A tunnel inspection robot cloud platform camera image anti-shake processing method

Abstract

The application discloses a kind of tunnel inspection robot cloud platform camera image anti-shake processing methods, it is related to computer processing technical field, the method comprises the following steps: obtaining the state data under the corresponding time stamp of the continuous image of cloud platform camera;According to the state data to the corresponding standard anti-shake processing parameter in database is matched to determine whether the current image anti-shake processing strategy needs to be adjusted;If it needs to be adjusted, then according to the anti-shake strategy table set in advance to determine the standard anti-shake processing parameter after correction, and the standard anti-shake processing parameter after correction is applied to image anti-shake processing.The application is by real-time analysis cloud platform camera's motion state and environmental illumination data, and intelligently dynamically adjusts anti-shake parameter, effectively suppresses the image jitter and exposure anomaly in tunnel inspection process, significantly improves picture stability and image quality.

Description

Technical Field

[0001] This invention relates to the field of computer processing technology, and in particular to a method for image stabilization processing of a gimbal camera on a tunnel inspection robot. Background Technology

[0002] The purpose of this tunnel inspection robot is to monitor high-speed tunnels to prevent delays in obtaining information and providing rescue after traffic accidents. Its detailed structure can be found in Chinese Patent No. CN109676625B, published on August 30, 2024, which discloses a tunnel inspection robot including a guide device, a mounting base, and an inspection device mounted on the mounting base. The inspection device includes a main control module and an execution module, with the main control module electrically connected to the execution module. The execution module includes a voice module, a communication module, a power supply module, a charging docking module, a motion module, a protection module, a monitoring module, a speed measurement module, and an auxiliary module. The inspection device can move along the guide device via the motion module. This application addresses the basic characteristics of current tunnels by adopting a side-mounted design with a guide rail. The tunnel robot operates along the guide rail, patrolling the tunnel continuously around the clock. Through various modules, it achieves real-time monitoring of the tunnel; captures evidence of illegal vehicles; facilitates rapid enforcement by relevant departments; and provides comprehensive monitoring of the tunnel surface and walls to prevent traffic accidents caused by tunnel factors.

[0003] The current technology is insufficient because the inspection robot moves back and forth within the tunnel along a track. During this process, factors such as friction between the wheels and the guide rails, vibration at the track joints, and the complex environment inside the tunnel (such as airflow and electromagnetic interference) can easily cause high-frequency, small-amplitude vibrations in the gimbal camera. These vibrations result in jitter, blurring, and misalignment in the captured continuous images, severely impacting image quality. Several methods exist for addressing image jitter, including: One approach is to use high-precision gimbals or optical image stabilization lenses at the hardware level. However, this significantly increases the manufacturing cost and maintenance complexity of the equipment, and its effectiveness is limited in harsh environments with continuous movement and vibration.

[0004] Secondly, at the software level, general electronic image stabilization algorithms are used (such as the video denoising method and electronic device disclosed in publication number WO2021135702A1, date of publication). However, these algorithms are mostly general designs and fail to fully consider the highly structured image content in tunnel inspection scenarios, such as regular tunnel outlines, road surface textures, and the special characteristics of jitter patterns, such as high-frequency linear jitter. The processing effect is unsatisfactory, and the fixed algorithm parameters are difficult to adapt to scenarios with dynamic changes in jitter levels, which may lead to over-cropping or insufficient stabilization. Summary of the Invention

[0005] To address the aforementioned technical problems, the present invention provides a method for image stabilization processing of a gimbal camera on a tunnel inspection robot, the method comprising the following steps: Acquire continuous images from the gimbal camera and the status data at the corresponding timestamps; Based on the status data, the corresponding standard image stabilization parameters are matched with those in the database to determine whether the current image stabilization strategy needs to be adjusted. If adjustments are needed, the corrected standard image stabilization parameters are determined according to the pre-set image stabilization strategy table, and then applied to the image stabilization process.

[0006] Preferably, the status data includes: Obtain the real-time sliding speed of the inspection robot along the guide rail at the corresponding timestamp; Obtain the previous frame's sliding speed from the real-time sliding speed; The speed difference of the gimbal camera is determined based on the difference between the real-time sliding speed and the sliding speed of the previous frame.

[0007] Preferably, the acquisition of continuous images from the gimbal camera and the status data at the corresponding timestamps further includes: The three-dimensional acceleration data of the vehicle sliding along the guide rail is collected in real time by a three-axis accelerometer deployed on the inspection robot. The three-dimensional acceleration data includes acceleration components in the longitudinal, lateral and vertical directions. The collected acceleration data is subjected to low-pass filtering to remove high-frequency noise and retain the low-frequency acceleration signal caused by the movement of the trolley. The low-frequency acceleration signal is compared with a preset acceleration threshold. If the acceleration in a certain direction exceeds the corresponding threshold, the velocity error correction mechanism is triggered. The speed error correction mechanism corrects the real-time sliding speed based on the integral value of the longitudinal acceleration. ; in, The sliding speed is directly measured by the sensor. Let t be the longitudinal acceleration, and t be the current time. t0 is the start time of exceeding the limit, and dt represents the change in time.

[0008] Preferably, the determination of whether the current image stabilization processing strategy needs to be adjusted includes: Based on the state data and the target stabilization capability value corresponding to the standard stabilization processing parameters, the current image jitter value is obtained and a judgment is made: If the jitter value is less than the first preset threshold, it is determined that there is no risk of jitter, and the standard anti-shake processing parameters are used as the corrected standard anti-shake processing parameters. If the jitter value is greater than or equal to the first preset threshold, the ratio of the state data to the target stabilization capability value corresponding to the standard stabilization processing parameters is calculated, and the stabilization processing strategy matching the database is determined based on the ratio result.

[0009] Preferably, the determination of whether the current image stabilization strategy needs to be adjusted is performed based on a machine learning model, and the steps include: Record the motion parameters and corresponding anti-shake processing logs of the inspection robot over the past 30 days. The motion parameters include multiple sliding speeds, accelerations, and guide rail positions within multiple window periods. The corresponding anti-shake processing logs include the anti-shake parameter adjustment time and the image shake score before and after adjustment. Based on the image quality assessment algorithm, historical footage from at least the past 30 days is labeled with jitter, and footage with a jitter score below the second threshold SSIM<0.8 is marked as a jitter event.

[0010] The system retrieves the current image and the status data at the corresponding timestamp, and obtains the sliding average speed over five seconds, the variance of acceleration, and the periodic fluctuation amplitude of the guide rail position. The predicted jitter probability within the next five seconds is determined based on the five-second sliding average speed, the variance of acceleration, and the periodic fluctuation amplitude of the guide rail position, where: If the predicted shake probability is >0.7 and the current image stabilization parameters are standard, then switch directly to high shake image stabilization parameters; If the predicted shake probability is >0.9, switch directly to ultra-high shake stabilization parameters and mark it as emergency stabilization mode.

[0011] Preferably, the acquired continuous images and the state data under the corresponding timestamps further include a high-sensitivity photosensor installed near the gimbal camera for real-time acquisition of ambient light intensity inside the tunnel, to obtain raw light intensity data, and the processing of the raw light intensity data includes: The original light intensity data is filtered using a moving average to remove high-frequency noise. Then, the mean ambient light intensity after filtering is calculated as a representative value for the current lighting conditions, where: The moving average filter has a window period of at least five seconds; Acquire a sequence of multiple images captured continuously in a dynamic scene; Extract the brightness information of each frame in the image sequence, and calculate the exposure evaluation index of the current scene based on the brightness information; Based on the deviation between the exposure evaluation index and the preset target exposure range, determine whether exposure correction needs to be initiated; If exposure correction needs to be activated, the corresponding exposure mapping function is selected according to the preset exposure correction parameter lookup table, and the selected mapping function is used to adjust the exposure of the subsequently acquired images in real time. The exposure correction parameter lookup table stores the correspondence between different exposure deviation ranges and exposure mapping functions.

[0012] Preferably, the step of extracting the brightness information of each frame in the image sequence and calculating the exposure evaluation index of the current scene based on the brightness information includes: Each frame of the image is sampled in sections to obtain the average brightness value of each section; Calculate the global image brightness average based on the average brightness value of all regions; The contrast index of the image is calculated based on the brightness values ​​of each region of the image. By combining the global image brightness average with the contrast index, a comprehensive exposure evaluation index is generated.

[0013] Preferably, the step of determining the corrected standard image stabilization parameters according to a pre-set image stabilization strategy table includes: If the ratio result is greater than a preset first ratio threshold and less than a preset second ratio threshold, then the corrected standard image stabilization parameters are determined to be high-jitter image stabilization parameters. If the ratio result is greater than or equal to the preset second ratio threshold, then the corrected standard image stabilization parameters are determined to be ultra-high shake stabilization parameters.

[0014] The present invention has at least the following beneficial effects: It achieves adaptive dynamic adjustment of image stabilization processing strategy. By acquiring continuous images from the gimbal camera in real time and the status data at the corresponding timestamps, and matching and judging them with standard parameters in the database, it can automatically and accurately switch to the most suitable image stabilization processing parameters according to the actual working conditions, which significantly improves the stability and reliability of image acquisition in complex tunnel environments; By combining speed difference, three-dimensional acceleration information and jitter probability prediction based on machine learning model, the system can not only identify the current jitter, but also predict the jitter risk in the future, thereby supporting the pre-adjustment of anti-shake parameters, reducing image delay and jitter processing lag, and enhancing the real-time performance and robustness of the system. Comprehensive anti-interference capabilities in variable tunnel environments. By introducing a photosensitive sensor to monitor changes in ambient light and combining image brightness analysis with exposure mapping adjustment, challenges such as uneven lighting and alternating light and dark areas within the tunnel are effectively overcome. Simultaneous image stabilization and exposure compensation are achieved, comprehensively improving image quality and providing a clearer and more stable image data foundation for subsequent tunnel surface defect identification and analysis. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart provided for Embodiment 1 of the present invention; Figure 2 This is a judgment block diagram provided in Embodiment 1 of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0019] Example 1

[0020] This embodiment provides a method for image stabilization processing of a tunnel inspection robot's gimbal camera, the method including the following steps, such as... Figure 1 and Figure 2 As shown: Acquire continuous images from the gimbal camera and the status data at the corresponding timestamps; Specifically, the aforementioned status data includes: Obtain the real-time sliding speed of the inspection robot along the guide rail at the corresponding timestamp; Get the previous frame's scrolling speed in real-time; The speed difference of the gimbal camera is determined by the difference between the real-time sliding speed and the sliding speed of the previous frame.

[0021] Secondly, in the above embodiments, obtaining continuous images from the gimbal camera and the corresponding status data at the timestamps also includes: The three-dimensional acceleration data of the vehicle sliding along the guide rail is collected in real time by a three-axis accelerometer deployed on the inspection robot. The three-dimensional acceleration data includes acceleration components in the longitudinal, lateral and vertical directions. The collected acceleration data is low-pass filtered to remove high-frequency noise and retain the low-frequency acceleration signal caused by the movement of the trolley. The low-frequency acceleration signal is compared with a preset acceleration threshold. If the acceleration in a certain direction exceeds the corresponding threshold, the velocity error correction mechanism is triggered. The speed error correction mechanism corrects the real-time sliding speed based on the integral value of the longitudinal acceleration. ; in, The sliding speed is directly measured by the sensor. Let t be the longitudinal acceleration, and t be the current time. t0 is the start time of exceeding the limit, and dt represents the change in time, that is, calculating the cumulative effect of acceleration on velocity from t0 to t.

[0022] The aforementioned technology continuously captures image sequences using a high frame rate mode on a gimbal camera, with each frame marked with a precise timestamp to ensure strict temporal alignment between image data and subsequent state data. Simultaneously, speed sensors (such as encoders or laser ranging modules) deployed on the inspection robot collect the robot's sliding speed along the guide rail in real time, forming a real-time sliding speed data stream. A caching mechanism stores the sliding speed value of the previous frame, and differential calculation is used to determine the speed difference between the real-time sliding speed and the previous frame's speed, yielding the gimbal camera's speed difference. This difference reflects abrupt changes in the robot's motion state. A three-axis accelerometer collects the robot's three-dimensional acceleration components in real time in the longitudinal (guide rail direction), lateral (horizontal to vertical guide rail direction), and vertical directions. The raw acceleration data undergoes low-pass filtering, using Butterworth or moving average filters to remove high-frequency noise and retain low-frequency motion signals, preventing mechanical vibration or environmental interference from affecting data accuracy. The filtered acceleration data is dynamically compared with a preset threshold: if the acceleration in any direction exceeds the threshold, a speed error correction mechanism is immediately triggered.

[0023] Secondly, the real-time sliding speed is corrected by integrating the longitudinal acceleration. Specifically, when the longitudinal acceleration exceeds the limit, the start time t0 of the exceedance is recorded, and the longitudinal acceleration is integrated over the time interval [t0, t] with the current time t as the end point of integration. The integration result is then superimposed on the sliding speed directly measured by the sensor to form the corrected real-time sliding speed. This process is achieved by the formula: Corrected speed = Sensor measured speed longitudinal acceleration The integral value reflects the cumulative effect of acceleration changes on velocity, effectively compensating for velocity measurement errors caused by sensor noise or sudden motion changes.

[0024] The system matches the status data with the corresponding standard image stabilization parameters in the database to determine whether the current image stabilization strategy needs to be adjusted. Specifically, the above-mentioned determination of whether the current image stabilization processing strategy needs to be adjusted includes: Based on the status data and the target image stabilization capability value corresponding to the standard image stabilization processing parameters, the current image jitter value is obtained and a judgment is made: If the jitter value is less than the first preset threshold, it is determined that there is no risk of jitter, and the standard image stabilization processing parameters are used as the corrected standard image stabilization processing parameters. If the jitter value is greater than or equal to the first preset threshold, the ratio of the state data to the target stabilization capability value corresponding to the standard stabilization processing parameters is calculated, and the stabilization processing strategy matched with it in the database is determined based on the ratio result.

[0025] It should be noted that the determination of whether to adjust the current image stabilization strategy is based on a machine learning model, and the steps include: Record the motion parameters and corresponding anti-shake processing logs of the inspection robot over the past 30 days. The motion parameters include multiple sliding speeds, accelerations and guide rail positions within multiple window periods. The corresponding anti-shake processing logs include the anti-shake parameter adjustment time and the image shake score before and after adjustment. Based on the image quality assessment algorithm, historical footage from at least the past 30 days is labeled with jitter, and footage with a jitter score below the second threshold SSIM<0.8 is marked as a jitter event.

[0026] The system retrieves the current image and the status data at the corresponding timestamp, and obtains the sliding average speed over five seconds, the variance of acceleration, and the periodic fluctuation amplitude of the guide rail position. The predicted jitter probability within the next five seconds is determined based on the five-second sliding average speed, the variance of acceleration, and the periodic fluctuation amplitude of the guide rail position, where: If the predicted shake probability is >0.7 and the current image stabilization parameters are standard, then switch directly to high shake image stabilization parameters; If the predicted shake probability is >0.9, switch directly to ultra-high shake stabilization parameters and mark it as emergency stabilization mode.

[0027] The aforementioned technology queries a database to match the standard stabilization parameters corresponding to the current state data and calculates the correlation between the target stabilization capability value and the real-time jitter value. When the jitter value is lower than a first preset threshold, the standard parameter is directly used as the correction value; if it is higher than the threshold, the ratio of the state data to the target stabilization capability is calculated, and a matching stabilization strategy is retrieved from the database accordingly. This judgment process is driven by a machine learning model. Model training requires integrating the inspection robot's motion parameters and stabilization logs from the past 30 days. The motion parameters cover sliding speed sequences, three-dimensional acceleration data, and guide rail position changes within multiple time windows, while the stabilization logs record the parameter adjustment time and the image jitter score before and after the adjustment (assessed by image quality algorithms such as SSIM, with images scoring below 0.8 marked as jitter events). During real-time operation, the system collects current image and corresponding timestamp status data, extracts three feature indicators: the five-second sliding average speed, acceleration variance, and the periodic fluctuation amplitude of the guide rail position. These are input into a prediction model to calculate the probability of jitter in the next five seconds. If the probability exceeds 0.7 and the current stabilization parameters are in standard mode, it automatically switches to high jitter stabilization parameters; if the probability exceeds 0.9, it directly activates ultra-high jitter stabilization parameters and activates emergency stabilization mode. The entire process, through a closed-loop mechanism of data synchronization, feature extraction, model inference, and parameter tuning, upgrades the stabilization strategy from passive response to active prediction, effectively improving the stability and clarity of the inspection images.

[0028] If adjustments are needed, the corrected standard image stabilization parameters are determined according to the pre-set image stabilization strategy table, and then applied to the image stabilization process.

[0029] Specifically, the steps for determining the corrected standard image stabilization parameters according to the pre-set image stabilization strategy table include: If the ratio result is greater than the preset first ratio threshold and less than the preset second ratio threshold, then the corrected standard image stabilization parameters are determined to be high-jitter image stabilization parameters. If the ratio result is greater than or equal to the preset second ratio threshold, then the corrected standard image stabilization parameters are determined to be ultra-high shake stabilization parameters.

[0030] In the aforementioned technology, the corrected standard image stabilization parameters are determined and applied based on a pre-configured image stabilization strategy table. Specifically, the image stabilization strategy table uses the ratio of the state data to the target image stabilization capability value corresponding to the standard image stabilization parameters as the core basis for parameter classification: if the ratio is between a preset first and second ratio threshold, a high-jitter image stabilization parameter is selected as the correction value. This type of parameter typically enhances the response speed and compensation amplitude of the motion compensation algorithm to adapt to moderate-intensity shaking scenarios; if the ratio is greater than or equal to the second ratio threshold, an ultra-high-jitter image stabilization parameter is directly matched. This parameter further activates advanced functions such as multi-axis collaborative image stabilization and dynamic inter-frame compensation, and may be accompanied by the linkage control of a hardware-level image stabilization module to cope with extreme shaking conditions. The corrected parameters are synchronized to the image processing pipeline in real time, replacing the original image stabilization configuration. They automatically take effect during the preprocessing stage after image acquisition. Through optimization algorithms such as inter-frame alignment and pixel-level motion smoothing, the stability and clarity of the output image are ensured. Simultaneously, the system records parameter adjustment logs to provide data support for subsequent strategy optimization.

[0031] This embodiment achieves adaptive dynamic adjustment of the image stabilization strategy. By acquiring continuous images from the gimbal camera in real time and their corresponding timestamps, and matching and judging them with standard parameters in the database, the system can automatically and accurately switch to the most suitable stabilization parameters according to the actual working conditions, significantly improving the stability and reliability of image acquisition in complex tunnel environments. Secondly, by combining speed difference, three-dimensional acceleration information, and jitter probability prediction based on a machine learning model, the system can not only identify current jitter but also predict jitter risks in the future, thus supporting pre-adjustment of stabilization parameters, reducing image latency and jitter processing lag, and enhancing the system's real-time performance and robustness. Furthermore, it demonstrates comprehensive anti-interference capabilities in variable tunnel environments. By introducing a photosensitive sensor to monitor changes in ambient light and combining image brightness analysis and exposure mapping adjustment, it effectively overcomes challenges such as uneven lighting and alternating light and dark conditions in tunnels, simultaneously achieving stabilization and exposure compensation, comprehensively improving image quality, and providing a clearer and more stable image data foundation for subsequent tunnel surface defect identification and analysis.

[0032] Example 2

[0033] Based on the above embodiment one, this example further includes, in addition to the state data in the acquired continuous images and the state data under the corresponding timestamps, the installation of a high-sensitivity photosensor near the gimbal camera for real-time acquisition of ambient light intensity inside the tunnel, to obtain raw light intensity data. The processing of the raw light intensity data includes: The original light intensity data is filtered using a moving average to remove high-frequency noise. Then, the mean ambient light intensity after filtering is calculated as a representative value for the current lighting conditions, where: The moving average filter should have a window period of at least five seconds. Acquire a sequence of multiple images captured continuously in a dynamic scene; Extract the brightness information of each frame in the image sequence, and calculate the exposure evaluation index of the current scene based on the brightness information; Based on the deviation between the exposure evaluation index and the preset target exposure range, determine whether exposure correction needs to be initiated; If exposure correction needs to be activated, the corresponding exposure mapping function is selected according to the preset exposure correction parameter lookup table, and the selected mapping function is used to adjust the exposure of the subsequently acquired images in real time. The exposure correction parameter lookup table stores the correspondence between different exposure deviation ranges and exposure mapping functions.

[0034] Furthermore, the steps of extracting the brightness information of each frame in the image sequence and calculating the exposure evaluation index of the current scene based on the brightness information include: Each frame of the image is sampled in sections to obtain the average brightness value of each section; Calculate the global image brightness average based on the average brightness value of all regions; The contrast index of the image is calculated based on the brightness values ​​of each region of the image. By combining the global image brightness mean and contrast index, a comprehensive exposure evaluation index is generated.

[0035] Specifically, a high-sensitivity photosensor is deployed around the gimbal camera to collect raw light intensity data in real time. High-frequency noise is filtered out using a five-second window of moving average filtering, and the filtered ambient light intensity is calculated as a representative value for the current lighting conditions. In dynamic scenes, the system continuously captures multiple image sequences, samples each frame in a partition, extracts the average brightness value of each partition, and calculates the global average brightness. Simultaneously, it calculates the image contrast index based on the brightness differences between partitions, and finally merges the global average brightness and contrast index to generate a comprehensive exposure evaluation index. The system compares this index with a preset target exposure range. If the deviation exceeds the allowable range, an exposure correction mechanism is triggered—matching the corresponding exposure mapping function according to an exposure correction parameter lookup table. This table predefines the mapping relationship between different exposure deviation ranges (such as overexposure / underexposure) and mapping functions (such as linear adjustment, non-linear compensation, or adaptive curves). During the correction process, the system applies the selected mapping function in real time to adjust the pixel-level brightness of subsequently acquired images, ensuring that the exposure value of the output image remains stable within the target range. The entire process, through the collaborative work of hardware sensors, software algorithms, and parameter lookup tables, achieves closed-loop control across the entire chain, from ambient light perception and image brightness analysis to dynamic exposure adjustment, effectively improving the quality and stability of tunnel inspection images under complex lighting conditions.

[0036] Example 3

[0037] This invention provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the following steps: Acquire continuous images from the gimbal camera and the status data at the corresponding timestamps; The system matches the status data with the corresponding standard image stabilization parameters in the database to determine whether the current image stabilization strategy needs to be adjusted. If adjustments are needed, the corrected standard image stabilization parameters are determined according to the pre-set image stabilization strategy table, and then applied to the image stabilization process.

[0038] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0039] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0040] Example 4

[0041] This invention provides an electronic device, including a processor and a memory, wherein the memory stores at least one instruction or at least one program segment, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the following steps: Acquire continuous images from the gimbal camera and the status data at the corresponding timestamps; The system matches the status data with the corresponding standard image stabilization parameters in the database to determine whether the current image stabilization strategy needs to be adjusted. If adjustments are needed, the corrected standard image stabilization parameters are determined according to the pre-set image stabilization strategy table, and then applied to the image stabilization process.

[0042] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for image stabilization processing of a gimbal camera on a tunnel inspection robot, characterized in that, The method includes the following steps: Acquire continuous images from the gimbal camera and the status data at the corresponding timestamps; Based on the status data, the corresponding standard image stabilization parameters are matched with those in the database to determine whether the current image stabilization strategy needs to be adjusted. If adjustments are needed, the corrected standard image stabilization parameters are determined according to the pre-set image stabilization strategy table, and then applied to the image stabilization process.

2. The image stabilization method for a tunnel inspection robot gimbal camera according to claim 1, characterized in that, The status data includes: Obtain the real-time sliding speed of the inspection robot along the guide rail at the corresponding timestamp; Obtain the previous frame's sliding speed from the real-time sliding speed; The speed difference of the gimbal camera is determined based on the difference between the real-time sliding speed and the sliding speed of the previous frame.

3. The image stabilization method for a tunnel inspection robot gimbal camera according to claim 2, characterized in that, The acquisition of continuous images from the gimbal camera and the corresponding status data at the timestamps also includes: The three-dimensional acceleration data of the vehicle sliding along the guide rail is collected in real time by a three-axis accelerometer deployed on the inspection robot. The three-dimensional acceleration data includes acceleration components in the longitudinal, lateral and vertical directions. The collected acceleration data is subjected to low-pass filtering to remove high-frequency noise and retain the low-frequency acceleration signal caused by the movement of the trolley. The low-frequency acceleration signal is compared with a preset acceleration threshold. If the acceleration in a certain direction exceeds the corresponding threshold, the velocity error correction mechanism is triggered. The speed error correction mechanism corrects the real-time sliding speed based on the integral value of the longitudinal acceleration.

4. The image stabilization method for a tunnel inspection robot gimbal camera according to claim 1, characterized in that, The determination of whether the current image stabilization strategy needs to be adjusted includes: Based on the state data and the target stabilization capability value corresponding to the standard stabilization processing parameters, the current image jitter value is obtained and a judgment is made: If the jitter value is less than the first preset threshold, it is determined that there is no risk of jitter, and the standard anti-shake processing parameters are used as the corrected standard anti-shake processing parameters. If the jitter value is greater than or equal to the first preset threshold, the ratio of the state data to the target stabilization capability value corresponding to the standard stabilization processing parameters is calculated, and the stabilization processing strategy matching the database is determined based on the ratio result.

5. The image stabilization method for a tunnel inspection robot gimbal camera according to claim 4, characterized in that, The determination of whether the current image stabilization strategy needs to be adjusted is based on a machine learning model, and its steps include: Record the motion parameters and corresponding anti-shake processing logs of the inspection robot over the past 30 days. The motion parameters include multiple sliding speeds, accelerations, and guide rail positions within multiple window periods. The corresponding anti-shake processing logs include the anti-shake parameter adjustment time and the image shake score before and after adjustment. Based on the image quality assessment algorithm, historical footage from at least the past 30 days is labeled with jitter, and footage with a jitter score below the second threshold SSIM<0.8 is marked as a jitter event; The system retrieves the current image and the status data at the corresponding timestamp, and obtains the sliding average speed over five seconds, the variance of acceleration, and the periodic fluctuation amplitude of the guide rail position. The predicted jitter probability within the next five seconds is determined based on the five-second sliding average speed, the variance of acceleration, and the periodic fluctuation amplitude of the guide rail position, where: If the predicted shake probability is >0.7 and the current image stabilization parameters are standard, then switch directly to high shake image stabilization parameters; If the predicted shake probability is >0.9, switch directly to ultra-high shake stabilization parameters and mark it as emergency stabilization mode.

6. The image stabilization method for a tunnel inspection robot gimbal camera according to claim 1, characterized in that, The acquired continuous images and the state data under the corresponding timestamps also include a high-sensitivity photosensor installed near the gimbal camera to collect ambient light intensity in the tunnel in real time, thereby obtaining raw light intensity data. The processing of the raw light intensity data includes: The original light intensity data is filtered using a moving average to remove high-frequency noise. Then, the mean ambient light intensity after filtering is calculated as a representative value for the current lighting conditions, where: The moving average filter has a window period of at least five seconds; Acquire a sequence of multiple images captured continuously in a dynamic scene; Extract the brightness information of each frame in the image sequence, and calculate the exposure evaluation index of the current scene based on the brightness information; Based on the deviation between the exposure evaluation index and the preset target exposure range, determine whether exposure correction needs to be initiated; If exposure correction needs to be activated, the corresponding exposure mapping function is selected according to the preset exposure correction parameter lookup table, and the selected mapping function is used to adjust the exposure of the subsequently acquired images in real time. The exposure correction parameter lookup table stores the correspondence between different exposure deviation ranges and exposure mapping functions.

7. The image stabilization method for a tunnel inspection robot gimbal camera according to claim 6, characterized in that, The steps of extracting the brightness information of each frame in the image sequence and calculating the exposure evaluation index of the current scene based on the brightness information include: Each frame of the image is sampled in sections to obtain the average brightness value of each section; Calculate the global image brightness average based on the average brightness value of all regions; The contrast index of the image is calculated based on the brightness values ​​of each region of the image. By combining the global image brightness average with the contrast index, a comprehensive exposure evaluation index is generated.

8. The image stabilization method for a tunnel inspection robot gimbal camera according to claim 1, characterized in that, The step of determining the corrected standard image stabilization parameters according to the pre-set image stabilization strategy table includes: If the ratio result is greater than a preset first ratio threshold and less than a preset second ratio threshold, then the corrected standard image stabilization parameters are determined to be high-jitter image stabilization parameters. If the ratio result is greater than or equal to the preset second ratio threshold, then the corrected standard image stabilization parameters are determined to be ultra-high shake stabilization parameters.

9. A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the image stabilization processing method for the gimbal camera of the tunnel inspection robot as described in any one of claims 1-8.

10. An electronic device, characterized in that, The method includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the image stabilization processing method for the gimbal camera of the tunnel inspection robot as described in any one of claims 1-8.

Images