Tower climbing obstacle recognition and positioning method based on machine vision

Abstract

The invention discloses a tower climbing obstacle identification and positioning method based on machine vision. The method mainly comprises the following steps: S1, acquiring comprehensive environment and obstacle information and monitoring illumination conditions in real time; s2, developing an adaptive illumination compensation algorithm, and dynamically adjusting image processing parameters according to real-time illumination conditions; s3, performing semantic segmentation on the image, identifying different types of obstacles, and taking different countermeasures according to the characteristics of the obstacles; s4, constructing a high-precision real-time positioning system, accurately positioning the identified obstacle, and generating three-dimensional coordinate information of the obstacle; s5, tracking the position change of the obstacle in real time in the tower climbing process, and continuously updating the position information of the obstacle; and S6, designing an intelligent path optimization algorithm, planning a safe tower climbing path, and dynamically adjusting the tower climbing path in the tower climbing process. According to the method, misjudgment caused by insufficient illumination or overexposure is greatly reduced, and the safety of tower climbing operation is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial automation and machine vision technology, and in particular to a tower climbing obstacle recognition and positioning method based on machine vision. Background Art

[0002] With the rapid development of modern industrial technology, tower climbing plays a vital role in many fields such as electricity, communications, and construction. However, due to its complexity and high risk, tower climbing has extremely high requirements for safety and efficiency. In the process of tower climbing, accurately identifying and locating obstacles is the key to ensuring operation safety and improving efficiency.

[0003] Traditional methods for identifying and locating tower climbing obstacles mainly rely on manual observation and empirical judgment. This method is not only inefficient, but also easily affected by lighting conditions, environmental complexity and human factors, leading to misjudgment and missed judgment, thereby increasing the risk of tower climbing operations. Especially under conditions of insufficient light or overexposure, traditional image processing methods often have difficulty in providing clear and accurate images, further exacerbating the difficulty of identification. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a tower climbing obstacle identification and positioning method based on machine vision, mainly comprising the following steps:

[0006] S1: Obtain comprehensive environment and obstacle information and monitor lighting conditions in real time;

[0007] S2: Develop an adaptive illumination compensation algorithm to dynamically adjust image processing parameters based on real-time illumination conditions;

[0008] S3: Pre-process the collected images, then use deep learning algorithms to perform semantic segmentation on the images, identify different types of obstacles, and then take different countermeasures based on the characteristics of the obstacles;

[0009] S4: Combine visual data and GPS data to build a high-precision real-time positioning system to accurately locate the identified obstacles and generate the three-dimensional coordinate information of the obstacles;

[0010] S5: Track the position changes of obstacles in real time during the tower climbing process and continuously update the position information of obstacles;

[0011] S6: Design an intelligent path optimization algorithm to plan a safe tower climbing path based on the identified obstacle information and real-time positioning results, and dynamically adjust the tower climbing path during the tower climbing process based on the real-time obstacle and environmental information obtained.

[0012] As a preferred solution of the tower climbing obstacle identification and positioning method based on machine vision described in the present invention, wherein: by installing a high-resolution camera and environmental sensors on the tower climbing equipment, comprehensive environment and obstacle information can be obtained; by installing a light sensor on the machine, the machine lighting conditions can be obtained in real time.

[0013] As a preferred solution of the tower climbing obstacle identification and positioning method based on machine vision described in the present invention, the image preprocessing step includes image denoising, contrast enhancement and image sharpening.

[0014] As a preferred solution of the tower climbing obstacle identification and positioning method based on machine vision described in the present invention, the development of the adaptive illumination compensation algorithm includes the following steps:

[0015] S21: Build a dynamic lighting model based on light source conditions to simulate the impact of different light sources on the image;

[0016] S22: Dynamically adjust the illumination model parameters according to the illumination data collected in real time;

[0017] S23: Build a lighting compensation parameter library, covering the optimal parameter settings under different lighting conditions;

[0018] S24: Use historical data and actual application data to train machine learning models to automatically generate parameter adjustment strategies under different lighting conditions;

[0019] S25: monitor the current lighting conditions in real time, extract corresponding compensation parameters from the parameter library, and dynamically adjust the image processing parameters;

[0020] S26: Develop an image quality assessment algorithm to evaluate the image quality after illumination compensation in real time; establish an adaptive feedback mechanism to dynamically adjust the illumination compensation strategy based on the image quality assessment results.

[0021] As a preferred solution of the tower climbing obstacle identification and positioning method based on machine vision described in the present invention, the formula of the adaptive illumination compensation algorithm is:

[0022]

[0023] in:

[0024] I final(x, y, t) represents the compensated image pixel value, the position is (x, y,) at time t;

[0025] I input (x,y,t) represents the original input image pixel value, position (x,y,) at time t;

[0026] α and β represent adaptive adjustment parameters, which are used to adjust the compensation amplitude based on historical data and machine learning model training results;

[0027] L real (t) represents the real-time light intensity;

[0028] L target Indicates the target light intensity, used as a compensation reference;

[0029] is the illumination compensation factor, and the illumination compensation is dynamically adjusted through parameters α and β;

[0030] When the real-time light intensity L real (t) Approaching target light intensity L target When the real-time light intensity L real (t) and target light intensity L target When the gap is large, the compensation factor will be appropriately adjusted according to the difference and the current light intensity to ensure the compensation effect;

[0031] (L target -L real (t)) reflects the difference between the current light intensity and the target light intensity. If L target >L real (t) then this part is positive and the brightness needs to be increased. If L target <L real (t), then this part is a negative value and the brightness needs to be reduced;

[0032] is a stable term;

[0033] In the stability term, β and L are added to the denominator real (t) 2 To ensure the stability of the formula under any lighting conditions, adding β can prevent the denominator from being zero, and adding L real (t) 2 This allows the compensation factor to change more smoothly when the light intensity is extremely high or low.

[0034] As a preferred solution of the tower climbing obstacle identification and positioning method based on machine vision described in the present invention, the adaptive illumination compensation algorithm also includes a preprocessing module, which uses spectral analysis technology to analyze the spectral characteristics of the light source in order to perform illumination compensation more accurately.

[0035] As a preferred solution of the tower climbing obstacle identification and positioning method based on machine vision described in the present invention, the construction of a high-precision real-time positioning system includes the following steps:

[0036] S41: Fusing multi-sensor data and performing time synchronization on data from different sensors;

[0037] S42: Create an initial environment map using the initial visual image and GPS data, extract feature points in each frame, match them with the feature points in the initial map, and estimate and optimize the pose of each frame by combining IMU data and LiDAR point cloud;

[0038] S43: Use deep learning algorithms to perform semantic segmentation on visual images, identify different types of obstacles, combine visual SLAM and radar point cloud data, perform 3D reconstruction on the identified obstacles, and generate 3D coordinate information of the obstacles;

[0039] S44: Use the fused multi-source data to perform high-precision positioning of obstacles and the robot's real-time position;

[0040] S45: During the tower climbing process, the location information of obstacles is continuously updated, and the path planning is dynamically adjusted according to the updated data.

[0041] As a preferred solution of the tower climbing obstacle identification and positioning method based on machine vision of the present invention, the semantic segmentation of the visual image includes the following steps:

[0042] S431: Collect a diverse set of visual image data, perform enhancement operations on the data, and then use semi-automatic annotation tools to semantically annotate the images and generate pixel-level labels to ensure that each obstacle category has a sufficient number of samples;

[0043] S432: Select a deep learning model architecture suitable for semantic segmentation tasks; introduce an adaptive feature extraction module and a multi-scale context fusion module, and use a large-scale dataset to train the model;

[0044] S433: Fusion of multi-source data

[0045] Multimodal input, that is, combining RGB image and depth image input, using depth image to enhance the segmentation effect of occluded objects and complex scenes;

[0046] Joint training, that is, joint training of multimodal inputs, using cross-modal feature fusion technology to improve the robustness and accuracy of the model;

[0047] S434: Deploy the model on an embedded device or edge computing platform and use hardware accelerators to achieve efficient real-time reasoning; use sliding window technology to segment large images to ensure the segmentation quality of image edge areas;

[0048] S435: Introduce an online learning mechanism to dynamically update the model weights based on feedback data in actual application scenarios to improve model adaptability; regularly collect new data for incremental learning and continuously optimize model performance to ensure high segmentation accuracy in new environments and new obstacle categories.

[0049] As a preferred solution of the tower climbing obstacle identification and positioning method based on machine vision described in the present invention, the three-dimensional reconstruction of the obstacle includes using stereo vision technology to fuse images from multiple perspectives to generate a higher resolution and more accurate three-dimensional model.

[0050] As a preferred solution of the tower climbing obstacle identification and positioning method based on machine vision described in the present invention, wherein: an attention mechanism is introduced into the adaptive feature extraction module.

[0051] Beneficial effects of the present invention:

[0052] 1. The present invention can provide clear and accurate images under various lighting conditions by acquiring environment and obstacle information in real time and combining it with an adaptive light compensation algorithm, making obstacle identification more reliable. This greatly reduces misjudgment caused by insufficient light or overexposure and improves the safety of tower climbing operations.

[0053] 2. The present invention can accurately identify different types of obstacles through semantic segmentation technology, and take different countermeasures according to their characteristics, effectively reducing the risks in the tower climbing process. At the same time, combined with a high-precision real-time positioning system, it can track and update the location information of obstacles in real time, making path planning more accurate and flexible, thereby improving the efficiency of tower climbing operations.

[0054] 3. The adaptive illumination compensation algorithm in the present invention can automatically adjust image processing parameters to adapt to different illumination conditions by constructing a dynamic illumination model and a machine learning model, thereby enhancing the adaptability of the system in different environments.

[0055] 4. The semantic segmentation model in this invention introduces mechanisms such as multimodal input, joint training, and online learning, which can continuously improve the segmentation accuracy and robustness of the model for complex scenes. At the same time, by regularly collecting new data for incremental learning, the model performance can be continuously optimized to ensure that the system maintains high performance in new environments and new obstacle categories. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0057] Figure 1 The present invention is a flowchart of the tower climbing obstacle identification and positioning method based on machine vision. DETAILED DESCRIPTION

[0058] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0059] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0060] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0061] Secondly, the present invention is described in detail with reference to the schematic diagram. When describing the embodiments of the present invention in detail, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0062] Example 1

[0063] Reference Figure 1 , which is the first embodiment of the present invention, provides a tower climbing obstacle identification and positioning method based on machine vision, which mainly includes the following steps:

[0064] S1: Obtain comprehensive environment and obstacle information and monitor lighting conditions in real time. In specific implementation, high-resolution cameras and environmental sensors are installed on the tower climbing equipment to obtain comprehensive environment and obstacle information, and lighting sensors are installed on the machine to obtain the machine lighting conditions in real time.

[0065] S2: Develop an adaptive illumination compensation algorithm to dynamically adjust image processing parameters such as exposure, contrast, brightness, etc. according to real-time illumination conditions, ensure the stability of image quality under different illumination conditions, and improve the accuracy of obstacle recognition.

[0066] Specifically, developing an adaptive illumination compensation algorithm includes the following steps:

[0067] S21: Build a dynamic lighting model based on light source conditions, which can simulate the impact of different light sources on the image, including sunlight, shadows, reflected light, etc., to provide a reference for lighting compensation.

[0068] S22: Dynamically adjust the illumination model parameters according to the illumination data collected in real time, so that the illumination model parameters can adapt to the current illumination conditions and improve the accuracy of illumination compensation.

[0069] S23: Build a lighting compensation parameter library, covering the optimal parameter settings under different lighting conditions;

[0070] S24: Use historical data and actual application data to train machine learning models to automatically generate parameter adjustment strategies under different lighting conditions;

[0071] S25: monitor the current lighting conditions in real time, extract corresponding compensation parameters from the parameter library, and dynamically adjust the image processing parameters;

[0072] S26: Develop an image quality assessment algorithm to evaluate the image quality after illumination compensation in real time to ensure the effectiveness of the compensation effect; establish an adaptive feedback mechanism to dynamically adjust the illumination compensation strategy according to the image quality assessment results to further optimize the image quality.

[0073] The formula of the adaptive illumination compensation algorithm is:

[0074]

[0075] in:

[0076] I final (x, y, t) represents the compensated image pixel value, the position is (x, y,) at time t;

[0077] I input (x,y,t) represents the original input image pixel value, position (x,y,) at time t;

[0078] α and β represent adaptive adjustment parameters, which are used to adjust the compensation amplitude based on historical data and machine learning model training results;

[0079] L real (t) represents the real-time light intensity;

[0080] Ltarget Indicates the target light intensity, used as a compensation reference;

[0081] is the illumination compensation factor, and the illumination compensation is dynamically adjusted through parameters α and β;

[0082] When the real-time light intensity L real (t) Approaching target light intensity L target When , the compensation factor approaches 1, indicating that no major adjustment is required;

[0083] When the real-time light intensity L real (t) and target light intensity L target When the gap is large, the compensation factor will be appropriately adjusted according to the difference and the current light intensity to ensure the compensation effect;

[0084] (L target -L real (t)) reflects the difference between the current light intensity and the target light intensity. If L target >L real (t) then this part is positive and the brightness needs to be increased. If L target <L real (t), then this part is a negative value and the brightness needs to be reduced;

[0085] is a stable term;

[0086] In the stability term, β and L are added to the denominator real (t) 2 To ensure the stability of the formula under any lighting conditions, adding β can prevent the denominator from being zero, and adding L real (t) 2 This makes the compensation factor change more smoothly when the light intensity is extremely high or low;

[0087] Overall effect:

[0088] When the real-time light intensity L real (t) Approaching target light intensity L target When , the compensation factor approaches 1 and the image brightness changes slightly.

[0089] When the real-time light intensity L real (t) When the difference between the target light intensity and the target light intensity is large, the compensation factor will deviate significantly from 1, resulting in a large adjustment of the image brightness to ensure that the image remains clearly visible under extreme lighting conditions.

[0090] The adjustment of the compensation factor is dynamic and can respond to illumination changes in real time, making image processing more robust and adaptable.

[0091] Practical application scenarios:

[0092] In bright light conditions (such as midday sunlight), the compensation factor reduces the brightness of the image to avoid overexposure.

[0093] In low-light conditions (e.g. at night), the compensation factor increases the brightness of the image, improving visibility.

[0094] In environments with changing lighting conditions, the compensation factor can be adjusted in real time to ensure stable image quality.

[0095] Furthermore, through this lighting compensation factor, the system can dynamically adjust the image brightness according to the actual lighting conditions, improving the accuracy and stability of image processing and obstacle recognition.

[0096] S3: Preprocess the collected images, including denoising, enhancement, edge detection and other operations to improve image quality. Use deep learning algorithms (such as U-Net, Mask R-CNN) to perform semantic segmentation on the images, identify different types of obstacles (such as wires, branches, bird nests, etc.), and then take different countermeasures such as avoidance and recording according to the characteristics of the obstacles (such as size, location, type).

[0097] S4: Combine visual data and GPS data to build a high-precision real-time positioning system to accurately locate identified obstacles, generate three-dimensional coordinate information of obstacles, and provide accurate location information for path planning and obstacle avoidance.

[0098] Specifically, building a high-precision real-time positioning system includes the following steps:

[0099] S41: Fuse multi-sensor data and synchronize the data from different sensors. Multi-sensor data fusion and time synchronization can improve the accuracy and consistency of data and provide a reliable data basis for subsequent processing.

[0100] S42: Create an initial environment map using the initial visual image and GPS data, extract feature points in each frame, match them with the feature points in the initial map, and estimate and optimize the pose of each frame by combining IMU data and LiDAR point cloud;

[0101] S43: Use deep learning algorithms to perform semantic segmentation on visual images, identify different types of obstacles, combine visual SLAM and radar point cloud data, reconstruct the identified obstacles in three dimensions, generate the three-dimensional coordinate information of the obstacles, and provide high-precision data support for path planning and obstacle avoidance.

[0102] S44: Use the fused multi-source data to perform high-precision positioning of obstacles and robots in real time to ensure system reliability and safety

[0103] S45: During the tower climbing process, the location information of obstacles is continuously updated, and the path planning is dynamically adjusted according to the updated data to ensure the safety and efficiency of the tower climbing process.

[0104] Furthermore, semantic segmentation of visual images includes the following steps:

[0105] S431: Collect a diverse set of visual image data, enhance the data set, and then use semi-automatic annotation tools to semantically annotate the images and generate pixel-level labels to ensure that each obstacle category has a sufficient number of samples, providing a high-quality data foundation for the training of deep learning models.

[0106] S432: Select a deep learning model architecture suitable for the semantic segmentation task; introduce an adaptive feature extraction module and a multi-scale context fusion module to improve the accuracy and robustness of semantic segmentation; use a large-scale dataset to train the model.

[0107] S433: Fusion of multi-source data

[0108] Multimodal input, that is, combining RGB image and depth image input, using depth image to enhance the segmentation effect of occluded objects and complex scenes. Through multimodal input, the segmentation effect in occluded objects and complex scenes can be improved, and the robustness of the model can be enhanced.

[0109] Joint training, that is, joint training of multimodal inputs, using cross-modal feature fusion technology to improve the robustness and accuracy of the model;

[0110] S434: Deploy the model on an embedded device or edge computing platform and use hardware accelerators to achieve efficient real-time reasoning; use sliding window technology to segment large images to ensure the segmentation quality of image edge areas;

[0111] S435: Introduce an online learning mechanism to dynamically update the model weights based on feedback data in actual application scenarios to improve model adaptability; regularly collect new data for incremental learning and continuously optimize model performance to ensure high segmentation accuracy in new environments and new obstacle categories.

[0112] S5: Track the position changes of obstacles in real time during the tower climbing process, and continuously update the position information of obstacles to ensure dynamic adjustments during path planning and execution to avoid collisions and other dangers.

[0113] S6: Design an intelligent path optimization algorithm to plan a safe tower climbing path based on the identified obstacle information and real-time positioning results. During the tower climbing process, dynamically adjust the tower climbing path based on the real-time obstacle and environmental information to ensure the timeliness and flexibility of path planning and avoid collisions and dangers.

[0114] Specifically, the adaptive illumination compensation algorithm further includes a preprocessing module, which uses a spectrum analysis technique to analyze the spectral characteristics of the light source so as to perform illumination compensation more accurately.

[0115] Specifically, the three-dimensional reconstruction of the obstacle includes fusing images from multiple perspectives using stereo vision technology to generate a higher resolution and more accurate three-dimensional model.

[0116] Specifically, an attention mechanism is introduced into the adaptive feature extraction module to enhance the ability to extract key features and improve the accuracy and efficiency of semantic segmentation.

[0117] In summary, the present invention can provide clear and accurate images under various lighting conditions by acquiring environment and obstacle information in real time and combining with an adaptive illumination compensation algorithm, making obstacle identification more reliable, which greatly reduces misjudgment caused by insufficient illumination or overexposure and improves the safety of tower climbing operations. The present invention can accurately identify different types of obstacles through semantic segmentation technology, and take different countermeasures according to their characteristics, effectively reducing the risks in the tower climbing process. At the same time, combined with a high-precision real-time positioning system, the location information of obstacles can be tracked and updated in real time, making path planning more accurate and flexible, thereby improving the efficiency of tower climbing operations. The adaptive illumination compensation algorithm in the present invention can automatically adjust image processing parameters to adapt to different lighting conditions by constructing a dynamic illumination model and a machine learning model, thereby enhancing the adaptability of the system in different environments. The semantic segmentation model in the present invention introduces mechanisms such as multimodal input, joint training, and online learning, which can continuously improve the segmentation accuracy and robustness of the model for complex scenes. At the same time, by regularly collecting new data for incremental learning, the model performance can be continuously optimized to ensure that the system maintains high performance in new environments and new obstacle categories.

[0118] Example 2

[0119] Referring to Table 1, which is a second embodiment of the present invention, this embodiment is different from the first embodiment in that, in order to verify its beneficial effects, operating data and related instructions of the present invention in an actual environment are provided.

[0120] In order to verify the performance of the present invention in a real environment, a number of experiments were conducted, covering different lighting conditions, various obstacle types, and complex tower climbing paths.

[0121] Experimental setup

[0122] The experimental equipment includes high-resolution cameras, environmental sensors, light sensors, IMUs and lidars. The experimental environment is set up under a variety of lighting conditions, such as strong sunlight, cloudy days and nights, to test image quality, obstacle recognition and positioning.

[0123] Experimental results table

[0124]

[0125] Table 1

[0126] Detailed description

[0127] Ⅰ. Image Quality and Lighting Compensation

[0128] Under different lighting conditions, the image quality is significantly improved through the adaptive illumination compensation algorithm. For example, in a strong sunlight environment, the image is overexposed before compensation, but after compensation, the exposure is balanced and the image details are clear.

[0129] II. Obstacle Identification and Positioning

[0130] Experimental results show that the obstacle recognition accuracy rate is over 95% under various lighting conditions. Using a high-precision real-time positioning system, the obstacle positioning error is less than 10 centimeters, and the path planning can be updated and adjusted dynamically in real time.

[0131] III. Effect of Adaptive Illumination Compensation Algorithm

[0132] By building a dynamic illumination model and a machine learning model, experiments show that the clarity of the compensated image is improved by about 30%. The illumination compensation parameter library covers the optimal parameter settings under different lighting conditions and can adapt to lighting changes in actual environments.

[0133] IV. Semantic Segmentation Model

[0134] By introducing multimodal input, joint training and online learning mechanisms, experiments show that segmentation accuracy is improved by 15% and occluder segmentation accuracy is improved by 20%. Incremental learning ensures that the model maintains high performance in new environments and new obstacle categories.

[0135] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A tower climbing obstacle recognition and positioning method based on machine vision, characterized in that: The following steps are involved: S1: Obtain comprehensive environment and obstacle information and monitor lighting conditions in real time; S2: Develop an adaptive illumination compensation algorithm to dynamically adjust image processing parameters based on real-time illumination conditions; S3: Pre-process the collected images, then use deep learning algorithms to perform semantic segmentation on the images, identify different types of obstacles, and then take different countermeasures based on the characteristics of the obstacles; S4: Combine visual data and GPS data to build a high-precision real-time positioning system to accurately locate the identified obstacles and generate the three-dimensional coordinate information of the obstacles; S5: Track the position changes of obstacles in real time during the tower climbing process and continuously update the position information of obstacles; S6: Design an intelligent path optimization algorithm to plan a safe tower climbing path based on the identified obstacle information and real-time positioning results, and dynamically adjust the tower climbing path during the tower climbing process based on the real-time obstacle and environmental information obtained.

2. The tower climbing obstacle identification and positioning method based on machine vision as claimed in claim 1, characterized in that: By installing high-resolution cameras and environmental sensors on the tower climbing equipment, comprehensive environment and obstacle information can be obtained. By installing light sensors on the machine, the machine's lighting conditions can be obtained in real time.

3. The tower climbing obstacle identification and positioning method based on machine vision as claimed in claim 1, characterized in that: The image preprocessing step includes image denoising, contrast enhancement and image sharpening.

4. The tower climbing obstacle identification and positioning method based on machine vision as claimed in claim 1, characterized in that: The development of the adaptive illumination compensation algorithm comprises the following steps: S21: Build a dynamic lighting model based on light source conditions to simulate the impact of different light sources on the image; S22: Dynamically adjust the illumination model parameters according to the illumination data collected in real time; S23: Build a lighting compensation parameter library, covering the optimal parameter settings under different lighting conditions; S24: Use historical data and actual application data to train machine learning models to automatically generate parameter adjustment strategies under different lighting conditions; S25: monitor the current lighting conditions in real time, extract corresponding compensation parameters from the parameter library, and dynamically adjust the image processing parameters; S26: Develop an image quality assessment algorithm to evaluate the image quality after illumination compensation in real time; establish an adaptive feedback mechanism to dynamically adjust the illumination compensation strategy based on the image quality assessment results.

5. The tower climbing obstacle identification and positioning method based on machine vision as claimed in claim 4, characterized in that: The formula of the adaptive illumination compensation algorithm is: in: I final (x, y, t) represents the compensated image pixel value, the position is (x, y,) at time t; I input (x,y,t) represents the original input image pixel value, position (x,y,) at time t; α and β represent adaptive adjustment parameters, which are used to adjust the compensation amplitude based on historical data and machine learning model training results; L real (t) represents the real-time light intensity; L target Indicates the target light intensity, used as a compensation reference; is the illumination compensation factor, and the illumination compensation is dynamically adjusted through parameters α and β; When the real-time light intensity L real (t) Approaching target light intensity L target When the real-time light intensity L real (t) and target light intensity L target When the gap is large, the compensation factor will be appropriately adjusted according to the difference and the current light intensity to ensure the compensation effect; (L target -L real (t)) reflects the difference between the current light intensity and the target light intensity. If L target >L real (t) then this part is positive and the brightness needs to be increased. If L target <L real (t), then this part is negative and the brightness needs to be reduced; is a stable term; In the stability term, β and L are added to the denominator real (t) 2 To ensure the stability of the formula under any lighting conditions, adding β can prevent the denominator from being zero, and adding L real (t) 2 This allows the compensation factor to change more smoothly when the light intensity is extremely high or low.

6. The tower climbing obstacle identification and positioning method based on machine vision as claimed in claim 1, characterized in that: The adaptive illumination compensation algorithm further includes a pre-processing module, which uses spectrum analysis technology to analyze the spectral characteristics of the light source so as to perform illumination compensation more accurately.

7. The tower climbing obstacle identification and positioning method based on machine vision as claimed in claim 1, characterized in that: The construction of a high-precision real-time positioning system includes the following steps: S41: Fusing multi-sensor data and performing time synchronization on data from different sensors; S42: Create an initial environment map using the initial visual image and GPS data, extract feature points in each frame, match them with the feature points in the initial map, and estimate and optimize the pose of each frame by combining IMU data and LiDAR point cloud; S43: Use deep learning algorithms to perform semantic segmentation on visual images, identify different types of obstacles, combine visual SLAM and radar point cloud data, perform 3D reconstruction on the identified obstacles, and generate 3D coordinate information of the obstacles; S44: Use the fused multi-source data to perform high-precision positioning of obstacles and the robot's real-time position; S45: During the tower climbing process, the location information of obstacles is continuously updated, and the path planning is dynamically adjusted according to the updated data.

8. The tower climbing obstacle identification and positioning method based on machine vision as claimed in claim 7, characterized in that: The semantic segmentation of the visual image comprises the following steps: S431: Collect a diverse set of visual image data, perform enhancement operations on the data, and then use semi-automatic annotation tools to semantically annotate the images and generate pixel-level labels to ensure that each obstacle category has a sufficient number of samples; S432: Select a deep learning model architecture suitable for semantic segmentation tasks; introduce an adaptive feature extraction module and a multi-scale context fusion module, and use a large-scale dataset to train the model; S433: Fusion of multi-source data Multimodal input, that is, combining RGB image and depth image input, using depth image to enhance the segmentation effect of occluded objects and complex scenes; Joint training, that is, joint training of multimodal inputs, using cross-modal feature fusion technology to improve the robustness and accuracy of the model; S434: Deploy the model on an embedded device or edge computing platform and use hardware accelerators to achieve efficient real-time reasoning; use sliding window technology to segment large images to ensure the segmentation quality of image edge areas; S435: Introduce an online learning mechanism to dynamically update the model weights based on feedback data in actual application scenarios to improve model adaptability; regularly collect new data for incremental learning and continuously optimize model performance to ensure high segmentation accuracy in new environments and new obstacle categories.

9. The tower climbing obstacle identification and positioning method based on machine vision according to claim 7, characterized in that: The obstacle three-dimensional reconstruction includes using stereo vision technology to fuse images from multiple perspectives to generate a higher resolution and more accurate three-dimensional model.

10. The tower climbing obstacle identification and positioning method based on machine vision according to claim 8, characterized in that: An attention mechanism is introduced into the adaptive feature extraction module.

Images