Vision-based hydraulic hoist oil leakage identification system and method

Abstract

The system comprises a base, a power distribution box and a three-dimensional holder are arranged on the base, an infrared thermal imaging module and a camera are installed at the mobile end of the three-dimensional holder, and the power distribution box is provided with a Bluetooth module, a controller and a driver. The infrared thermal imaging module and the camera are respectively connected with the input end of the controller through network cables, the output end of the controller is connected and communicated with the computer through a network cable, the controller is connected with the three-dimensional holder through the driver, and the controller is connected with the IMU sensor mounted on the hoist through the Bluetooth module; the infrared thermal imaging module and the camera are used for synchronously collecting infrared thermal imaging data and visible light image data of a hydraulic hoist area. The system is used for solving the problems that the omission ratio of manual inspection is high, single infrared detection is easily interfered by the environment, fixed monitoring has a blind area, and static image analysis is poor in vibration resistance in oil leakage detection.

Description

Technical Field

[0001] This invention relates to the field of hydraulic equipment leakage detection technology, specifically to a vision-based hydraulic gate opening and closing mechanism oil leakage identification system and method. Background Technology

[0002] Oil leakage is a common problem during the operation of hydraulic gate hoists. It not only degrades the performance of the hydraulic system and affects the normal operation of the equipment, but can also cause environmental pollution and safety hazards. Traditional oil leakage detection methods have the following shortcomings: Manual inspection: Manual inspection suffers from a high rate of missed detections, making it difficult to detect even minor leaks in a timely manner. Due to the complex structure of hydraulic gate hoists, some leak points may be hidden in hard-to-observe areas, making these areas prone to being overlooked during manual inspection. Furthermore, manual inspection relies heavily on the experience and eyesight of personnel, making it difficult to guarantee the consistency and accuracy of the detection results due to subjective factors.

[0003] Single infrared detection: Infrared detection is a commonly used non-destructive testing method that identifies temperature anomalies by detecting infrared radiation from an object's surface. However, single infrared detection is easily affected by factors such as ambient temperature changes and sunlight interference in complex environments, leading to inaccurate results. For example, in high ambient temperatures, infrared detection may misjudge temperature changes caused by oil leaks, thus increasing the false negative rate.

[0004] Fixed monitoring: Traditional fixed monitoring equipment suffers from blind spots. Due to its fixed viewing angle, it cannot provide comprehensive coverage of the complex structure of hydraulic gate hoists, resulting in some areas remaining unmonitored. Furthermore, once the installation location and angle of fixed monitoring equipment are determined, it is difficult to adjust them according to actual conditions, failing to meet the monitoring needs of different scenarios.

[0005] Static image analysis: Static image analysis performs poorly in terms of vibration resistance. In industrial settings, hydraulic gate hoists vibrate during operation, causing image blurring or jitter, thus affecting the accuracy of static image analysis. Therefore, in dynamic environments, static image analysis struggles to reliably identify oil leakage characteristics.

[0006] Given the many shortcomings of traditional detection methods, it is particularly important to develop a hydraulic gate opening and closing mechanism oil leakage identification system and method that can effectively solve the above problems. Summary of the Invention

[0007] The purpose of this invention is to provide a vision-based hydraulic gate opening and closing mechanism oil leakage identification system and method to solve the problems of high missed detection rate of manual inspection, susceptibility of single infrared detection to environmental interference, blind spots of fixed monitoring, and poor vibration resistance of static image analysis in oil leakage detection.

[0008] To solve the above problems, the technical solution of the present invention is as follows: like Figures 1 to 3 As shown, a vision-based hydraulic gate hoist oil leakage identification system includes a base, on which a power distribution box and a 3D pan-tilt unit are mounted. The 3D pan-tilt unit's moving end is equipped with an infrared thermal imaging module and a camera. The power distribution box includes a Bluetooth module, a controller, and a driver. The infrared thermal imaging module and camera are connected to the controller's input end via network cables. The controller's output end communicates with a computer via a network cable. The controller is connected to the 3D pan-tilt unit via the driver. The IMU sensor communicates with the controller via Bluetooth using the UART protocol, with a sampling rate of 100Hz. The infrared thermal imaging module and camera are used to simultaneously acquire infrared thermal imaging data and visible light image data of the hydraulic gate hoist area. The 3D pan-tilt unit is used to drive the movement of the infrared thermal imaging module and camera. The computer has a built-in image processing module for processing the infrared thermal imaging data and visible light image data, including dynamic background modeling, dual-channel feature fusion, and leakage feature extraction. The IMU sensor is used to acquire vibration data in real time and fuse it with visual image data.

[0009] Furthermore, an audible and visual alarm module is provided on the base. The audible and visual alarm module is connected to the output of the controller and is used to trigger a leakage alarm based on the output of the image processing module.

[0010] Furthermore, an air pump is installed on the base, and a hollow ring plate is installed on the outer shell of the camera lens. Multiple air jet holes are opened on the inner wall of the ring plate. The ring plate is connected to the air pump through an air inlet pipe, and the controller controls the start and stop of the air pump.

[0011] Furthermore, the infrared thermal imaging module employs a 640×480 resolution uncooled microbolometer with an operating wavelength of 8-14μm and a temperature sensitivity ≤0.05℃. It also integrates a temperature compensation circuit and an adaptive ROI algorithm. The 640×480 resolution uncooled microbolometer can clearly present temperature field details with high pixel density in environments ranging from -20℃ to 60℃. Combined with the 8-14μm operating wavelength, it penetrates media such as oil mist and matches the peak thermal radiation of hydraulic oil. The ≤0.05℃ temperature sensitivity can capture the slight temperature difference at the initial stage of micro-leakage. The temperature compensation circuit calibrates the ambient temperature drift in real time to ensure measurement accuracy. The adaptive ROI algorithm focuses on high-leakage areas such as valve groups, reducing data processing volume and improving real-time detection, thereby achieving early and accurate identification and dynamic tracking of hydraulic gate oil micro-leakage.

[0012] Furthermore, the camera is equipped with a 20-megapixel CMOS sensor, a polarizing filter, and a multispectral fusion lens. The stainless steel pipes of the hydraulic hoist are prone to specular reflection; the polarizing filter (adjustable from 0° to 180°) can filter polarized light in a specific direction, improving image contrast in oil-stained areas. For example, by adjusting the polarization angle to 45°, reflections from the pipes can be effectively suppressed, highlighting the diffuse reflection characteristics of the oil stains.

[0013] Furthermore, the computer transmits image data to the mobile phone via a wireless network. Based on the two-dimensional coordinates and three-dimensional positioning results of the leakage area, the computer overlays annotations (such as red boxes marking the leak location, along with parameters such as area and leakage volume) onto the visible light image or infrared thermal image, generating an annotated composite image. The computer connects to a local area network or 4G / 5G network via its built-in wireless network module and packages the annotated image using the TCP / IP protocol to transmit it to the mobile app. After receiving the image, the mobile app displays a visual interface with leak annotations, supporting zooming and dragging to view details, and simultaneously displays information such as the leakage area (e.g., 2-5 cm² for Level 1 alarm), three-dimensional coordinates, etc. Combined with the preset observation pose of the 3D pan-tilt unit (e.g., top view of the valve assembly, side view of the pipe joint), the mobile app can generate a navigation path from the current location to the leak point, assisting staff in quickly locating the leak.

[0014] A vision-based method for identifying hydraulic gate valve oil leakage includes the following steps: Step 1: Simultaneously acquire infrared thermal imaging data and visible light image data of the hydraulic hoist area through the infrared thermal imaging module and camera. At the same time, the 3D gimbal drives the movement of the infrared thermal imaging module and camera. Based on the vibration data of triaxial acceleration and angular velocity detected by the IMU sensor, eliminate equipment vibration interference through dynamic background modeling. Step 2: Extract temperature gradient features from infrared thermal imaging data, analyze texture changes from visible light image data, and fuse the two types of features using DS evidence theory; Step 3: When the joint confidence level exceeds the threshold, trigger a multi-level alarm based on the area of ​​the leakage area, and calculate the three-dimensional coordinates of the leakage point based on multi-view vision positioning technology.

[0015] Furthermore, the dynamic background modeling in step 1 includes: updating the background pixel probability model based on the improved ViBe algorithm, calculating the pixel motion vector between adjacent frames by combining optical flow field estimation, and the amplitude compensation accuracy is 0.05 pixels.

[0016] Furthermore, in step 2, the infrared thermal imaging data processing also includes: when the ambient temperature is >30℃, the ΔT threshold is automatically lowered to 1.2℃ / s; the visible light image data processing also includes: using the multi-scale Retinex algorithm to enhance the oil stain texture and extract HOG feature descriptors.

[0017] The beneficial effects of this invention are as follows: 1. Multispectral Collaborative Detection: This system integrates an infrared thermal imaging module and a high-resolution camera, leveraging their complementary strengths. The infrared thermal imaging module captures the subtle temperature differences that arise in the early stages of micro-leakage, while the camera provides clear visible light images. Compared to a single infrared detection solution, this multispectral collaborative approach effectively improves the sensitivity of micro-leakage identification, enabling earlier detection of oil leaks. This allows for early and accurate identification and dynamic tracking of micro-leakage in hydraulic gate hoists, buying valuable time for timely maintenance.

[0018] 2. Strong anti-interference capability: Through dynamic background modeling technology, the system can eliminate interference from equipment vibration and complex environments. It utilizes an improved ViBe algorithm to update the background pixel probability model and combines optical flow field estimation to calculate pixel motion vectors, thereby accurately distinguishing between real leakage and interference factors such as vibration noise. This feature enables the system to maintain stable detection performance in complex industrial operating environments, significantly improving recognition accuracy and reducing the probability of false alarms and missed alarms.

[0019] 3. Full-Area Coverage Monitoring: Leveraging the multi-degree-of-freedom motion capabilities of a 3D pan-tilt unit, equipped with an infrared thermal imaging module and camera, the unit can scan the hydraulic gate hoist area from all angles, effectively reducing blind spots present in traditional fixed-view monitoring. By pre-setting multiple observation poses, the system can flexibly monitor various key components, achieving comprehensive and blind-spot-free monitoring of the hydraulic gate hoist. This improves the coverage and reliability of the entire detection system, ensuring that no potential leakage point is missed.

[0020] 4. Low maintenance cost design: On the one hand, the camera lens is equipped with an air pump and a hollow ring plate to prevent oil and dust from entering the lens. This allows for timed air jet cleaning of oil and dust from the lens surface, extending the lens cleaning cycle and reducing the frequency and labor intensity of manual lens cleaning. On the other hand, an audible and visual alarm module is installed in the base. Once a leak is detected, an alarm is issued in time to remind staff to carry out maintenance, preventing further damage to the equipment from continuous oil leakage. This reduces maintenance costs and downtime caused by equipment failure, and improves the overall operating efficiency and economic benefits of the equipment.

[0021] 5. Precise Positioning and Efficient Navigation: The system combines multi-view visual positioning technology with a Kalman filter algorithm to accurately calculate the three-dimensional coordinates of the leak point. When a leak is detected, it not only displays the leak location and related information visually on a computer or mobile device, but also provides staff with a navigation path from their current location to the leak point. This function greatly facilitates maintenance personnel in quickly and accurately locating leaks, shortens maintenance response time, improves the efficiency and focus of maintenance work, helps to promptly repair leaks, ensures the normal operation of hydraulic gate hoists, and enhances the safety and maintenance efficiency of industrial equipment. Attached Figure Description

[0022] The invention will be further described below with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the structure of the present invention. Figure 2 This is a front view structural diagram of the hollow ring plate of the present invention. Figure 3 This is a schematic diagram showing the connection relationship between the various electrical components of the present invention.

[0023] The diagram includes: base 1, power distribution box 2, computer 3, mobile phone 4, air pump 5, IMU sensor 7, hollow ring plate 8, infrared thermal imaging module 9, 3D pan-tilt unit 10, camera 11, sound and light alarm module 12, and air jet 13. Detailed Implementation

[0024] 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.

[0025] The vision-based hydraulic gate hoist 6 oil leakage identification system aims for accurate detection and efficient early warning. It includes a base 1, on which a power distribution box 2 and a 3D pan-tilt unit 10 are mounted. The power distribution box 2 houses a controller, a Bluetooth module, and a driver. The Bluetooth module, model HC-05, handles wireless data transmission. The controller is an STM32F407, and the driver, model TB6600, drives the 3D pan-tilt unit 10. The 3D pan-tilt unit 10, model MG400, has an infrared thermal imaging module 9 and a camera 11 mounted on its moving end. The infrared thermal imaging module 9 is a FLIR A320, equipped with a 640×480 resolution uncooled microbolometer operating in the 8-14μm wavelength range, used to capture subtle temperature changes in the early stages of micro-leakage. The camera 11 is a Sony IMX586, used to acquire visible light image data of the hydraulic gate hoist 6 area. The infrared thermal imaging module 9 and the camera 11 are connected to the controller input via Cat5e Ethernet cables. The controller output also communicates with computer 3 via a Cat5e network cable. Computer 3 has a built-in image processing module based on a deep learning framework, which can perform processing on infrared thermal imaging data and visible light image data. This includes using the improved ViBe algorithm for dynamic background modeling to effectively eliminate equipment vibration and environmental interference; and using DS evidence theory to achieve dual-channel feature fusion to accurately extract leakage features.

[0026] Meanwhile, the controller connects to the MPU6050 IMU sensor installed in key parts of the hoist via Bluetooth. The MPU6050 can collect vibration data during the hoist's operation in real time, deeply integrate it with visual image data, and use a Kalman filter algorithm to compensate for vibration in the image, ensuring stable and accurate identification of oil leakage even under equipment vibration conditions, significantly improving the system's reliability and accuracy in complex industrial environments.

[0027] Furthermore, an audible and visual alarm module 12 is provided on the base 1. The audible and visual alarm module 12 is connected to the controller output and is used to trigger a leakage alarm based on the output of the image processing module. The image processing module fuses and analyzes infrared thermal imaging data and visible light images to calculate the leakage area and leakage rate. When the leakage rate exceeds 0.8 L / min, a trigger signal is transmitted to the controller, which then sends a start command to the audible and visual alarm module 12, causing the buzzer in the alarm module to emit a high-frequency alarm sound, while the LED lights flash red to indicate the abnormality. This improves the safety and maintenance efficiency of industrial equipment operation.

[0028] Furthermore, an air pump 5 is installed on the base 1, and a hollow ring plate is installed on the lens end housing of the camera 11. Multiple air jet holes 13 are opened on the inner wall of the ring plate. The ring plate is connected to the air pump 5 through an air inlet pipe, and the controller controls the start and stop of the air pump 5. In practice, the controller starts the air pump 5 at a set time, using compressed air to blow clean the lens end glass through the air jet holes 13, periodically removing surface oil and dust, keeping the lens clean, and ensuring image quality.

[0029] Furthermore, the infrared thermal imaging module 9 employs a 640×480 resolution uncooled microbolometer with a working wavelength of 8-14μm and a temperature sensitivity of ≤0.05℃, and integrates a temperature compensation circuit and an adaptive ROI algorithm.

[0030] Furthermore, the camera 11 is equipped with a 20-megapixel CMOS sensor, a polarizing filter, and a multispectral fusion lens.

[0031] Furthermore, the computer 3 transmits image data to the mobile phone 4 via a wireless network.

[0032] A vision-based method for identifying oil leakage in a hydraulic gate hoist includes the following steps: Step 1: Synchronous image acquisition and multispectral data registration; Step 1.1. Hardware synchronization: The infrared thermal imaging module 9 (640×480 resolution, 8-14μm band) and the 20-megapixel visible light camera 11 shoot synchronously at a frame rate of 30fps, and the synchronization is achieved with a time deviation of <1ms through the synchronization trigger circuit.

[0033] The visible light camera 11 is equipped with a polarizing filter (adjustable from 0° to 180°) to suppress the specular reflection of stainless steel pipes and improve the contrast of oil stain areas; the infrared module integrates a temperature compensation circuit to monitor the ambient temperature from -20°C to 60°C in real time and dynamically adjust the gain parameters. Step 1.2. Data registration processing: First, coarse alignment of infrared and visible light images is completed by SIFT feature point matching. Then, sub-pixel level fine alignment is achieved by using thin plate spline (TPS) transformation. Finally, the registration error is controlled within 0.3 pixels. An adaptive ROI algorithm is used to focus on parts such as pipe joints and sealing rings, reducing data processing by 40% and improving real-time performance.

[0034] Step 2: Vibration compensation and dynamic background modeling; Step 2.1. Inertial Data Fusion: Vibration data of the equipment is collected through IMU sensor 7 (sampling rate 100Hz, including a triaxial accelerometer ±8g and a gyroscope ±2000° / s). The inertial data and visual features are fused by Kalman filtering to predict image offset and compensate in advance. This ensures the accuracy of the leakage area contour extraction and reduces the false detection rate of the inter-frame difference method.

[0035] The methods for dynamically adjusting the weights of IMU data and visual features include: When the IMU detects the vibration frequency f >10 Hz At that time, the weight of visual features in the Kalman filter decreased from 0.6 to 0.4, while the weight of inertial data increased from 0.4 to 0.6. A sliding window (window size 50 frames) was introduced to calculate the root mean square of vibration (RMS). When RMS > 0.5g, a super-resolution reconstruction algorithm (such as SRCNN) was automatically activated to repair image blur.

[0036] Step 2.2. Dynamic background modeling: Based on the improved ViBe algorithm, the background pixel probability model is updated every frame, and the pixel motion vector between adjacent frames is calculated by combining optical flow field estimation (accuracy 0.01 pixels / frame) to eliminate pseudo motion interference caused by device vibration (2-15Hz frequency).

[0037] When using the background model difference method to separate the dynamic foreground, the amplitude compensation accuracy reaches 0.05 pixels, effectively distinguishing between real leakage and vibration noise. In addition, the background model is initialized with three frames of difference to solve the "ghosting" problem, and dynamic weights of pixel neighborhoods are introduced (such as a weight of 0.6 for the center pixel and a weight of 0.4 for the 8-neighborhood) to improve the robustness of background updates in vibration scenarios.

[0038] Step 3: Dual-channel feature extraction and fusion analysis; Step 3.1 Infrared channel feature extraction, calculate the temperature difference matrix between adjacent frames, and use adaptive morphological processing: use a 3×3 circular kernel for discrete noise points <5 pixels, use a 5×1 horizontal kernel for strip artifacts, and combine the candidate leakage areas with the region growing algorithm.

[0039] Temperature gradient features are extracted: the leakage area ΔT≥1.5℃ / s. When the ambient temperature>30℃, the threshold is automatically lowered to 1.2℃ / s. Dynamic judgment criteria are constructed by using a 3×3 neighborhood temperature difference matrix and the time dimension change rate curve. Step 3.2 Visible light channel feature extraction: The multi-scale Retinex algorithm is used to enhance the oil stain texture. Three levels of Gaussian wrapping scale are set: 15, 80, and 250 pixels, which respectively enhance the micro-texture, the contrast of the medium area, and the global brightness balance. HOG feature descriptor extraction: The image is divided into 16×16 pixel cell units, and a 9-direction gradient histogram is calculated for each unit to form a 1764-dimensional feature vector; 12-dimensional texture features (contrast, energy, etc.) are extracted by combining LBP and GLCM. The SVM classifier achieves an accuracy of 94.3% on a dataset of 2000 images.

[0040] Step 3.3 Feature fusion decision: The DS evidence theory is used to fuse two types of features. The confidence threshold is set to 0.85, and the weights are dynamically adjusted according to the ambient light intensity (acquired in real time by the light sensor): the infrared weight is increased from 0.6 to 0.8 in low illumination (<50 lux). A time-domain cumulative decision mechanism is introduced: the confidence of a single frame must exceed the threshold for 5 consecutive frames to trigger an alarm, avoiding false alarms due to instantaneous interference; if the leakage area expansion rate of adjacent frames is >10%, the alarm level is immediately upgraded.

[0041] The dual-channel feature fusion step includes: dynamically adjusting the infrared channel weighting coefficients according to the ambient temperature. Visible light channel weights The vibration compensation step includes: utilizing the triaxial angular velocity data from the IMU. , , Through rotation matrix Predict image offset.

[0042] The weighting coefficient of the infrared channel is used to measure the importance of infrared thermal imaging data in feature fusion, and the value ranges from 0.6 to 0.8. 30 represents the ambient temperature (unit: °C), which is collected in real time by the temperature sensor built into the infrared thermal imaging module 9; 30 represents the temperature threshold (unit: °C), which is the benchmark critical point for weight adjustment.

[0043] Step 4: Multi-level alarm and three-dimensional coordinate positioning; Step 4.1 Progressive alarm mechanism: Level 1 alarm (leakage area 2-5cm²): Log only; Level 2 alarm (5-10cm²): Send SMS alert; Level 3 alarm (>10cm², corresponding to a leakage rate of 0.8L / min): Activates audible and visual alarms and links to shutdown protection. The alarm threshold can be flexibly configured through the HMI interface according to the oil type (such as fire-resistant hydraulic oil). Step 4.2. Precise three-dimensional coordinate positioning: Calculate the three-dimensional coordinates of the leak point based on the PnP algorithm, with a positioning error of ≤ ±5mm; Use multi-view vision positioning technology to calculate depth information through parallax of two or more cameras. During the prior calibration phase, a checkerboard target (accuracy 0.01mm) is used to establish the camera coordinate system mapping. During on-site deployment, a laser rangefinder is used for rapid calibration (time <3 minutes). When the 3D gimbal 10 drives the camera 11 to change pose, the coordinate transformation matrix is ​​updated in real time through the Denavit-Hartenberg parameter model to ensure positioning accuracy. Step 5: System self-maintenance and adaptive optimization; Step 5.2. The 3D pan-tilt unit 10 (±30° pitch, 360° horizontal rotation, 0.5-2m telescopic) is driven by a servo motor, with a positioning repeatability of ±0.1°. It has 12 preset typical observation poses (such as top view of valve group, side view of pipe joint), and can receive PLC commands via Modbus protocol to automatically switch the viewing angle, covering more than 80% of traditional monitoring blind spots.

[0044] The embodiments described in this specification are merely examples of implementations of the inventive concept. The scope of protection of this invention should not be considered as limited to the specific forms stated in the embodiments. The scope of protection of this invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A vision-based hydraulic gate hoist oil leakage identification system, characterized in that: The system includes a base, a power distribution box, and a 3D pan-tilt unit. The 3D pan-tilt unit's mobile end is equipped with an infrared thermal imaging module and a camera. The power distribution box includes a Bluetooth module, a controller, and a driver. The infrared thermal imaging module and camera are connected to the controller's input via network cables. The controller's output is connected to a computer via a network cable. The controller is connected to the 3D pan-tilt unit via the driver and to an IMU sensor mounted on the gate hoist via the Bluetooth module. The infrared thermal imaging module and camera are used to simultaneously acquire infrared thermal imaging data and visible light image data of the hydraulic gate hoist area. The 3D pan-tilt unit drives the movement of the infrared thermal imaging module and camera. The computer has a built-in image processing module for processing the infrared thermal imaging data and visible light image data, including dynamic background modeling, dual-channel feature fusion, and leakage feature extraction. The IMU sensor is used to acquire vibration data in real time and fuse it with visual image data.

2. The vision-based hydraulic gate opening and closing mechanism oil leakage identification system according to claim 1, characterized in that: An audible and visual alarm module is provided on the base. The audible and visual alarm module is connected to the output of the controller and is used to trigger a leakage alarm based on the output of the image processing module.

3. The vision-based hydraulic gate opening and closing mechanism oil leakage identification system according to claim 1 or 2, characterized in that: An air pump is installed on the base, and a hollow ring plate is installed on the outer shell of the camera lens. Multiple air jet holes are opened on the inner wall of the ring plate. The ring plate is connected to the air pump through an air inlet pipe, and the controller controls the start and stop of the air pump.

4. The vision-based hydraulic gate opening and closing mechanism oil leakage identification system according to claim 1 or 2, characterized in that: The infrared thermal imaging module uses a 640×480 resolution uncooled microbolometer with a working wavelength of 8-14μm and a temperature sensitivity of ≤0.05℃. It also integrates a temperature compensation circuit and an adaptive ROI algorithm.

5. The vision-based hydraulic gate opening and closing mechanism oil leakage identification system according to claim 1 or 2, characterized in that: The camera is equipped with a 20-megapixel CMOS sensor, a polarizing filter, and a multispectral fusion lens.

6. The vision-based hydraulic gate opening and closing mechanism oil leakage identification system according to claim 1 or 2, characterized in that: The computer transmits image data to the mobile phone via a wireless network.

7. A vision-based method for identifying hydraulic gate opening and closing oil leakage, characterized in that, Includes the following steps: Step 1: Simultaneously acquire infrared thermal imaging data and visible light image data of the hydraulic hoist area through the infrared thermal imaging module and camera. At the same time, the 3D gimbal drives the movement of the infrared thermal imaging module and camera. Based on the vibration data of triaxial acceleration and angular velocity detected by the IMU sensor, eliminate equipment vibration interference through dynamic background modeling. Step 2: Extract temperature gradient features from infrared thermal imaging data, analyze texture changes from visible light image data, and fuse the two types of features using DS evidence theory; Step 3: When the joint confidence level exceeds the threshold, trigger a multi-level alarm based on the area of ​​the leakage area, and calculate the three-dimensional coordinates of the leakage point based on multi-view vision positioning technology.

8. The vision-based hydraulic gate opening and closing mechanism oil leakage identification method according to claim 7, characterized in that, The dynamic background modeling in step 1 includes: updating the background pixel probability model based on the improved ViBe algorithm, calculating the pixel motion vector between adjacent frames by combining optical flow field estimation, and the amplitude compensation accuracy is 0.05 pixels.

9. A vision-based method for identifying hydraulic gate oil leakage according to claim 7, characterized in that, In step 2, the infrared thermal imaging data processing also includes: when the ambient temperature is > 30℃, the ΔT threshold is automatically lowered to 1.2℃ / s; the visible light image data processing also includes: using the multi-scale Retinex algorithm to enhance the oil stain texture and extract HOG feature descriptors.

Images