into the second risk level, and classify the target cable with average deformation development rate V≥V2 into the third risk level; set the early warning coefficient corresponding to the first risk level as Q1, set the early warning coefficient corresponding to the second risk level as Q2, set the early warning coefficient corresponding to the third risk level as Q3, and Q1<Q2<Q3;

[0094] Step S400: Calculate the priority of defect detection for the target cable based on the warning coefficient of the risk level corresponding to the target cable and the current degree of deformation of the target cable. The robot automatically plans the detection path according to the calculated priority.

[0095] Step S400 includes:

[0096] Step S401: Calculate the priority of the target cable defect detection according to the preset priority calculation formula. The calculation formula is as follows:

[0097] ;

[0098] Where P represents the priority of the target cable, Q i G represents the warning coefficient corresponding to the i-th risk level of the target cable, and G represents the degree of deformation of the current target cable.

[0099] In an embodiment of the present invention, the warning coefficient of a certain target cable is Q3=0.9, and the priority calculated according to the formula is P=0.9×32.5=29.25;

[0100] Step S402: Sort all the calculated priority of the target cables in descending order, and at the same time, the robot extracts the geographical location of the target cables according to the priority order, and performs defect detection on the target cables according to the geographical location of each target cable.

[0101] Example 2: Figure 2 As shown, in order to better implement the above method, an artificial intelligence-based cable defect image recognition system is also provided. The system includes: an acquisition module, a prediction module, an evaluation module, and an operation module.

[0102] The acquisition module is used to acquire cable images and ice thickness data, analyze the acquired cable images based on the overburden thickness data, calculate the cable sag angle, and mark the target cable based on the calculated sag angle.

[0103] The prediction module obtains temperature and humidity prediction values ​​from the meteorological department, generates a cover layer thickness prediction curve, calculates the degree of deformation of the target cable at each time point based on the cover layer thickness prediction value corresponding to each time point in the generated cover layer thickness prediction curve, and generates a target cable deformation degree prediction curve.

[0104] The assessment module calculates the deformation development rate of the target cable based on the deformation degree prediction curve of the target cable, classifies the risk level of the target cable, and calculates the early warning coefficient corresponding to each risk level.

[0105] The operation module calculates the priority of defect detection for the target cable based on the warning coefficient of the risk level corresponding to the target cable and the current degree of deformation of the target cable. The robot automatically plans the detection path according to the calculated priority.

[0106] The acquisition module includes: an image acquisition unit, an overlay thickness acquisition unit, a coordinate establishment unit, and a target marking unit;

[0107] The image acquisition unit uses a high-definition camera to capture images of cables below zero degrees Celsius in winter.

[0108] The cover layer thickness acquisition unit obtains the cover layer thickness on the cable surface through sensors;

[0109] The coordinate establishment unit extracts the collected cable images below zero degrees Celsius in winter, identifies the cable outline in the cable image using image recognition technology, selects the cable along the cable outline using a detection box, and establishes a coordinate system in the cable image with the lowest point of the detection box as the origin.

[0110] The target marking unit extracts the coating thickness of the cable surface collected by sensors mounted on the robot. It translates the detection frame in the coordinate system according to the coating thickness to generate a target detection frame. It reads the coordinates of the lowest point, left endpoint, and right endpoint of the target detection frame from the coordinate system and calculates the cable sag angle. It then identifies the cable model in the target detection frame using image recognition technology. Based on the cable model, it retrieves the corresponding cable sag angle threshold from the database. It compares the calculated cable sag angle with the sag angle threshold. If the sag angle is greater than the threshold, the cable is marked as a target cable. The unit obtains the geographical location of the target cable and stores the cable model, geographical location, and sag angle of the target cable.

[0111] The prediction module includes: a meteorological processing unit, a cover layer prediction unit, and a deformation prediction unit.

[0112] The meteorological processing unit extracts the geographical location of the target cable, retrieves the temperature and humidity forecasts for the geographical location from the meteorological department, and sorts the temperature and humidity forecasts according to the time series.

[0113] The cover layer prediction unit marks the time period in which both the temperature prediction value and the humidity prediction value remain unchanged, divides it into several time periods, and calculates the predicted value of the cover layer thickness of the target cable for each time period.

[0114] The deformation prediction unit sorts the calculated predicted values ​​of the cover layer thickness of the target cable according to the time series and generates a cover layer thickness prediction curve; it extracts the cover layer thickness corresponding to each time point from the cover layer thickness prediction curve, calculates the degree of deformation of the target cable at the time point, and plots the degree of deformation prediction curve of the target cable.

[0115] The evaluation module includes: a rate calculation unit, a risk classification unit, and a coefficient setting unit.

[0116] The rate calculation unit extracts the deformation variables of the target cable corresponding to two adjacent time points from the deformation degree prediction curve of the target cable, and calculates the deformation development rate of the target cable.

[0117] The risk classification unit calculates the weighted average of the deformation development rates of the target cables to obtain the average deformation development rate of the target cables; it extracts the average deformation development rates of all target cables, sorts the average deformation development rates in ascending order to obtain an average deformation development rate sequence, and obtains the average value and standard deviation of the average deformation development rate sequence according to the preset average value calculation formula and standard deviation calculation formula; and it sets a first risk threshold and a second risk threshold.

[0118] The coefficient setting unit classifies target cables whose average deformation development rate is less than or equal to the first risk threshold into the first risk level, target cables whose average deformation development rate is greater than the first risk threshold but less than the second risk threshold into the second risk level, and target cables whose average deformation development rate is greater than or equal to the second risk threshold into the third risk level; and sets the early warning coefficient corresponding to each risk level.

[0119] The task module includes: a priority calculation unit and a path planning unit;

[0120] The priority calculation unit calculates the priority of the target cable defect detection according to the preset priority calculation formula;

[0121] The path planning unit sorts the calculated priorities of all target cables in descending order. At the same time, the robot extracts the geographical locations of the target cables according to the priority order and performs defect detection on the target cables based on their geographical locations.

[0122] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A cable defect image recognition method based on artificial intelligence, characterized in that, The method includes: Step S100: Acquire cable images; obtain the thickness of the covering layer on the cable surface; analyze the cable images based on the acquired covering layer thickness, calculate the cable sag angle, and capture the target cable based on the cable sag angle. Step S200: Obtain the predicted temperature and humidity values, and generate a predicted curve for the cover layer thickness. Based on the predicted cover layer thickness values ​​corresponding to each time point in the generated predicted curve, calculate the degree of deformation of the target cable at each time point, and generate a predicted curve for the degree of deformation of the target cable. Step S300: Based on the deformation degree prediction curve of the target cable, calculate the deformation development rate of the target cable, classify the risk level of the target cable, and calculate the early warning coefficient corresponding to each risk level. Step S400: Calculate the priority of defect detection for the target cable based on the warning coefficient of the risk level corresponding to the target cable and the current degree of deformation of the target cable. The robot automatically plans the detection path according to the calculated priority.

2. The artificial intelligence-based cable defect image recognition method according to claim 1, characterized in that, Step S100 includes: Step S101: Extract the collected cable images at and below zero degrees Celsius, identify the cable contour in the cable image using image recognition technology, select the cable along the cable contour using a detection box, and establish a coordinate system in the cable image with the lowest point of the detection box as the origin. Step S102: Extract the thickness d of the cable surface covering layer collected by the sensor on the robot. In the coordinate system, translate the lower edge of the detection frame upward by d units and translate the upper edge of the detection frame downward by d units. The upper and lower edge lines after translation by d units form the target detection frame. Read the coordinates of the lowest point (x0, y0), the left endpoint (x1, y1), and the right endpoint (x2, y2) of the target detection frame from the coordinate system. Step S103: Based on the read coordinates of the lowest point (x0, y0), the left endpoint (x1, y1), and the right endpoint (x2, y2), calculate the cable sag angle using the following formula: ； in, Indicates the sag angle of the cable; Step S104: Identify the cable model in the target detection box using image recognition technology. Retrieve the sag angle threshold of the cable corresponding to the cable model from the database. Compare the calculated sag angle of the cable with the sag angle threshold. If the sag angle is greater than the sag angle threshold, mark the cable as the target cable. Obtain the geographical location of the target cable and store the cable model, geographical location, and sag angle of the target cable.

3. The artificial intelligence-based cable defect image recognition method according to claim 1, characterized in that, Step S200 includes: Step S201: Extract the geographical location of the target cable, retrieve the temperature and humidity forecasts for the geographical location from the meteorological department, sort the temperature and humidity forecasts according to time series, mark the time periods in which the temperature and humidity forecasts remain unchanged, and divide the time into several time periods. The predicted thickness of the cover layer of the target cable for a certain time period is calculated using the following formula: ； Wherein, D represents the predicted value of the cover layer thickness of the target cable corresponding to the time period, d represents the cover layer thickness of the target cable at the current time point, k1 represents the growth coefficient, T represents the predicted value of the temperature during the time period, β represents the humidity influence factor, and H represents the predicted value of the humidity during the time period. Step S202: Sort the calculated predicted values ​​of the target cable's sheath thickness according to the time series to generate a sheath thickness prediction curve; extract the sheath thickness corresponding to a certain time point from the sheath thickness prediction curve, and calculate the degree of deformation of the target cable at that time point using the following formula: ； Wherein, G represents the degree of deformation of the target cable at the time point, and D represents the thickness of the covering layer on the surface of the target cable at the time point. y1 is the x-coordinate of the left endpoint of the target detection box, y2 is the x-coordinate of the right endpoint of the target detection box, and y3 is the y-coordinate of the right endpoint of the target detection box. Step S203: Obtain the calculated deformation degree of the target cable and the corresponding time point, and plot the deformation degree prediction curve of the target cable.

4. The artificial intelligence-based cable defect image recognition method according to claim 1, characterized in that, Step S300 includes: Step S301: Extract the deformation degrees G1 and G2 of the target cable corresponding to two adjacent time points t1 and t2 from the deformation degree prediction curve of the target cable, and calculate the deformation development rate of the target cable. The calculation formula is as follows: ； Where v represents the deformation development rate of the target cable at two adjacent time points; Step S302: Calculate the weighted average of the calculated deformation development rates of the target cable to obtain the average deformation development rate of the target cable; extract the average deformation development rates of all target cables, sort the average deformation development rates in ascending order to obtain an average deformation development rate sequence; and obtain the average value of the average deformation development rate sequence according to the preset average value calculation formula and standard deviation calculation formula. Sum and standard deviation α; set the first risk threshold as Set the second risk threshold as ; Step S303: Classify the target cable with average deformation development rate V≤V1 into the first risk level, classify the target cable with average deformation development rate V1<V<V2 into the second risk level, and classify the target cable with average deformation development rate V≥V2 into the third risk level; set the early warning coefficient corresponding to the first risk level as Q1, set the early warning coefficient corresponding to the second risk level as Q2, set the early warning coefficient corresponding to the third risk level as Q3, and Q1<Q2<Q3.

5. The artificial intelligence-based cable defect image recognition method according to claim 1, characterized in that, Step S400 includes: Step S401: Calculate the priority of the target cable defect detection according to the preset priority calculation formula. The calculation formula is as follows: ； Where P represents the priority of the target cable, Q i G represents the warning coefficient corresponding to the i-th risk level of the target cable, and G represents the degree of deformation of the current target cable. Step S402: Sort all the calculated priority of the target cables in descending order, and at the same time, the robot extracts the geographical location of the target cables according to the priority order, and performs defect detection on the target cables according to the geographical location of each target cable.

6. An artificial intelligence-based cable defect image recognition system, used to execute the artificial intelligence-based cable defect image recognition method according to any one of claims 1-5, characterized in that, The system includes: a data acquisition module, a prediction module, an evaluation module, and an operation module; The acquisition module is used to acquire cable images and overlay thickness data, analyze the acquired cable images based on the overlay thickness data, calculate the cable sag angle, and mark the target cable based on the calculated sag angle. The prediction module obtains temperature and humidity prediction values ​​from the meteorological department, generates a cover layer thickness prediction curve, and calculates the degree of deformation of the target cable at each time point based on the cover layer thickness prediction value corresponding to each time point in the generated cover layer thickness prediction curve, thereby generating a deformation degree prediction curve of the target cable. The evaluation module calculates the deformation development rate of the target cable based on the deformation degree prediction curve of the target cable, classifies the risk level of the target cable, and calculates the early warning coefficient corresponding to each risk level. The operation module calculates the priority of defect detection for the target cable based on the warning coefficient of the risk level corresponding to the target cable and the current degree of deformation of the target cable. The robot automatically plans the detection path according to the calculated priority.

7. The artificial intelligence-based cable defect image recognition system according to claim 6, characterized in that, The acquisition module includes: an image acquisition unit, a cover layer thickness acquisition unit, a coordinate establishment unit, and a target marking unit; The image acquisition unit captures cable images below zero degrees Celsius in winter using a high-definition camera; The cover layer thickness acquisition unit obtains the cover layer thickness on the cable surface through a sensor; The coordinate establishment unit extracts the collected cable images below zero degrees Celsius in winter, identifies the cable outline in the cable image using image recognition technology, selects the cable along the cable outline using a detection box, and establishes a coordinate system in the cable image with the lowest point of the detection box as the origin. The target marking unit extracts the coating thickness of the cable surface collected by sensors mounted on the robot, translates the detection frame in the coordinate system according to the coating thickness to generate a target detection frame, reads the coordinates of the lowest point, left endpoint, and right endpoint of the target detection frame from the coordinate system, and calculates the cable sag angle; identifies the cable model in the target detection frame using image recognition technology, retrieves the sag angle threshold corresponding to the cable model from the database based on the cable model, compares the calculated cable sag angle with the sag angle threshold, and if the sag angle is greater than the sag angle threshold, marks the cable as a target cable, obtains the geographical location of the target cable, and stores the cable model, geographical location, and sag angle of the target cable.

8. The artificial intelligence-based cable defect image recognition system according to claim 6, characterized in that, The prediction module includes: a meteorological processing unit, a cover layer prediction unit, and a deformation prediction unit; The meteorological processing unit extracts the geographical location of the target cable, retrieves the temperature and humidity forecasts for the geographical location from the meteorological department, and sorts the temperature and humidity forecasts according to time series. The cover layer prediction unit marks the time periods in which both the temperature prediction value and the humidity prediction value remain unchanged, divides them into several time periods, and calculates the predicted value of the cover layer thickness of the target cable for each time period. The deformation prediction unit sorts the calculated predicted values ​​of the target cable's cover layer thickness according to the time series to generate a cover layer thickness prediction curve; it extracts the cover layer thickness corresponding to each time point from the cover layer thickness prediction curve, calculates the degree of deformation of the target cable at that time point, and plots a deformation degree prediction curve of the target cable.

9. The artificial intelligence-based cable defect image recognition system according to claim 6, characterized in that, The evaluation module includes: a rate calculation unit, a risk classification unit, and a coefficient setting unit; The rate calculation unit extracts the deformation degree of the target cable corresponding to two adjacent time points from the deformation degree prediction curve of the target cable, and calculates the deformation development rate of the target cable. The risk classification unit calculates the deformation development rate of the target cable by weighted averaging to obtain the average deformation development rate of the target cable; extracts the average deformation development rate of all target cables, sorts the average deformation development rate in ascending order to obtain an average deformation development rate sequence, and obtains the average value and standard deviation of the average deformation development rate sequence according to the preset average value calculation formula and standard deviation calculation formula; and sets a first risk threshold and a second risk threshold. The coefficient setting unit classifies target cables with an average deformation development rate less than or equal to a first risk threshold into a first risk level, target cables with an average deformation development rate greater than the first risk threshold but less than a second risk threshold into a second risk level, and target cables with an average deformation development rate greater than or equal to the second risk threshold into a third risk level; and sets a warning coefficient corresponding to each risk level.

10. The artificial intelligence-based cable defect image recognition system according to claim 6, characterized in that, The task module includes: a priority calculation unit and a path planning unit; The priority calculation unit calculates the priority of the target cable defect detection according to the preset priority calculation formula; The path planning unit sorts the calculated priorities of all target cables in descending order, while the robot extracts the geographical locations of the target cables according to the priority order and performs defect detection on each target cable based on its geographical location.

Images