CN115542362A - High-precision space positioning method, system, equipment and medium for electric power operation site - Google Patents

Abstract

The invention discloses a high-precision spatial positioning method, system, equipment, and medium for an electric power operation site. The invention belongs to the field of spatial positioning technology, including obtaining the positioning of the moving target of the electric power operation based on the visual positioning technology; according to the positioning of the moving target of the electric power operation, Carry out image analysis on the moving target of electric power work, obtain the characteristic information of the moving target of electric power work and construct the three-dimensional model of the moving target of electric power work; refine the three-dimensional model of the moving target of electric power work according to the characteristic information, obtain the high-precision three-dimensional model and update the described in real time Positioning of high-precision 3D models in 3D scenes. By attaching UWB or Beidou positioning tags to the dynamic target, image processing is performed on the basis of the positioning of the moving target, and the refined reconstruction of the moving target in the 3D scene is realized, so as to obtain the real-time position information of the dynamic target and display it in the 3D scene. Update the dynamic target model position status in real time.

Description

High-precision spatial positioning method, system, equipment and medium for electric power work site

technical field

The invention relates to the technical field of space positioning, in particular to a high-precision space positioning method, system, equipment and medium for electric power work sites.

Background technique

Research on vision-based target detection, recognition and tracking algorithms is a hotspot in the field of artificial intelligence. In recent years, great progress has been made with the wide application of deep learning. Deep learning technology has been gradually applied in the power production environment to realize the identification of some targets such as insulators, overhead arcs, foreign objects in substations, and equipment temperature. However, the scene of substations is complex, and the identification of key targets in all aspects of power operations still needs further research.

Integrating real scene video information into 3D virtual model is a branch of virtual reality technology, or a stage of development of virtual reality. 3D video fusion technology refers to the matching and fusion of one or more video sequences of camera images and related 3D virtual scenes to generate a new dynamic virtual scene or model of this scene, realizing virtual scene and real-time video fusion. The 3D video fusion technology can rely on a separate 3D engine to realize the fusion application of resources such as small-scale or partial 3D scenes and videos, and can also rely on the 3D geographic information system to realize the visual fusion of 3D geographic information in a wide area around the world application.

In the electric power operation site, accurate and effective positioning of the operator's position is an important basis for preventing accidents and ensuring the safety of workers. At present, GPS, Beidou positioning system and UWB positioning system are mostly used to locate the workers on the electric power work site, but only relying on GPS, Beidou positioning system and UWB positioning system cannot extract accurate spatial position information. There are problems such as occlusion, target posture change, and lighting on the job site, which will also affect the positioning of the job target at the power job site.

In summary, the research on vision-based target detection, recognition and tracking algorithms has made some research progress, but how to deal with the impact of occlusion, target pose changes, lighting and other issues is still the focus of research. The video image data processed by the existing visual positioning algorithm is the projection of the target from the three-dimensional space to the two-dimensional space, which cannot extract accurate spatial position information.

The problems of prior art are as follows:

The video image data processed by the existing visual positioning algorithm used in the electric power operation site is the projection of the target from the three-dimensional space to the two-dimensional space, and the precise spatial position information in the three-dimensional space cannot be extracted.

Contents of the invention

The technical problem to be solved in this application is that the video image data processed by the existing visual positioning algorithm for the electric power operation site is the projection of the target from the three-dimensional space to the two-dimensional space, and the accurate spatial position information of the three-dimensional space cannot be extracted. On-site high-precision spatial positioning methods, systems, equipment and media, by attaching UWB or Beidou positioning tags to dynamic targets, image processing is performed on them on the basis of positioning of moving targets, so as to realize fine precision of moving targets in 3D scenes In order to obtain the real-time position information of the dynamic target and update the position status of the dynamic target model in the 3D scene in real time, it solves the problem that the existing visual positioning algorithm cannot extract the accurate spatial position information in the 3D space.

The application is realized through the following technical solutions:

The first aspect of this application provides a high-precision spatial positioning method for an electric power operation site, including

S1. Based on the visual positioning technology, obtain the positioning of the electric power operation moving target; among them, the UWB auxiliary visual positioning technology is used for the indoor electric power operation moving target, and the Beidou auxiliary visual positioning technology is used for the outdoor electric power operation moving target; UWB auxiliary visual positioning technology Including UWB-assisted passive video positioning;

S2. According to the positioning of the moving target of the electric work, image analysis is performed on the moving target of the electric work, the characteristic information of the moving target of the electric work is obtained and a three-dimensional model of the moving target of the electric work is constructed;

S3. Refining the three-dimensional model of the moving target of electric power work according to the characteristic information, obtaining a high-precision three-dimensional model and updating the positioning of the high-precision three-dimensional model in the three-dimensional scene in real time.

In the above technical scheme, UWB-assisted visual positioning technology is used for indoor electric power operation moving targets, and Beidou auxiliary visual positioning technology is used for outdoor electric power work moving targets. By combining simple UWB positioning technology, Beidou positioning technology and visual positioning technology Combined, the positioning of the moving target of the electric power operation in the three-dimensional space can be obtained, and problems such as interference and occlusion of the electric power operation scene can be reduced, thereby achieving centimeter-level positioning. By wearing UWB tags or Beidou positioning tags on moving targets in electric power operations, image analysis of moving targets is performed on the basis of moving target positioning to achieve refined reconstruction of moving targets in 3D scenes, thereby obtaining real-time position information of moving targets And in the 3D scene, the position state of the moving target model is updated in real time, so as to obtain the precise positioning of the moving target of the electric power work in the 3D space.

In an optional embodiment, the method for UWB-assisted passive video positioning includes:

The binocular vision system formed by two cameras locates the moving target of electric power work, and obtains the position information of the moving target of electric power work in space;

Based on the UWB model, the moving target of the electric power operation is positioned, and the UWB positioning information of the moving target of the electric power work is obtained;

The position information of the moving target in electric power work and the UWB positioning information of the moving target in electric power work are integrated to obtain the positioning of the moving target in electric power work.

In an optional embodiment, the method of integrating the position information of the moving object in electric power work in space and the UWB positioning information of the moving object in electric power work is as follows:

The position and heading measured by the binocular vision system and the position measured by the UWB model are used as observation values to establish the UWB/vision fusion observation equation; the UWB/vision fusion observation equation is as follows:

表示双目视觉系统vision测量的平面坐标，

表示UWB测量的平面坐标，

表示双目视觉系统的位置测量误差，

表示 UWB 位置测量误差，

表示双目视觉系统测量的平面坐标与UWB测量的平面坐标之间的偏转角，

表示中间矩阵。In the above formula,

Represents the plane coordinates measured by the binocular vision system vision,

Indicates the plane coordinates of the UWB measurement,

Indicates the position measurement error of the binocular vision system,

Indicates the UWB position measurement error,

Indicates the deflection angle between the plane coordinates measured by the binocular vision system and the plane coordinates measured by UWB,

represents the intermediate matrix.

In an optional embodiment, the construction method of the UWB model is as follows:

Establish an improved robust EKF model, and use the improved robust EKF model as the standard model of the UWB model;

Use statistical methods to judge whether there are gross errors in the improved robust EKF model;

If there is a gross error, call the robust EKF model as an improved robust EKF model;

If there is no gross error, call the EKF model as an improved robust EKF model.

In an optional embodiment, the positioning method of the Beidou-assisted visual positioning technology includes:

Based on the Kalman filter model and the predicted position of the operation boundary line, the region of interest is obtained;

Perform image edge detection on the region of interest to obtain the positioning region;

Build a multi-eye vision measurement model through the multi-eye vision sensor, and build multi-eye vision coordinates in the positioning area through the multi-eye vision measurement model;

Based on the weighted LM algorithm, optimize the multi-eye vision coordinates constructed in the positioning area through the multi-eye vision measurement model, and obtain the optimized multi-eye vision coordinates;

Based on Beidou positioning technology, the optimized multi-eye visual coordinates are converted into coordinates of the global positioning system.

In an optional embodiment, based on the Kalman filter model, the method for dynamically acquiring the region of interest through multiple visual sensors is as follows:

Predict the current straight line position based on Kalman filter to obtain the first dynamic region of interest;

Based on the projection method, the column number of the operation boundary line in the image coordinates is predicted, and the second dynamic ROI is obtained based on the column number and combined with the current camera call situation.

In an optional embodiment, based on the weighted LM algorithm, the method for optimizing the multi-eye vision coordinates constructed in the positioning area by the multi-eye vision measurement model is as follows:

Non-linear optimization is performed on the image coordinates obtained after the object point is transformed to the i-th multi-eye vision sensor by minimizing the re-projection error;

Normalize the distance from the multi-eye vision sensor to the object, and convert the reciprocal of the normalized distance into a weighting factor;

The objective function is constructed by weighting factors, and the multi-eye vision coordinates are substituted into the objective function for calculation, and the optimized multi-eye vision coordinates are obtained.

The second aspect of the application provides a high-precision spatial positioning system for electric power work sites, including

UWB assisted video positioning module, the UWB assisted active video locating module is used to obtain the positioning of indoor power work moving targets, and the UWB assisted active video locating module includes an initialization unit for initializing environmental images and IMU data, and is used for obtaining visual A combined visual/inertial unit for coordinates and an ultra-wideband unit for acquiring UWB coordinates;

The Beidou auxiliary visual positioning module, the Beidou auxiliary visual positioning module is used to obtain the positioning of the moving target of the outdoor electric power operation, and the Beidou auxiliary visual positioning module includes a dynamic area of interest module for reducing environmental factors to extract the operation boundary line. A multi-eye vision module for constructing multi-eye vision area coordinates and a Beidou positioning module for converting multi-eye vision area coordinates into GPS coordinates.

The third aspect of the present application provides an electronic device, including a memory, a processor, and a computer program stored on the memory and operable on the processor, characterized in that, when the processor executes the program, it is used to realize power A high-precision spatial positioning method for a job site.

The fourth aspect of the present application provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it is used to implement a high-precision spatial positioning method for an electric power operation site.

Compared with the prior art, the present application has the following advantages and beneficial effects:

By attaching UWB or Beidou positioning tags to dynamic targets, image processing is performed on them based on the positioning of the moving targets to achieve fine reconstruction of the moving targets in the 3D scene, so as to obtain the real-time position information of the dynamic targets and In the 3D scene, the position state of the dynamic target model is updated in real time, and the verification experiment proves that the error between the reconstructed dynamic model and the actual size is less than 1%.

Description of drawings

In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention. Therefore, it should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained according to these drawings without creative work. In the attached picture:

FIG. 1 is a schematic flow diagram of a high-precision spatial positioning method for an electric power operation site provided by an embodiment of the present application;

Fig. 2 is a schematic diagram of coordinate transformation of multi-eye vision fusion Beidou positioning provided by an embodiment of the present application;

Fig. 3 is the auxiliary navigation tracking result diagram in the test site when the speed provided by an embodiment of the present application is 0.4m/s;

Fig. 4 is the auxiliary navigation tracking result diagram in the test field when the speed provided by an embodiment of the present application is 0.8m/s;

Fig. 5 is a diagram of the auxiliary navigation tracking results in the test site when the speed is 1.2m/s provided by an embodiment of the present application.

detailed description

In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the examples and accompanying drawings. As a limitation of the present invention.

Example 1

This embodiment provides a high-precision spatial positioning method for an electric power operation site. The flow of the method is shown in Figure 1, including:

Step S1. Based on the visual positioning technology, the positioning information of the moving target of the electric power operation is obtained.

Among them, UWB-assisted visual positioning technology is used for indoor electric power operation moving targets, and Beidou-assisted visual positioning technology is used for outdoor electric power operation moving targets; UWB-assisted visual positioning technology includes UWB-assisted active video positioning and UWB-assisted passive video positioning.

Among them, for wearable devices that are not equipped with a vision system, the positioning of moving targets can be realized through the camera installed in the substation combined with UWB technology.

Among them, the improved robust EKF model is adopted as the standard model of the UWB model.

In the improved robust EKF positioning model, the robust EKF gain matrix is constructed as follows:

和

是稳健参数，

取2.5-3.5，

取3.5-4.5，

为EKF模型的增益矩阵，

为抗差优化后的增益矩阵，

为残差相关阈值。

and

is a robust parameter,

Take 2.5-3.5,

Take 3.5-4.5,

is the gain matrix of the EKF model,

is the gain matrix optimized for robustness,

is the residual correlation threshold.

表示观测向量维度。

、

和

分别表示观测向量

的预测残差、冗余观测分量和测量标准差。冗余观测分量可表示为：

Indicates the observation vector dimension.

,

and

Respectively represent the observation vector

The prediction residuals, redundant observation components, and measurement standard deviations of . Redundant observation components can be expressed as:

和

分别是残差向量的协方差矩阵和观测值的权重矩阵。每次更新迭代给定迭代次数，状态预测值

，如下：

and

are the covariance matrix of the residual vector and the weight matrix of the observations, respectively. For each update iteration given the number of iterations, the state prediction value

,as follows:

，如下：prediction residual

,as follows:

在t时刻迭代的状态预测值由t-1时刻的状态滤波值及其预测残差确定，

表示某时刻迭代结束的计算赋值，

表示初始时刻的预测值，

为分布特性的矩阵，

为测量值观测。根据上述公式

计算等效增益矩阵，抗差滤波值为：

The state prediction value iterated at time t is determined by the state filter value and its prediction residual at time t-1,

Indicates the calculation assignment at the end of the iteration at a certain moment,

Indicates the predicted value at the initial moment,

is a matrix of distribution properties,

Observations for measured values. According to the above formula

Calculate the equivalent gain matrix, and the robust filter value is:

为抗差EKF增益矩阵，

为t时刻迭代结束的计算值，

为由t-1时刻预测的t时刻预测值。如果

与

之差小于给定的极限差，则迭代结束。当t=1时，

是k时刻标准EKF的赋值。后验协方差矩阵为：

is the robust EKF gain matrix,

is the calculated value at the end of the iteration at time t,

is the predicted value at time t predicted from time t-1. if

and

If the difference is less than the given limit difference, the iteration ends. When t=1,

is the assignment of the standard EKF at time k. The posterior covariance matrix is:

表示状态向量维度，

是迭代结束时的最终等价卡尔曼滤波增益矩阵，

为协方差矩阵。

represents the state vector dimension,

is the final equivalent Kalman filter gain matrix at the end of the iteration,

is the covariance matrix.

In order to improve the positioning angle, a vision/UWB fusion EKF positioning model is constructed considering the visual scale factor and initial direction, and the state and measurement equations are expressed as:

和

分别表示状态方程和量测方程，

和

分别是协方差阵

和

的独立、零均值、高斯噪声过程，

为状态转移矩阵，H为分布特性的矩阵。In the above formula,

and

represent the state equation and the measurement equation, respectively,

and

covariance matrix

and

Independent, zero-mean, Gaussian noise process of ,

Is the state transition matrix, and H is the matrix of distribution characteristics.

，

表示平面坐标，

表示行人速度，

表示移动方向角，

表示比例尺模糊度，

表示视觉计算的平面坐标与UWB计算的平面坐标之间的偏转角。In the above formula,

,

represent plane coordinates,

represents the pedestrian speed,

represents the direction angle of movement,

represents the scale ambiguity,

Indicates the deflection angle between the plane coordinates calculated by vision and the plane coordinates calculated by UWB.

According to the error equation of vision and UWB, the corresponding state model is:

If the vision-measured position, heading and UWB-measured position are taken as observations, the UWB/vision fusion observation equation can be expressed as:

表示双目视觉系统vision测量的平面坐标，

表示UWB测量的平面坐标，

表示双目视觉系统的位置测量误差，

表示 UWB 位置测量误差，

表示双目视觉系统测量的平面坐标与UWB测量的平面坐标之间的偏转角。In the above formula,

Represents the plane coordinates measured by the binocular vision system vision,

Indicates the plane coordinates of the UWB measurement,

Indicates the position measurement error of the binocular vision system,

Indicates the UWB position measurement error,

Indicates the deflection angle between the plane coordinates measured by the binocular vision system and the plane coordinates measured by UWB.

In an optional embodiment, Beidou-assisted visual positioning technology can be used for outdoor electric power work moving targets. Restricted by the positioning accuracy of Beidou and the movement of on-site construction equipment and the geographical environment, currently only relying on Beidou positioning can only achieve meter-level precision positioning, which cannot meet the centimeter-level high-precision positioning requirements of dynamic power operation scenarios. Therefore, the Beidou positioning technology is combined with the multi-eye vision-assisted positioning technology to provide positioning accuracy.

，已知行驶航线上的一点P在

坐标系下的坐标为

、G点在

坐标系下的坐标为

，以G点为原点建立

坐标系，则P点在

坐标系下的坐标

为As shown in Figure 2, the coordinate system in this figure is the coordinate system established after the iteration of the optimized weighted LM algorithm, and a point in the coordinate system is taken

, a point P on the known driving route is at

The coordinates in the coordinate system are

, G point at

The coordinates in the coordinate system are

, establish with point G as the origin

coordinate system, then point P is at

coordinates in the coordinate system

for

中坐标的单位是像素，可由此得出单位为长度m的坐标(x _l，y _l)：in,

The unit of the coordinates is the pixel, so the coordinates ( x _l , y _l ) whose unit is the length m can be obtained from this:

The coordinate system is established with the fixed point C on the camera as the origin, and the horizontal distance from point C to point G on the ground can be measured as l ₁ is 2.5m. It can be known that the coordinates of point P O _c in the coordinate system are:

The construction machinery is equipped with a Beidou positioning and navigation device. Through the calculation, the coordinates of the vehicle control point M in the earth plane coordinate system O in units of m and the heading angle of the vehicle can be obtained. The coordinate system O _m is established with point M as the origin, the measured distance from the fixed point O _c on the camera is l ₂ , and the coordinates of point P in the O _m coordinate system are:

Through the positioning system, it can be known that the coordinates of M in the earth coordinate system O are ( x _w , y _w ), and the coordinates ( x _o , y _o ) of point P in the earth coordinate system can be obtained through calculation, namely:

To sum up, given the image coordinates ( x _p , y _p ) of a point P on the flight path, its coordinates ( x _o , y _o ) in the GPS can be obtained, namely:

为矩阵

，

为矩阵

，

为矩阵

，

为矩阵

，

为矩阵

。in,

for the matrix

,

for the matrix

,

for the matrix

,

for the matrix

,

for the matrix

.

In the multi-eye vision measurement system, the method of minimizing the reprojection error is usually used for nonlinear optimization, that is,

表示调整前物点变换至第i个相机后得到的图像坐标，

表示调整后的物点变换至第i个相机后得到的图像坐标。当各相机参数、测量环境等因素均相同时，相机到物点的距离对成像噪声影响较为明显，距离越远，噪声越大，反之噪声越小。在设计目标函数时，将相机到物点距离的倒数转化为加权因子，同时为了增加不同相机之间的可比性，对距离信息进行归一化处理。In the above formula:

Indicates the image coordinates obtained after the object point is transformed to the i-th camera before adjustment,

Indicates the image coordinates obtained after the adjusted object point is transformed to the i-th camera. When the camera parameters, measurement environment and other factors are the same, the distance from the camera to the object point has a more obvious impact on the imaging noise. The farther the distance is, the greater the noise is, and vice versa. When designing the objective function, the reciprocal of the distance from the camera to the object point is converted into a weighting factor, and in order to increase the comparability between different cameras, the distance information is normalized.

为加权因子，

为物点o到相机i的距离，

为oi物点的坐标，

为ci相机的坐标。因此，最终的目标函数为In the above formula,

is the weighting factor,

is the distance from object point o to camera i,

is the coordinates of the oi object point,

is the coordinates of the ci camera. Therefore, the final objective function is

代入目标函数并转化为方程组形式，可得Will

Substituting into the objective function and transforming into the form of equations, we can get

为参量系数，F_ui、F_vi为优化的子目标函数，u_i，v_i为迭代求解的中间变量。In the above formula, x _w , y _w , and z _w represent unit vectors in three-dimensional space;

are parameter coefficients, F _ui and F _vi are optimized sub-objective functions, u _i and v _i are intermediate variables for iterative solution.

的导数，可得雅可比矩阵为Calculations About

The derivative of , the Jacobian matrix can be obtained as

In the above formula, the values in the Jacobian matrix J are the first-order partial derivatives of F _ui and F _vi with respect to x _w , y _w , and z _w directions respectively. Substituting the above formula into the following formula, the increment can be obtained as

In the above formula, μ is the updated damping coefficient. Therefore, the weighted LM algorithm is used to optimize the coordinates constructed in the positioning area through the multi-eye vision measurement model. The specific calculation steps are as follows:

(1) Calculate the initial value of the object point coordinates according to the orthogonal projection method, set the iteration termination constant as e, the actual drop effect threshold as ε, the number of iterations as k, and initialize μ;

的导数求解雅可比矩阵J，并计算μ;(2) According to the current estimated object point world coordinates and camera parameters, use the

The derivative of solves the Jacobian matrix J, and calculates μ;

(3) Calculate the increment ΔP;

(4) Calculate ρ and evaluate the current decline effect;

(5) If ρ<ε, then μ=0.5μ, and return to step (3), otherwise, proceed to step (6);

或迭代次数大于等于k，停止迭代，输出优化后的结果;否则返回步骤（2）进入下一轮迭代。(6) If

Or if the number of iterations is greater than or equal to k , stop the iteration and output the optimized result; otherwise, return to step (2) to enter the next iteration.

In order to analyze the real-time performance of the operation boundary line extraction algorithm, this embodiment counts the image processing steps of 100 construction process pictures before and after the algorithm improvement, and the average time of the entire image processing process. The results are shown in the following table:

It can be seen from the data in the above table that the average processing time of each image is shortened from 0.176087s before the algorithm improvement to 0.064547s, which is 63.3% shorter than the original processing time. processing time requirements.

When the speed is 0.4m/s, the auxiliary navigation tracking results in the field are shown in Figure 3, and the auxiliary navigation results in the field plot when the speed is 0.4m/s are shown in the following table:

It can be seen from Figure 3 and the above table that when the construction machinery is working in a straight line at the speed of 0.4m/s in the test site, the maximum lateral deviation is 6.5cm after the first operation until the lateral deviation reaches 0 when the initial deviation is 30cm , with a mean of 0.49cm and a standard deviation of 3.05cm. The results show that the construction machinery can continuously update the Beidou positioning data and visual picture information in the case of a large initial deviation in the test site, and the system simulation trajectory converges with the actual position of the construction machinery. With the extension of the job tracking time, the positioning accuracy of the simulation job becomes more stable, and the auxiliary navigation effect can well meet the requirements.

When the speed is 0.8m/s, the auxiliary navigation tracking results in the field are shown in Figure 4, and the auxiliary navigation results in the field plot when the speed is 0.8m/s are shown in the following table:

It can be seen from Figure 4 and the above table that when the construction machine is working in a straight line at the speed of 0.8m/s in the test site, the maximum lateral deviation is 13.7cm when the initial deviation is 30cm, after the first operation until the lateral deviation reaches 0 , with a mean of 0.82cm and a standard deviation of 5.98cm. In the case of a large initial deviation, adjust the pose of the construction machine so that the construction machine can walk according to the expected trajectory. The results show that the construction machinery can continuously update the Beidou positioning data and visual picture information in the case of a large initial deviation in the test site, and the system simulation trajectory converges with the actual position of the construction machinery. With the extension of the job tracking time, the positioning accuracy of the simulation job becomes more stable, and the auxiliary navigation effect can well meet the design requirements.

When the speed is 1.2m/s, the auxiliary navigation tracking results in the field are shown in Figure 5, and the auxiliary navigation results in the field plot when the speed is 0.8m/s are shown in the following table:

It can be seen from Figure 5 and the above table that when the construction machine is working in a straight line at a speed of 1.2m/s in the test site, the maximum lateral deviation is 19.2cm after the first operation until the lateral deviation reaches 0 when the initial deviation is 30cm , with a mean of 1.40cm and a standard deviation of 7.06cm. The result can basically meet the design requirements of auxiliary navigation, but there are small-scale obvious fluctuations in the auxiliary navigation process. The follow-up analysis of the wheel marks is caused by the braking and steering of construction machinery during the formal process. This deviation does not affect the imitation Beidou auxiliary visual positioning technology The application effect in the simulation system.

The field test results show that when the construction machinery travels at speeds of 0.4m/s, 0.8m/s, and 1.2m/s, the maximum lateral deviations are 6.5cm, 13.7cm, and 19.2cm, and the average deviations are 0.49cm, 0.87cm, 1.40cm, and the deviation and standard deviation are 3.05cm, 5.98cm, 7.06cm respectively. The positioning accuracy is higher than that of Beidou navigation alone, which can meet the positioning accuracy requirements of simulation system-assisted navigation in the power operation environment.

Step S2, by analyzing the image of the electric power work moving target at the location, the feature information of the electric power work moving target is obtained and a three-dimensional model of the electric power work moving target is constructed.

Using real-world 3D model scanning technology, point cloud fusion technology, and refined modeling technology to model various maintenance vehicles and operators, and restore their real structures and textures. The maintenance car model is named according to the model, and forms a dynamic target refined model library with the operator model.

For the overhaul vehicle, its model code is identified through image recognition technology, and its model code is used as characteristic information.

Specifically, the steps are as follows:

A. Use dynamic collection to collect vehicle information from real-time video streaming images of surveillance video.

B. Perform preprocessing such as noise filtering, contrast enhancement, and image scaling on the collected maintenance vehicle model coded images.

C. Based on information such as license plate texture features, color features, and shape features, algorithms such as projection analysis, connected domain analysis, and machine learning are used to detect license plates. The projection analysis method locates the model code of the overhaul vehicle through the projection analysis of the image in the horizontal and vertical directions according to the feature that the license plate characters and the background alternately appear more than other parts.

Connected domain analysis According to the characteristics that each character in the maintenance vehicle model code is a connected domain and the structure and color of these connected domains are consistent, the maintenance vehicle model code is located by detecting and merging these connected domains.

D. After extracting the code area of the maintenance vehicle model, it is necessary to divide the code area of the maintenance vehicle model in units of one character.

E. Normalize the grayscale image of the segmented characters, extract features, and then perform machine learning or match with character database templates, and finally select the result with the highest matching degree as the recognition result.

For workers, through image recognition technology, individual representative appearance and geometric features are identified as feature information.

Specifically, the steps are as follows:

A. Establish a face image database.

B. According to the face feature information, extract the face image features in the face image database.

Step S3: refine the 3D model of the moving target in electric power work according to the characteristic information, obtain a high-precision 3D model and update the positioning of the high-precision 3D model in the 3D scene.

The dynamic target refinement model library contains various maintenance vehicle models and operator models. For the overhaul vehicle, it is extracted in the database through the identified model code; for the operator, it is extracted through the identified name of the personnel, and the 3D model is adjusted according to the height of the operator in the information.

By attaching UWB or Beidou positioning tags to dynamic targets, on the basis of moving target positioning, a series of algorithm operations including image acquisition, preprocessing, model code positioning, character segmentation, character recognition, and result output are used to achieve moving targets. The refined reconstruction in the 3D scene obtains the real-time position information of the dynamic target and updates the position status of the dynamic target model in real time in the 3D scene. The verification experiment proves that the error between the reconstructed dynamic model and the actual size is less than 1%.

Example 2

On the basis of Embodiment 1, this embodiment provides a high-precision spatial positioning system for an electric power operation site, including:

UWB assisted video positioning module, the UWB assisted video positioning module is used to obtain the positioning of indoor power work moving targets, and the UWB assisted video positioning module includes an initialization unit for initializing environmental images and IMU data, and a visual/visual system for obtaining visual coordinates Inertial combination unit and ultra-wideband unit for obtaining UWB coordinates;

The Beidou auxiliary visual positioning module, the Beidou auxiliary visual positioning module is used to obtain the positioning of the moving target of the outdoor electric power operation. Multi-eye vision module for multi-eye vision area coordinates and Beidou positioning module for converting multi-eye vision area coordinates into GPS coordinates.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the scope of the present invention. Protection scope, within the spirit and principles of the present invention, any modification, equivalent replacement, improvement, etc., shall be included in the protection scope of the present invention.

Claims (10)

1. A high-precision spatial positioning method for an electric power operation site, characterized in that it includes: S1. Based on the visual positioning technology, obtain the positioning of the electric power operation moving target; among them, the UWB auxiliary visual positioning technology is used for the indoor electric power operation moving target, and the Beidou auxiliary visual positioning technology is used for the outdoor electric power operation moving target; UWB auxiliary visual positioning technology Including UWB-assisted passive video positioning; S2. According to the positioning of the moving target of the electric work, image analysis is performed on the moving target of the electric work, the characteristic information of the moving target of the electric work is obtained and a three-dimensional model of the moving target of the electric work is constructed; S3. Refining the three-dimensional model of the moving target of electric power work according to the characteristic information, obtaining a high-precision three-dimensional model and updating the positioning of the high-precision three-dimensional model in the three-dimensional scene in real time. 2. The high-precision spatial positioning method of an electric power operation site according to claim 1, wherein the method for passive video positioning assisted by the UWB comprises: The binocular vision system formed by two cameras locates the moving target of electric power work, and obtains the position information of the moving target of electric power work in space; Based on the UWB model, the moving target of the electric power operation is positioned, and the UWB positioning information of the moving target of the electric power work is obtained; The position information of the moving target in electric power work and the UWB positioning information of the moving target in electric power work are integrated to obtain the positioning of the moving target in electric power work. 3. The high-precision spatial positioning method of the electric power work site according to claim 2, wherein the method of integrating the position information of the electric power work moving target in space and the UWB positioning information of the electric power work moving target is as follows: The position and heading measured by the binocular vision system and the position measured by the UWB model are used as observation values to establish the UWB/vision fusion observation equation; the UWB/vision fusion observation equation is as follows:

In the above formula,

Represents the plane coordinates measured by the binocular vision system vision,

Indicates the plane coordinates of the UWB measurement,

Indicates the position measurement error of the binocular vision system,

Indicates the UWB position measurement error,

Indicates the deflection angle between the plane coordinates measured by the binocular vision system and the plane coordinates measured by UWB,

represents the intermediate matrix. 4. the high-precision spatial positioning method of the electric power work site according to claim 3, is characterized in that, the construction method of UWB model is as follows: Establish an improved robust EKF model, and use the improved robust EKF model as the standard model of the UWB model; Use statistical methods to judge whether there are gross errors in the improved robust EKF model; If gross errors exist, the robust EKF model is called as the improved robust EKF model. 5. The high-precision spatial positioning method of an electric power operation site according to claim 1, wherein the positioning method of the Beidou-assisted visual positioning technology comprises: Based on the Kalman filter model and the predicted position of the operation boundary line, the region of interest is obtained; Perform image edge detection on the region of interest to obtain the positioning region; Build a multi-eye vision measurement model through the multi-eye vision sensor, and build multi-eye vision coordinates in the positioning area through the multi-eye vision measurement model; Based on the weighted LM algorithm, optimize the multi-eye vision coordinates constructed in the positioning area through the multi-eye vision measurement model, and obtain the optimized multi-eye vision coordinates; Based on Beidou positioning technology, the optimized multi-eye visual coordinates are converted into coordinates of the global positioning system. 6. the high-precision spatial positioning method of the electric power work site according to claim 5, is characterized in that, based on the Kalman filter model, the method for dynamically obtaining the region of interest by a plurality of visual sensors is as follows: Predict the current straight line position based on Kalman filter to obtain the first dynamic region of interest; Based on the projection method, the column number of the operation boundary line in the image coordinates is predicted, and the second dynamic ROI is obtained based on the column number and combined with the current camera call situation. 7. The high-precision spatial positioning method of the electric power work site according to claim 5, characterized in that, based on the weighted LM algorithm, the method for optimizing the multi-eye vision coordinates constructed in the positioning area by the multi-eye vision measurement model is as follows : Non-linear optimization is performed on the image coordinates obtained after the object point is transformed to the i-th multi-eye vision sensor by minimizing the re-projection error; Normalize the distance from the multi-eye vision sensor to the object, and convert the reciprocal of the normalized distance into a weighting factor; The objective function is constructed by weighting factors, and the multi-eye vision coordinates are substituted into the objective function for calculation, and the optimized multi-eye vision coordinates are obtained. 8. A high-precision spatial positioning system for electric power work sites, characterized in that it includes: UWB assisted video positioning module, the UWB assisted video locating module is used to obtain the positioning of indoor power work moving targets, and the UWB assisted video locating module includes an initialization unit for initializing environmental images and IMU data, and an initialization unit for obtaining visual coordinates combined visual/inertial unit and ultra-wideband unit for acquiring UWB coordinates; The Beidou auxiliary visual positioning module, the Beidou auxiliary visual positioning module is used to obtain the positioning of the moving target of the outdoor electric power operation, and the Beidou auxiliary visual positioning module includes a dynamic area of interest module for reducing environmental factors to extract the operation boundary line. A multi-eye vision module for constructing multi-eye vision area coordinates and a Beidou positioning module for converting multi-eye vision area coordinates into GPS coordinates. 9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and operable on the processor, wherein the processor implements any of claims 1 to 7 when executing the program. The high-precision spatial positioning method of the electric power work site described in the item. 10. A computer-readable storage medium, on which a computer program is stored, characterized in that, when the program is executed by a processor, the high-precision spatial positioning method for an electric power work site according to any one of claims 1 to 7 is realized .

Images