Laser-assisted focusing system of power transmission line inspection camera

Abstract

The invention relates to the field of power transmission line inspection, and discloses a laser-assisted focusing system of a power transmission line inspection camera, and the system comprises a laser ranging module, a visual sensor, a main control module, a target analysis module and a strategy optimization module, and the laser ranging module measures the distance between a target part on a power transmission line and a high-speed zoom lens in real time; the vision sensor synchronously collects definition characteristic data of a target part, and the main control module fuses the two types of data to generate an initial focusing parameter matrix and controls the high-speed zoom lens to adjust the focal length; the target analysis module determines a focusing strategy instruction based on the current detection task and a target priority database; the strategy optimization module corrects the initial parameter matrix according to the instruction, generates an optimized parameter matrix and feeds back the optimized parameter matrix to the main control module; the system improves the accuracy, pertinence and efficiency of focusing in a complex environment through the cooperation of multiple modules, and is suitable for the intelligent inspection work of a power transmission line.

Description

Technical Field

[0001] This application relates to the field of power transmission line inspection technology, and in particular to a laser-assisted focusing system for a power transmission line inspection camera. Background Technology

[0002] In power transmission line inspection, the accuracy and timeliness of camera focusing directly affect inspection efficiency and data quality. Traditional focusing methods rely heavily on image sharpness information acquired by visual sensors alone for judgment. However, the environment in which power transmission lines are located is often complex and variable, easily interfering with visual focusing. For example, under strong light, the surface of target components may reflect light, leading to a decrease in the contrast of the image acquired by the visual sensor, making it difficult to accurately capture sharpness features. In low-light environments such as rain, fog, or haze, image noise increases, and edge information becomes blurred, further increasing the difficulty of focusing.

[0003] Transmission line components are diverse, including conductors, insulators, and tower connectors. These components differ in shape and material, resulting in varying sharpness characteristics even under the same shooting conditions. Traditional systems lack differentiated focusing strategies for different targets, often leading to over-focusing on minor components while lagging behind in focusing on critical ones. Furthermore, during inspections, cameras are typically in motion, and the distance between the target component and the camera changes in real time. Relying solely on visual feedback for focus adjustment can easily cause calculation delays, resulting in focusing speed failing to keep pace with the changing distance and frequent defocusing, severely impacting the effectiveness and continuity of inspection images.

[0004] In existing technologies, some focusing systems attempt to introduce a single distance measurement method to assist focusing. However, because they are not effectively integrated with visual sharpness features, it is difficult to handle complex scenarios such as differences in surface texture of target components and uneven lighting when determining the focal length solely based on distance data, resulting in insufficient focusing accuracy. These shortcomings lead to a significant amount of time being spent on image re-taking and post-processing during power transmission line inspections, increasing inspection costs and hindering the further development of intelligent inspection technology. Summary of the Invention

[0005] To address the aforementioned technical problems, this application proposes a laser-assisted focusing system for a power transmission line inspection camera.

[0006] The technical solution adopted in this application is: a laser-assisted focusing system for a power transmission line inspection camera, used to control the high-speed zoom lens of a power transmission line inspection drone to perform automatic focusing, including:

[0007] The laser ranging module is used to acquire real-time distance measurements between the high-speed zoom lens and the target components to be inspected on the power transmission line.

[0008] A visual sensor is used to simultaneously acquire sharpness feature data of the target component;

[0009] The main control module receives distance measurement values ​​and sharpness feature data, generates an initial focus parameter matrix through fusion processing, and controls the high-speed zoom lens to perform focus adjustment based on the initial focus parameter matrix.

[0010] The target analysis module is used to call the built-in target priority database and determine the focusing strategy instructions for the target components based on the current detection task.

[0011] The strategy optimization module is used to modify the initial focus parameter matrix according to the focus strategy instructions, generate the optimized focus parameter matrix and feed it back to the main control module. Finally, the main control module outputs the control signal to drive the high-speed zoom lens according to the focus parameter matrix.

[0012] Furthermore, the main control module includes:

[0013] The data fusion unit is used to weightedly fuse distance measurements and sharpness feature data, and output a fused feature matrix.

[0014] The parameter generation unit is used to query a preset focal length mapping relationship table based on the fusion feature matrix to generate an initial focusing parameter matrix containing the focal length adjustment range.

[0015] The execution control unit is used to drive the microelectromechanical system drive unit inside the high-speed zoom lens based on the initial focus parameter matrix, thereby controlling the focal length change of the high-speed zoom lens.

[0016] Furthermore, the target analysis module performs the following operations:

[0017] Parse the task type identifier of the current detection task and match the corresponding component priority weight from the target priority database;

[0018] Based on component priority weights, the spatial distribution of target components is sorted to generate a target focusing sequence;

[0019] Based on the target focusing sequence, determine the focusing strategy instructions that include the priority execution order.

[0020] Furthermore, the strategy optimization module includes:

[0021] The strategy parsing unit is used to extract the priority execution order in the focus strategy instructions and generate strategy constraints by combining sharpness feature data.

[0022] The matrix correction unit is used to adjust the focal length adjustment range in the initial focus parameter matrix according to the strategy constraints, and generate the optimized focus parameter matrix.

[0023] The iterative feedback unit is used to send the optimized focus parameter matrix to the execution control unit of the main control module.

[0024] Furthermore, it also includes:

[0025] The feedback calibration module is used to obtain the actual sharpness assessment value collected by the visual sensor after the high-speed zoom lens performs a focal length adjustment.

[0026] The comparison unit is used to calculate the deviation between the actual sharpness assessment value and the target sharpness threshold, and to trigger a calibration command when the deviation exceeds the preset tolerance range.

[0027] The parameter readjustment unit is used to respond to calibration commands, generate a calibrated focus parameter matrix based on the optimized focus parameter matrix after deviation correction, and input it to the main control module.

[0028] Furthermore, the feedback calibration module performs the following operations:

[0029] Construct a sharpness distribution heatmap based on the actual sharpness assessment values;

[0030] Extract the coordinates of low-resolution regions below the target resolution threshold from the resolution distribution heatmap;

[0031] The coordinates of the low-resolution area are matched with the spatial distribution of the target component, generating an area calibration command and sending it to the parameter readjustment unit.

[0032] Furthermore, the parameter readjustment unit performs the following operations:

[0033] Locate the identifier of the target component to be calibrated according to the area calibration instructions;

[0034] Calculate the focal length compensation value corresponding to the target component identifier based on the deviation amount;

[0035] Update the focal length compensation value in the optimized focus parameter matrix to generate the calibrated focus parameter matrix.

[0036] Furthermore, the target priority database includes:

[0037] A component type index table is used to associate the type identifiers and defect detection priority coefficients of transmission line hardware, conductors, and insulators.

[0038] The task strategy mapping table is used to associate the task identifiers and corresponding focus accuracy parameters of defect identification tasks and texture acquisition tasks.

[0039] Furthermore, the target analysis module also performs:

[0040] Query the task strategy mapping table based on the task identifier of the current detection task to obtain the target focus accuracy parameters;

[0041] Adjust the sampling frequency of the laser ranging module and the exposure parameters of the vision sensor based on the target focusing accuracy parameters.

[0042] Furthermore, the strategy optimization module also performs:

[0043] The convergence termination condition of the optimized focus parameter matrix is ​​set according to the target focus accuracy parameters.

[0044] During the focal length adjustment process of a high-speed zoom lens, the matching degree between the actual focal length change and the optimized focus parameter matrix is ​​compared in real time, and the control signal output is terminated when the convergence termination condition is met.

[0045] The advantages of this application over the prior art are as follows:

[0046] By working collaboratively with the laser ranging module and the vision sensor, the simultaneous acquisition of distance measurements and sharpness feature data is achieved, providing multi-dimensional foundational information for determining focusing parameters. The main control module fuses the two types of data to generate an initial focusing parameter matrix. This allows focus adjustment to reference both the actual distance to the target and the visual sharpness, effectively avoiding potential biases that may arise from a single data source in complex environments. For example, when strong lighting reduces visual image contrast, the distance data obtained from laser ranging can provide a stable reference for focusing; while when target components are close in distance but have significantly different surface textures, the sharpness features of the vision sensor can help distinguish between different targets, improving the targeting accuracy.

[0047] The target analysis module calls upon the built-in target priority database to determine the focusing strategy instructions for the target components based on the current inspection task, enabling the focusing process to prioritize the needs of critical components. During the inspection process, different inspection tasks may focus on different components; for example, some tasks focus on insulators, while others focus on conductor connections. Through priority judgment, the system can automatically allocate focusing resources to more important components when multiple targets appear simultaneously, reducing ineffective focusing on secondary components and improving the purposefulness and efficiency of focusing.

[0048] The strategy optimization module modifies the initial focus parameter matrix based on the focus strategy instructions, generates an optimized focus parameter matrix, and feeds it back to the main control module, further improving the adaptability of the focus parameters. The initial focus parameter matrix may only be generated based on basic data, while the optimization process incorporates specific task requirements and target priorities, ensuring that focus adjustment is not only accurate but also better suited to the requirements of actual inspection scenarios. For example, for high-priority target components, even if their distance is similar to other components, the optimized parameters will ensure faster and more accurate focusing, reducing focus delays caused by lag in target priority judgment.

[0049] Through this series of collaborative mechanisms, the system can more quickly adjust its focus when dealing with rapid changes in target distance during mobile inspections, reducing the number of times the image is out of focus due to distance changes. The real-time distance data provided by the laser ranging module can quickly reflect changes in target position. Combined with dynamic analysis of visual features, this allows for more timely updates to focus parameters, ensuring image continuity and clarity during inspections. Simultaneously, the collaborative operation of multiple modules reduces interference from complex environmental factors on focus. Whether in strong light, low light, or densely packed target scenarios, the system maintains stable focusing performance, reducing invalid images caused by focusing issues. This provides more reliable image data support for power line inspections, helps reduce post-processing workload, and promotes efficient inspection work. Attached Figure Description

[0050] The following description, in conjunction with the accompanying drawings, further illustrates this application:

[0051] Figure 1 This is a timing diagram of the laser-assisted focusing system of a power transmission line inspection camera in operation, as shown in an embodiment of this application.

[0052] Figure 2 A flowchart illustrating the operation of the main control module;

[0053] Figure 3 A flowchart illustrating the work of the strategy optimization module;

[0054] Figure 4 A flowchart for the regional calibration of the feedback calibration module;

[0055] Figure 5 This is a flowchart for parameter readjustment unit calibration. Detailed Implementation

[0056] like Figures 1 to 5 As shown, this application provides a laser-assisted focusing system for a power transmission line inspection camera, used to control the high-speed zoom lens of a power transmission line inspection drone to automatically focus. The system includes a laser ranging module, a vision sensor, a main control module, a target analysis module, a strategy optimization module, a feedback calibration module, and auxiliary components such as a high-definition auxiliary camera and a lens position encoder.

[0057] The core components of a high-speed zoom lens include the lens element group (used to achieve focal length changes and imaging), the actuator of the microelectromechanical system drive unit (which is controlled by the system to move the lens element group), the lens position encoder (which collects and feeds back the lens position information), and the lens mechanical structure (which fixes and supports the components to ensure stable movement). These components work together to achieve precise focal length adjustment.

[0058] The laser ranging module uses a combination of a pulsed laser emitter and an avalanche photodiode receiver to output distance measurements between the high-speed zoom lens and the target components on the power transmission line to be inspected at a fixed frequency. The vision sensor uses a global shutter CMOS image sensor, calculating image edge gradients in real time as sharpness feature data using the Sobel operator. The main control module synchronously receives distance measurements and sharpness feature data via an SPI interface, performs data fusion using a Kalman filter algorithm, generates an initial focus parameter matrix containing zoom step size and direction, and controls the high-speed zoom lens to adjust its focus based on this matrix. The target analysis module has a built-in SQLite database storing priority weights for different component types, matching the target component's corresponding focus strategy instruction based on the GPS location information of the current inspection task. The strategy optimization module uses a genetic algorithm to perform multi-objective optimization on the initial focus parameter matrix, generating an optimized focus parameter matrix and feeding it back to the main control module. Finally, the main control module outputs a PWM control signal to drive the motor of the high-speed zoom lens.

[0059] In one specific embodiment, the laser ranging module uses a 905nm pulsed laser in conjunction with an avalanche photodiode receiver to output real-time distance measurements of target components on the transmission line at a fixed frequency of 100Hz. This laser ranging module incorporates a temperature compensation circuit to eliminate ranging errors caused by ambient temperature fluctuations. The distance data processing procedure is as follows: First, outlier filtering is performed: when the measurement deviation exceeds ±10cm within three consecutive sampling periods, a dynamic threshold correction mechanism is triggered, replacing the current outlier with the sliding average of the previous five valid data sets. A 5-megapixel global shutter CMOS image sensor is used as the vision sensor. Image acquisition is triggered synchronously with each laser ranging measurement, and the gradient variance of the image region is calculated using the Sobel edge detection algorithm as the sharpness feature data for quantitative evaluation. This vision sensor is equipped with an automatic exposure control loop that dynamically adjusts the integration time according to the ambient light intensity to maintain the image grayscale value within a linear response range of 80-160.

[0060] See Figure 2The main control module includes a data fusion unit, a parameter generation unit, and an execution control unit. The data fusion unit synchronously receives ranging data and sharpness feature data via dual-channel DMA (Direct Memory Access). The ranging data is first preprocessed with median filtering, with the filtering window set to 5 sampling points to eliminate interference from transient obstructions such as birds. The sharpness feature data is normalized to eliminate the influence of brightness differences in different scenes. The fusion process uses a weighted algorithm: a confidence weight of 0.7 is assigned to the ranging data, and a confidence weight of 0.3 is assigned to the sharpness feature data. A 12-dimensional fusion feature matrix is ​​generated using a Kalman filter model. Each column of this fusion feature matrix corresponds to a comprehensive confidence score for different spatial locations, and the time dimension covers the measurement sequence of the most recent 200ms.

[0061] The parameter generation unit stores 256 pre-calibrated focal length mapping relationship tables. Each focal length mapping relationship includes ideal focal length parameters corresponding to three input dimensions: target distance, ambient illumination, and lens temperature. The focal length mapping relationship table adopts a non-linear interpolation storage structure, dividing the range of 1-50 meters into query nodes with 0.5-meter accuracy. After fusing the feature matrix input, the nearest neighbor search algorithm matches the closest environmental state parameter group in the focal length mapping relationship table, outputting an initial focusing parameter matrix containing 1024 discrete focusing positions. This initial focusing parameter matrix adopts a circular buffer storage structure, with row indices representing the execution timing and column indices representing the lens movement direction and step size parameters.

[0062] The execution control unit converts the initial focus parameter matrix into physical drive signals. A microelectromechanical system (MEMS) drive unit drives the lens group of the high-speed zoom lens to achieve rapid and precise focus adjustment. The MEMS drive unit is a hardware module independent of the high-speed zoom lens and controlled by the execution control unit of the main control module. It converts the focus parameters output by the main control module into mechanical actions, driving the high-speed zoom lens to complete focus changes. The MEMS drive unit uses a MEMS voice coil motor, and the control circuit converts the focus parameters into a 0-5V analog voltage signal with a voltage resolution of 2.44mV. The control process uses a closed-loop PID algorithm, reading the lens position encoder feedback signal at 100µs intervals for position correction. When the target distance is within 10 meters, the system drives the lens to complete the full-stroke focusing action within 200ms; at a distance of 50 meters, the focusing time is extended to 500ms to avoid mechanical overshoot. The lens motion trajectory planning uses an S-shaped velocity curve algorithm, with acceleration linearly increasing from 0 to a peak of 2m / s². 2 Then, it decreases linearly as it approaches the target position.

[0063] The target analysis module loads a priority database in SQLite format upon startup. This database contains a component type index table and a task strategy mapping table. The component type index table uses string hash values ​​as component identifiers, associated with numerical weight coefficients. The task strategy mapping table uses the task ID as the primary key and stores preset focusing accuracy indicators and execution order rules. Upon receiving a detection task instruction, the target analysis module first parses the task type identifier and matches the corresponding component priority weight from the target priority database. For example, for the "insulator string crack detection" task, the task type identifier corresponds to the preset weight configuration: insulator string weight is set to 0.8, conductor suspension point weight to 0.15, and equalizing ring weight to 0.05. Then, based on the component priority weights, the spatial distribution of the target components is sorted to generate a target focusing sequence. The spatial sorting uses a geometric optimization algorithm, with the UAV's current hovering position as the reference origin, to calculate the three-dimensional vector from the center point of each component to the lens. The priority score for each component is determined by the product of the reciprocal of the geometric distance and the weight coefficient, ultimately generating the target focusing sequence in descending order of score. Finally, based on the target focusing sequence, a focusing strategy instruction containing the priority execution order is determined.

[0064] The target focusing sequence is encoded into structured instructions: the instruction header contains a 16-bit checksum, and the subsequent payload area is divided into time slices at 0.1-second intervals. Each time slice stores three bytes of data, with the highest bit indicating the action type and the lower fifteen bits storing the target component code. The execution controller extracts the component code in time slice order and obtains its spatial coordinates in the target area by looking up a table. When the target component is an insulator string, the sub-region scanning mode is automatically activated: based on the global coordinates, five auxiliary focusing points are generated at 0.2-meter intervals along the vertical direction. An overtime protection parameter is added to the end of the instruction: if the focusing action of a single component lasts for more than 1.5 seconds without achieving the preset sharpness, the system is forced to switch to the next priority target. A state lock mechanism is introduced during the entire process to ensure that the focusing strategy instruction generation is not interfered with by new ranging data.

[0065] The execution controller is the control component in the laser-assisted focusing system responsible for executing the focusing strategy instructions. It is not an auxiliary component of a single device. Its core function is to extract the target component code from the focusing strategy instructions, determine the spatial coordinates of the component by querying preset data, and control the high-speed zoom lens to complete the focusing action on the target component in sequence according to the instructions.

[0066] See Figure 3 The strategy optimization module includes:

[0067] The strategy parsing unit is used to extract the priority execution order in the focus strategy instructions and generate strategy constraints by combining sharpness feature data.

[0068] The matrix correction unit is used to adjust the focal length adjustment range in the initial focus parameter matrix according to the strategy constraints, and generate the optimized focus parameter matrix.

[0069] The iterative feedback unit is used to send the optimized focus parameter matrix to the execution control unit of the main control module.

[0070] During system operation, the strategy optimization module continuously monitors changes in environmental parameters. When vibration interference with wind speed exceeding 12 m / s is detected, a rearrangement command is automatically sent to the target analysis module, triggering a dynamic update of the target focusing sequence. During the sequence execution phase (the phase where the system performs focusing operations based on the target focusing sequence generated by the target analysis module), the strategy optimization module collects the actual lens position information every 50 ms and generates a statistical report on the deviation between the execution trajectory and the command requirements. If the deviation at three consecutive detection points exceeds 5% of the focal length value, the system automatically switches to safety mode: stopping the current sequence execution, re-measuring the target distance, updating the initial focusing parameter matrix, and then continuing operation.

[0071] The main control module and the strategy optimization module use redundant communication channels. The main channel uses RS485 differential transmission of control commands, while the alternative channel transmits key status flags through PWM pulses to ensure control reliability in electromagnetic interference environments.

[0072] In one specific embodiment, after receiving the focusing strategy instruction in JSON format, the strategy parsing unit first performs syntax verification and semantic parsing. The priority execution order field in the focusing strategy instruction is converted into state machine control parameters: eight state registers are initialized, each corresponding to the operation state of a component type. When executing the "conductor-insulator-fitting" sequence, the state machine sets the priority flag bit according to binary encoding. After the sharpness feature data is input, the system automatically selects the processing channel according to the current ambient light level: under strong light conditions above 50000 Lux, the overexposure suppression algorithm is activated, and the saturation component of the HSV color space is extracted as an auxiliary sharpness indicator; below 1000 Lux, the digital gain compensation module is activated to perform adaptive histogram equalization on the brightness channel. The strategy constraints generated after data processing are stored in a structured manner, containing three sets of key parameters: the allowable focal length error range is stored as a percentage value of ±0.3%; the maximum focusing time is accurate to the 5ms level in milliseconds; and the energy consumption limit index is converted into the maximum allowable current value of the motor.

[0073] The matrix correction unit runs a quadratic programming solver, loading a priority weighting factor onto the initial focusing parameter matrix. This priority weighting factor is dynamically set according to the target component type: 1.0 for conductors, 1.2 for insulators, and 0.8 for small hardware such as vibration dampers. The constraint injection process uses the Lagrange multiplier method to establish a three-dimensional optimization surface in the solution space. The focus adjustment amplitude correction algorithm consists of two stages: the preliminary correction stage uses the gradient descent method to solve for local optima, with accuracy controlled within 2% of the target focus length; the fine correction stage switches to the conjugate direction method, improving the accuracy to 0.5% through five iterations. During the solution process, the processor temperature of the matrix correction unit in the strategy optimization module of the laser-assisted focusing system is monitored in real time. When the temperature exceeds 85℃, the number of iterations is automatically reduced, and a lookup table method is used instead of real-time calculation.

[0074] The iterative feedback unit employs a dual-channel communication mechanism. The forward transmission channel uses a CRC-checked 485 bus protocol, sending a 128-byte optimized focus parameter matrix frame every 50ms. The matrix data structure uses a timestamp alignment mechanism, with the frame header recording a microsecond-level time stamp from the system clock. The reverse feedback channel receives lens position encoder data via GPIO interrupts. The lens position encoder uses a 17-bit absolute value encoding scheme, achieving a position resolution of 0.001 degrees. The feedback processing module establishes two circular buffer queues: the input queue buffers the latest 20 sets of encoder sample values, and the output queue stores pending drive instructions. A data synchronization mechanism is implemented between the two queues; when three consecutive feedback data deviations are detected, a retransmission protocol is initiated to update the forward channel data. The feedback processing module is a hardware module independent of the iterative feedback unit.

[0075] In one embodiment, the laser-assisted focusing system of this application further includes:

[0076] The feedback calibration module is used to obtain the actual sharpness assessment value collected by the visual sensor after the high-speed zoom lens performs a focal length adjustment.

[0077] The comparison unit is used to calculate the deviation between the actual sharpness assessment value and the target sharpness threshold, and to trigger a calibration command when the deviation exceeds the preset tolerance range.

[0078] The parameter readjustment unit is used to respond to calibration commands, generate a calibrated focus parameter matrix based on the optimized focus parameter matrix after deviation correction, and input it to the main control module.

[0079] See Figure 4 The feedback calibration module can perform the following operations:

[0080] Construct a sharpness distribution heatmap based on the actual sharpness assessment values;

[0081] Extract the coordinates of low-resolution regions below the target resolution threshold from the resolution distribution heatmap;

[0082] The coordinates of the low-resolution area are matched with the spatial distribution of the target component, generating an area calibration command and sending it to the parameter readjustment unit.

[0083] The feedback calibration module activates 300ms after the lens completes focusing, ensuring complete attenuation of mechanical vibration. Actual sharpness assessment values ​​are acquired using a dynamic region selection strategy: conductor targets are divided into 5-meter equal-length detection segments; insulator targets are partitioned using a nine-square grid. The evaluation algorithm employs an improved Tenengrad operator, calculating the edge gradients of the luminance and chrominance components in the YUV color space, and merging the scoring results with weights of 70% and 30%, respectively. The scoring output is normalized to the 0-1.0 range, retaining two decimal places of precision. The comparison unit uses a two-level judgment logic: the main judge calculates the absolute difference between the actual score and the target threshold; the auxiliary judge calculates the coefficient of variation of the three frames of scoring data. When the insulator target threshold is set to 0.85, the absolute difference tolerance is set to ±0.04; when the conductor target threshold is 0.75, the tolerance is extended to ±0.06. A calibration command is triggered only if two consecutive frames exceed the threshold.

[0084] See Figure 5 The parameter readjustment unit performs the following operations:

[0085] Locate the identifier of the target component to be calibrated according to the area calibration instructions;

[0086] Calculate the focal length compensation value corresponding to the target component identifier based on the deviation amount;

[0087] Update the focal length compensation value in the optimized focus parameter matrix to generate the calibrated focus parameter matrix.

[0088] The parameter readjustment unit employs a multi-level response mechanism. The first-level response handles simple offset cases: when the deviation is within the 20% threshold range, the focal length compensation value is calculated directly according to a linear ratio, with compensation accuracy controlled at the micrometer level. The second-level response handles complex scenarios: for cases with vibration interference, the spectrum analysis module is activated to obtain the current dominant mechanical vibration frequency, and a phase offset correction term is added to the compensation value. The third-level response handles persistent defocusing: if the target is not met after three consecutive compensations, the lens calibration mode is automatically invoked, driving the lens to perform a full-stroke zeroing operation and re-initializing the focusing process. The focal length compensation value uses an incremental update method, marking the parameter indices that need modification in the optimized focusing parameter matrix, and only replacing data points with changes greater than 0.1%. Every ten compensation operations, the system automatically generates parameter change records and stores them in non-volatile memory for subsequent strategy optimization reference.

[0089] The calibration process incorporates multiple protection mechanisms, including a current monitoring circuit, a temperature sensor, and a motion trajectory monitoring module. The current monitoring circuit monitors the motor drive current waveform in real time; if a sudden current surge exceeds a set threshold, the calibration process is immediately paused for short-circuit protection checks. The temperature sensor samples the motor radiator temperature 100 times per second; if the temperature exceeds 90°C, the drive voltage is automatically reduced for thermal protection. The motion trajectory monitoring module uses dual-channel detection with a photoelectric encoder and a Hall sensor; an emergency stop is triggered if the difference between the two channels exceeds 0.5%. Calibration result verification requires two complete focusing cycles: after the initial focusing, sharpness feature data is collected (the actual sharpness assessment value collected by the calibration module via a visual sensor after focus adjustment on the high-speed zoom lens, compared with the target sharpness threshold to determine if a calibration command needs to be triggered), and a second focusing test is performed after fine-tuning the compensation value. The overall calibration process is limited to within 800ms; if it is not completed within this time, the current calibration sequence is skipped. The system maintains a calibration status counter; after more than ten calibration operations are triggered in a single day, a maintenance warning report is automatically generated and uploaded to the data center.

[0090] In one specific embodiment, the feedback calibration module is activated after the high-speed zoom lens performs focus adjustment, with a delay time set to 1.8 times the lens's mechanical vibration decay period. The visual sensor switches to high-resolution mode, acquiring image sequences of the target area at a rate of 30 frames per second. The calculation of the actual sharpness evaluation value employs an adaptive windowing technique: for conductor-type targets, a sliding detection band is established along the direction, with a bandwidth 1.3 times the actual diameter; for insulator-type targets, gridded analysis is enabled, dividing the area into 5×5 unit detection zones. The evaluation algorithm integrates spatial frequency analysis and edge sharpness as dual indicators, processing the lightness and chroma channels separately in the CIELAB color space. The lightness channel uses the Laplacian operator to calculate the second derivative response, and the chroma channel detects the chromaticity gradient in the ab plane. The final score is calculated using the following dual-channel weighted fusion:

[0091] ;

[0092] Where: S represents the overall sharpness score, S luminance S represents the sharpness value of the brightness channel. chroma This represents the sharpness value of the chroma channel. The scoring result is quantized on an integer scale of 0-100 to eliminate floating-point calculation errors.

[0093] A sharpness distribution heatmap is generated using polar coordinates. A coordinate system is established with the geometric center of the target component as the origin, with radial distances divided into 0.1-meter accuracy intervals and angular directions divided into 0.5° resolution intervals. The value of each cell in the sharpness distribution heatmap is determined by a kernel density estimation algorithm, which uses spatial interpolation based on a three-dimensional Gaussian function.

[0094] ;

[0095] in: This indicates the weight value for the contribution of sharpness. Represents angular coordinates (unit: degrees). Indicates radial distance (unit: meters). Represents the angular difference component. Represents the distance difference component. This is the smoothing coefficient parameter (default value 0.1).

[0096] The sharpness distribution heatmap is rendered as a 512×512 pixel matrix, and the sharpness values ​​are mapped using the HSL color model: 100 points corresponds to cyan (H=180°), and 0 points corresponds to dark red (H=0°).

[0097] Low-resolution area detection employs a dynamic discrimination mechanism. A baseline threshold is also included. Dynamically configured based on component type: Wire connector area: Insulator skirt area: Surface of the vibration damper: .

[0098] The discrimination process includes four steps: The first step is to scan the sharpness distribution heatmap and find that three consecutive cells have a score lower than the target score. The first step uses the region as the seed point; the second step performs eight-neighborhood region growing centered on the seed point and merges similar scoring units; the third step calculates the geometric features of the region and filters out noise regions with an area less than 10 pixels; the fourth step optimizes the boundaries of the effective region and removes jagged edges. The final output region is represented by the minimum bounding rectangle, and its center coordinates are recorded. With dimensional parameters .

[0099] The coordinate matching process establishes a spatial mapping model. The tilt angle of the current lens optical axis is obtained through a tilt sensor mounted on the gimbal. and yaw angle The coordinate converter will transform the sharpness distribution heatmap into polar coordinates. Convert to absolute space coordinates: convert angles Superimposed gimbal yaw angle Obtain the absolute azimuth angle; convert the radial distance... Combined with the camera's tilt angle Converted to horizontal projected distance; real-time UAV positioning data is used as the origin of the world coordinate system.

[0100] Transformed spatial coordinates The input component matching engine stores a library of CAD models of standard components. The matching process employs a three-level screening process: the first level filters candidate components with an axial distance error of less than 0.3 meters; the second level compares component contour feature points; and the third level performs texture template matching. When the matching confidence exceeds 90%, the component identifier and surface position code are output.

[0101] The regional calibration instructions employ an optimized coding scheme. The instruction header contains a 32-bit timestamp and a 16-bit location check code, while the data area uses a structure array to store the calibration region parameters. Each region record includes: component type code (4 bits): classification of conductor / insulator / fitting; surface partition code (6 bits): accurate to a 2cm×2cm grid; sharpness deviation value (8 bits): storing the difference between the actual score and the target score; and calibration priority (3 bits): indicating the urgency of processing.

[0102] Calibration commands are transmitted to the policy optimization module via fiber optic channel, using a time-sensitive network standard to ensure delivery within 2ms. A CRC-32 checksum is appended to the data packet, and the receiving end automatically implements an error retransmission mechanism. The command buffer is designed as a depth-first queue structure; when multiple calibration areas appear simultaneously, the system automatically merges requests based on spatial proximity.

[0103] The feedback calibration module also includes an environmental compensation subsystem for real-time correction of measurement errors. This subsystem includes temperature and humidity sensors, a triaxial accelerometer, and a vibration compensator. The temperature and humidity sensors sample environmental data 10 times per second to establish an atmospheric refractive index correction model. :

[0104] ;

[0105] in: This is the air pressure value (unit: hectopascals). This is the temperature value (unit: degrees Celsius). This refractive index parameter is used to correct the optical path calculation for laser ranging. The vibration compensator receives data from the triaxial accelerometer and uses an adaptive filtering algorithm to separate mechanical vibration noise. When the dominant oscillation frequency is detected to be within the range of 8-12Hz, the motion blur compensation algorithm is automatically activated to adjust the exposure synchronization strategy of the vision sensor.

[0106] In this embodiment, a closed-loop verification mechanism is established for calibration quality monitoring. After each calibration command is executed, the system performs a 500ms retest verification. During the retest, three consecutive frames of sharpness data are collected, and the median value is taken as the final score after removing the maximum and minimum values. The calibration results are recorded in non-volatile memory, containing three data items: the original score, compensation parameters, and the post-calibration score. The monitoring terminal calculates the calibration success rate for each type of component and periodically generates calibration performance reports for system maintenance reference. When a specific component type fails five consecutive calibrations, an expert diagnostic mode is automatically triggered, freezing the system status and awaiting manual intervention.

[0107] See Figure 5 The target component identifier localization employs multi-feature fusion recognition technology. When the region calibration command is passed to the parameter readjustment unit, the system first activates the high-definition auxiliary camera mounted on the gimbal to acquire images of the target area at a 12-megapixel resolution. The acquired target area images are then preprocessed, including geometric distortion correction and dynamic range compression to eliminate lens barrel distortion and ambient halo interference. Feature extraction is then performed on the preprocessed images, simultaneously running three algorithms: extracting at least 64 scale-invariant feature points using a SIFT-based keypoint detector; calculating the texture encoding of an 8×8 pixel block using a local binary pattern analyzer; and outputting a 128-dimensional semantic feature vector through a deep convolutional network. This provides high-quality image data for accurate target component identifier localization and is a pre-processing step. The component matching engine compares these features with templates in the component CAD model library and uses a hierarchical filtering strategy to accurately locate the target component. This strategy describes the specific operation method of feature comparison performed by the component matching engine during the target component identifier localization process. Its purpose is to improve the accuracy and efficiency of component matching through multiple rounds of screening. The specific process is as follows: the first level of screening is based on axial distance, selecting candidate components whose distance error meets the requirements; the second level compares the contour feature points of the candidate components to further narrow down the range; the third level performs texture template matching to finally determine the target component identifier. This strategy is a specific implementation method based on multi-feature fusion recognition technology, ensuring accurate location of the target component even in complex environments. See Table 1.

[0108] Table 1: Component Feature Matching Parameter Table.

[0109] .

[0110] The matching process continuously optimizes the search strategy: when the confidence level is at the critical value (80-85%), the system automatically adds three more frames of images for multi-temporal verification; for components with repetitive structures such as insulators, the spatial periodic analysis mode is enabled to detect the distribution pattern of feature points. The identifier output after a successful match contains a 16-bit code: the first 4 bits represent the category (e.g., "1000" represents a conductor joint), the middle 8 bits identify the specific model, and the last 4 bits record the surface partition number.

[0111] The calculation of focal length compensation incorporates an environmental compensation mechanism. The system maintains a real-time updated environmental parameter table, including current temperature, humidity, air pressure, and light intensity readings. The compensation algorithm first queries this environmental parameter table to obtain the latest data: when the temperature exceeds 40℃, the lens thermal expansion compensation coefficient is activated; when the humidity is greater than 80%, the light refraction compensation mode is enabled. The calculation process uses an iterative approximation method: the initial compensation value is based on calibration data under standard atmospheric conditions, and is subsequently adjusted according to the environmental parameters. The compensation calculation for wire connectors considers the influence of the surface oxide layer: the compensation amount is reduced by 5% for newly installed connectors and increased by 8% for connectors that have been in operation for more than five years. The compensation value is output in the format of a signed 16-bit integer, in units of 0.1 micrometers; positive numbers indicate forward focusing, and negative numbers indicate backward focusing.

[0112] The matrix update operation employs differential synchronization technology. The optimized focus parameter matrix is ​​stored in dual-port RAM, allowing read and write operations to be executed in parallel. The update process follows a strict transaction processing flow: first, the matrix block to be modified is locked and marked as "editing"; then, the new focus parameters are written, while a backup of the original data is retained; finally, the CRC checksum of the written data is verified, and the block is unlocked after confirmation. For multi-component combined targets such as insulator strings, the system implements a batch update function: after selecting an insulator unit, its adjacent units are automatically identified, and the same compensation strategy is applied to synchronously update the positions of five matrices. Each update operation records a detailed log, including a timestamp, operator ID, value before modification, and value after modification.

[0113] In one specific embodiment, the component type index table adopts a relational database structure. The core table contains five fields: type_id stores the unique identifier of the component generated using the IEEE standard hash algorithm; priority_coef defines a priority coefficient ranging from 0.1 to 1.0; defect_sensitivity records the component's sensitivity level to imaging blur, divided into levels 1-5; material_type indicates the material category (e.g., ceramic, steel-cored aluminum, etc.); and age_factor represents the impact coefficient of service life. The system loads this component type index table into memory each time it starts, and builds a B+ tree index to accelerate queries. A typical priority configuration is as follows: the priority_coef of insulator strings for newly built transmission lines is set to 0.9, decreasing to 0.8 after ten years of operation; the initial value for conductor joints is 0.7, increasing to 0.85 when slight corona discharge is detected.

[0114] The task strategy mapping table employs a dynamic loading mechanism. This table, stored in an encrypted configuration file, contains three core fields: task identifier, focus mode, and accuracy level. When the UAV receives a new inspection task, the control terminal sends an encrypted command packet containing the task identifier. After decryption, the system queries the task strategy mapping table to obtain the corresponding configuration parameters: the insulator contamination inspection task is associated with "high-precision mode," requiring a lens dwell time of no less than 800ms; the conductor damage inspection task uses "fast scan mode," limiting single-point focusing time to within 300ms. The table also defines anomaly handling strategies, such as automatically switching to a backup focus algorithm when the image signal-to-noise ratio is below 35dB.

[0115] In one specific embodiment, the parameter readjustment unit also integrates a fault self-diagnosis function. Before each calibration operation, the system automatically runs a pre-check procedure: checking whether the motor drive current is within the normal range of 5-20mA; verifying that the standard deviation of 10 consecutive samples of the lens position encoder reading is less than 3 minimum resolution units; and confirming that the temperature sensor reading is within the permissible range of -20℃ to 60℃. Upon detecting an anomaly, the unit enters a safety mode: disconnecting the motor power supply, saving a snapshot of the current state, and sending a fault code via a wireless channel. The fault code uses hexadecimal encoding, with the first digit indicating the fault category (e.g., "B" represents a mechanical fault), and the last three digits identifying the specific problem (e.g., "B12A" indicates abnormal lens guide resistance). Maintenance personnel can read detailed diagnostic logs through a dedicated interface, including system parameter change curves for the 30 seconds prior to the fault occurrence.

[0116] In one specific embodiment, the system also includes a historical data analysis module for periodically optimizing calibration strategies. The system stores a complete record of the most recent 1000 calibration operations, including environmental conditions, component type, compensation value, and final sharpness score. An analysis program runs automatically weekly: identifying component types with consistently below-average compensation performance; discovering calibration failure patterns under specific environmental conditions (such as morning backlight); and statistically analyzing the time distribution of different focusing strategies. The analysis results are used to dynamically adjust parameters in the component type index table: for example, when the calibration success rate of a certain type of fitting drops by more than 15% during the rainy season, the system automatically increases its priority_coef by 0.1 temporarily; if a certain insulator model is detected to perform well under high-temperature conditions, its compensation is reduced by 10%.

[0117] In one specific embodiment, the acquisition of target focusing accuracy parameters begins with the parsing of the task identifier. The detection task command sent by the ground control station is encapsulated using the AES-256 encryption protocol, and the system decrypts and verifies it through a hardware security module. The task identifier is a 16-bit code; the first 3 bits represent the power grid voltage level, the middle 7 bits specify the inspection area number, and the last 6 bits define the specific detection item type. The process of querying the task strategy mapping table after parsing uses preprocessed SQL statements to establish a read-only cursor to traverse the memory-mapped file. For insulator contamination detection tasks with identifiers ending in "01", the system extracts the associated high-precision parameter set: this parameter set defines a focal length repeatability accuracy requirement of ±0.003 mm and a spatial resolution requirement of 0.05 mm per pixel. The conductor sag measurement task identifier ends in "05", and the associated medium-precision parameter requirement is ±0.01 mm positioning accuracy.

[0118] In one specific embodiment, the parameter adjustment of the laser ranging module is achieved by modifying the internal registers of the FPGA. After acquiring the target focusing accuracy parameters, the clock divider register is reconfigured according to the mode code: in defect recognition mode, the system writes a division coefficient K=8, increasing the laser emission frequency from the base 100Hz to 200Hz; in texture acquisition mode, the division coefficient is set to K=32, and the frequency is reduced to 50Hz. When the system adjusts the sampling frequency of the laser ranging module according to the target focusing accuracy parameters (e.g., increasing the frequency in defect recognition mode and decreasing the frequency in texture acquisition mode), the temperature compensation module inside the laser ranging module will start simultaneously with the sampling frequency adjustment operation. It no longer works only according to a fixed cycle, but synchronously adjusts the temperature data acquisition interval according to the new sampling frequency (e.g., shortening the acquisition interval when the frequency increases), ensuring that the influence of temperature fluctuations on the ranging value can be corrected in a timely manner under different sampling frequencies, thus guaranteeing the ranging accuracy. The platinum resistance sensor inside the module acquires temperature data every 100ms and automatically corrects the ranging value by looking up the temperature-drift compensation table. When the temperature exceeds 45°C, the system additionally activates a secondary compensation strategy, shortening the sampling interval to 50ms and taking the median of three measurements as the output value.

[0119] In one specific embodiment, the exposure control of the vision sensor employs a multi-channel collaborative mechanism. The system's main controller configures the row exposure register of the vision sensor via the I2C bus: first, it reads ambient light sensor data; if the illuminance exceeds 80,000 lux, it activates a strong light suppression mode, setting the minimum exposure time to 37 μs; in twilight conditions below 1000 lux, it automatically enables hardware gain compensation, extending the exposure time to 8 ms. Exposure parameter generation follows a non-linear curve: within the medium illuminance range of 2000-50000 lux, the optimal exposure time is calculated using a logarithmic function of the illuminance value. White balance parameters are adjusted synchronously: an automatic white balance lock function is enabled, and a reference white board is sampled three seconds before detection begins, stabilizing the color temperature within the range of 5500K±200K.

[0120] In one specific embodiment, the convergence termination condition of the optimized focus parameter matrix adopts a dual-criteria design. The first criterion is an absolute accuracy limit: the condition is met when the absolute deviation between the measured focal length and the target value is less than 0.5% of the focal length value. The second criterion is a relative stability criterion: stability is determined when the focal length fluctuation of five consecutive frames is less than 0.05% of the focal length value. The system calculates the dual-criteria status flag in real time: a 16-bit status register is set, with bits 0-7 storing the current deviation percentage value, bits 8-12 recording the number of consecutive stable frames, and bit 15 being the convergence achievement flag. This register is refreshed every 10ms, and a status change triggers a hardware interrupt signal.

[0121] In one specific embodiment, the strategy optimization module also includes a real-time matching degree analysis module, which runs a dynamic time warping algorithm. The microcontroller of the real-time matching degree analysis module collects actual motion trajectory data point sequences from the lens position encoder and calculates their similarity with the expected trajectory of the optimized focus parameter matrix. The sequence alignment process uses sliding window matching: a 200ms observation window is established, and the morphological similarity between the two sequences within the window is calculated. The matching result is quantized into integer values ​​from 0 to 100: when the real-time score value is greater than 95 for 10 consecutive sampling periods, the motion trajectory is considered to meet expectations. When a sudden mismatch is detected, the system automatically switches to emergency control mode: the current control parameters are frozen, laser ranging and image feature acquisition are re-performed, the focus parameter matrix is ​​updated, and control is resumed.

[0122] In one specific embodiment, the drive termination control of the high-speed zoom lens employs a multi-level protection strategy. The main control processor, dedicated to drive termination control, sends a three-stage stop command to the motor driver after the convergence condition is met (controlling drive voltage reduction, maintaining position closed-loop, and disconnecting the drive circuit): First, the drive voltage is reduced to 70% of the maintenance voltage within 2ms via a PWM signal, and then maintained for 10ms to allow mechanical resonance to decay; the second stage cuts off power output while maintaining position closed-loop control; finally, before completely disconnecting the drive circuit, a position lock verification is performed: the lens position encoder value is read three times consecutively, and after confirming that the offset is less than 5 minimum resolution units, the electromagnetic brake (a safety locking component of the high-speed zoom lens, integrated into the mechanical structure of the lens) is activated. Simultaneously, the temperature monitoring system monitors the temperature of the driver's heatsink. If the temperature exceeds the limit, an emergency stop operation is performed. This system is only responsible for drive termination and temperature protection after focus adjustment is completed, and does not participate in the control of initial focus parameter generation or focus adjustment. Temperature monitoring is involved in the entire control process: when the motor heatsink temperature exceeds 85℃, a forced cooling program is immediately initiated; when it reaches 90℃, an emergency stop is performed regardless of whether convergence has occurred.

[0123] In one specific embodiment, the data logging mechanism employs a circular buffer and flash memory working in tandem. The system allocates a 256KB memory buffer to store complete logs of the most recent 200 focusing operations, including the initial parameter matrix, actual motion trajectory, environmental parameters, and termination condition trigger status. Whenever the buffer is full, the processor initiates an encryption and compression program: compressing the data using the LZ77 algorithm and encrypting it using elliptic curve cryptography before writing it to the industrial-grade flash memory chip. The flash memory area is divided into 100 logical sectors, with each sector storing 50 complete records. Under favorable wireless communication conditions, the system automatically selects idle periods to upload historical data to the central server. Before each focusing program is initiated, the system automatically checks the remaining flash memory space; if it falls below 10%, a maintenance alarm is generated and sent to the ground control station.

[0124] In one specific embodiment, the anomaly recovery process of this application system also establishes a hierarchical processing mechanism. When the convergence process times out (the preset time threshold is 800ms), the system enters Level 1 recovery: saving a snapshot of the current state and reverting to the initial focus parameter matrix for re-execution. If recovery fails three times consecutively, Level 2 processing is activated: completely resetting the laser ranging module and vision sensor driver, and recalibrating the lens zero position. The most severe mechanical jamming triggers Level 3 processing: cutting off the power supply to the motor driver, activating the backup piezoelectric ceramic fine-tuning mechanism (the backup position adjustment component of the laser-assisted focusing system) to maintain basic functions, and simultaneously sending fault diagnosis data through the emergency communication channel.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A laser-assisted focusing system for a power transmission line inspection camera, used to control the high-speed zoom lens of a power transmission line inspection drone for automatic focusing, characterized in that: include: The laser ranging module is used to acquire real-time distance measurements between the high-speed zoom lens and the target components to be inspected on the power transmission line. A visual sensor is used to simultaneously acquire sharpness feature data of the target component; The main control module receives distance measurement values ​​and sharpness feature data, generates an initial focus parameter matrix through fusion processing, and controls the high-speed zoom lens to perform focus adjustment based on the initial focus parameter matrix. The target analysis module is used to call the built-in target priority database and determine the focusing strategy instructions for the target components based on the current detection task. The strategy optimization module is used to modify the initial focus parameter matrix according to the focus strategy instructions, generate the optimized focus parameter matrix and feed it back to the main control module. Finally, the main control module outputs the control signal to drive the high-speed zoom lens according to the focus parameter matrix.

2. The laser-assisted focusing system for a power transmission line inspection camera according to claim 1, characterized in that: The main control module includes: The data fusion unit is used to weightedly fuse distance measurements and sharpness feature data, and output a fused feature matrix. The parameter generation unit is used to query a preset focal length mapping relationship table based on the fusion feature matrix to generate an initial focusing parameter matrix containing the focal length adjustment range. The execution control unit is used to drive the microelectromechanical system drive unit inside the high-speed zoom lens based on the initial focus parameter matrix, thereby controlling the focal length change of the high-speed zoom lens.

3. The laser-assisted focusing system for a power transmission line inspection camera according to claim 1, characterized in that: The target analysis module performs the following operations: Parse the task type identifier of the current detection task and match the corresponding component priority weight from the target priority database; Based on component priority weights, the spatial distribution of target components is sorted to generate a target focusing sequence; Based on the target focusing sequence, determine the focusing strategy instructions that include the priority execution order.

4. The laser-assisted focusing system for a power transmission line inspection camera according to claim 2, characterized in that: The strategy optimization module includes: The strategy parsing unit is used to extract the priority execution order in the focus strategy instructions and generate strategy constraints by combining sharpness feature data. The matrix correction unit is used to adjust the focal length adjustment range in the initial focus parameter matrix according to the strategy constraints, and generate the optimized focus parameter matrix. The iterative feedback unit is used to send the optimized focus parameter matrix to the execution control unit of the main control module.

5. A laser-assisted focusing system for a power transmission line inspection camera according to any one of claims 1-4, characterized in that: Also includes: The feedback calibration module is used to obtain the actual sharpness assessment value collected by the visual sensor after the high-speed zoom lens performs a focal length adjustment. The comparison unit is used to calculate the deviation between the actual sharpness assessment value and the target sharpness threshold, and to trigger a calibration command when the deviation exceeds the preset tolerance range. The parameter readjustment unit is used to respond to calibration commands, generate a calibrated focus parameter matrix based on the optimized focus parameter matrix after deviation correction, and input it to the main control module.

6. The laser-assisted focusing system for a power transmission line inspection camera according to claim 5, characterized in that: The feedback calibration module performs the following operations: Construct a sharpness distribution heatmap based on the actual sharpness assessment values; Extract the coordinates of low-resolution regions below the target resolution threshold from the resolution distribution heatmap; The coordinates of the low-resolution area are matched with the spatial distribution of the target component, generating an area calibration command and sending it to the parameter readjustment unit.

7. The laser-assisted focusing system for a power transmission line inspection camera according to claim 6, characterized in that: The parameter readjustment unit performs the following operations: Locate the identifier of the target component to be calibrated according to the regional calibration instructions; Calculate the focal length compensation value corresponding to the target component identifier based on the deviation amount; Update the focal length compensation value in the optimized focus parameter matrix to generate the calibrated focus parameter matrix.

8. The laser-assisted focusing system for a power transmission line inspection camera according to claim 1, characterized in that: The target priority database includes: A component type index table is used to associate the type identifiers and defect detection priority coefficients of transmission line hardware, conductors, and insulators. The task strategy mapping table is used to associate the task identifiers and corresponding focus accuracy parameters of defect identification tasks and texture acquisition tasks.

9. The laser-assisted focusing system for a power transmission line inspection camera according to claim 8, characterized in that: The target analysis module also performs: Query the task strategy mapping table based on the task identifier of the current detection task to obtain the target focus accuracy parameters; Adjust the sampling frequency of the laser ranging module and the exposure parameters of the vision sensor based on the target focusing accuracy parameters.

10. The laser-assisted focusing system for a transmission line inspection camera according to claim 9, characterized in that: The strategy optimization module also performs: The convergence termination condition of the optimized focus parameter matrix is ​​set according to the target focus accuracy parameters. During the focal length adjustment process of a high-speed zoom lens, the matching degree between the actual focal length change and the optimized focus parameter matrix is ​​compared in real time, and the control signal output is terminated when the convergence termination condition is met.

Images