A method, system, and medium for oil leak inspection based on periodic image acquisition.

Abstract

This invention relates to the field of aircraft fluid pipeline leak detection technology, specifically to a method, system, and medium for oil leak inspection based on periodic image acquisition. The method involves periodically acquiring images from a camera using a motorized gimbal. Images acquired in the Tth period are identified using an oil leak detection algorithm model, and the oil area ratio matrix for all images in the Tth period is calculated and stored as `current`. Images acquired in the (T+1)th period are also identified using the same algorithm model, and the oil area ratio matrix for all images in the (T+1)th period is calculated and stored as `next`. The norm difference between `current` and `next` is compared to see if it exceeds a threshold `q`. If it does, an oil leak is detected, the oil image is output, `current` is cleared, and `next` is stored in `current`. Otherwise, `current` is cleared, and `next` is stored in `current`, and the process continues for the next period. This invention can perform the task of monitoring and detecting micro-leaks in large-scale fluid pipelines.

Description

Technical Field

[0001] This invention relates to the field of aircraft fluid pipeline leakage detection technology, specifically to an oil leakage inspection method, system, and medium based on periodic image acquisition. Background Technology

[0002] Fluid systems provide power for airborne actuation, fuel for engines, and heat dissipation for airborne equipment. They are critical systems related to flight safety, and their reliability is a key factor affecting aircraft safety and availability. Machine vision-based detection of aircraft fluid line leaks suffers from a limited field of view due to the lens's angle of view. Furthermore, wide-angle lenses can distort oil targets, affecting identification. Therefore, a gimbal is needed to expand the detector's field of view and enable large-scale, periodic leak detection in fluid lines. Summary of the Invention

[0003] The purpose of this invention is to provide a method, system, and medium for oil leak inspection based on periodic image acquisition, for monitoring micro-leakage detection tasks in large-scale fluid pipelines.

[0004] The technical solution of this invention:

[0005] An oil leak inspection method based on periodic image acquisition is proposed. The method uses an electric pan-tilt head to acquire periodic images from a camera. The images acquired in the Tth period are identified by an oil leak detection algorithm model, and the oil area ratio matrix of all images in the Tth period is calculated and stored as current.

[0006] The images acquired in the (T+1)th period are identified using the oil leakage detection algorithm model, and the oil area ratio matrix of all images in the (T+1)th period is calculated and stored as next.

[0007] Compare the norm difference between current and next to see if it exceeds the threshold q. If it does, it is determined that an oil leak has occurred, the oil screen is output, current is cleared and next is stored in current; otherwise, current is cleared directly and next is stored in current, and the next cycle operation is continued.

[0008] Furthermore, the camera acquires images periodically using a motorized pan-tilt head, with the number of images acquired in each period being [number missing]. Every One frame is captured at a time, where α is the horizontal field of view of the camera. This is the time it takes for the camera to rotate one full circle.

[0009] Furthermore, the images collected in each cycle are identified using an oil leak detection algorithm model.

[0010] Furthermore, the oil leak detection algorithm model adopts the U²-Net network model, which includes an encoder, a decoder, and a significant image fusion module.

[0011] The encoder consists of 6 RSU modules, with the first four layers used for feature extraction and dimensionality reduction, and the last two layers expanding the receptive field while maintaining the spatial size through dilated convolution.

[0012] The decoder is used to achieve feature fusion and progressive recovery of spatial resolution;

[0013] The saliency image fusion module generates multi-scale saliency maps through the Sigmoid activation function, upsamples them to the size of the original image, and then further fuses them through concatenation and 1×1 convolution to generate the final saliency prediction map.

[0014] Furthermore, the method for calculating the oil area ratio matrix in the image within the period is as follows:

[0015] ,

[0016]

[0017] In the formula This is the matrix representing the percentage of oil area in the image during the Tth period. For the Tth period, the first The percentage of oil area in the diagram. Let be the number of pixels representing the oil area in the i-th image within the T-th period. This represents the total number of pixels in the entire image.

[0018] Furthermore, the norms of the oil area proportion matrix in two adjacent periods are compared. If the difference exceeds the threshold q, an oil leak is determined to have occurred, an oil leak alarm is output, and the oil area proportion matrix of the current period is updated; otherwise, only the oil area proportion matrix of the current period is updated.

[0019] Furthermore, after updating the current cycle's oil area percentage matrix, the above operation is repeated in the next cycle.

[0020] An oil leak inspection system based on periodic image acquisition, applied to the method described, includes:

[0021] Image acquisition and recognition module:

[0022] It includes an electric gimbal, an image processing oil leakage detection algorithm model, and a calculation module; the electric gimbal performs periodic image acquisition from the camera, identifies the images acquired in the Tth period using the oil leakage detection algorithm model, and calculates the oil area ratio matrix of all images in the Tth period using the calculation module, storing it as current;

[0023] The images acquired in the T+1 period are identified by the oil leakage detection algorithm model, and the oil area ratio matrix of all images in the T+1 period is calculated by the calculation module and stored as next.

[0024] Judgment and Update Module:

[0025] This is used to compare whether the norm difference between current and next exceeds the threshold q. If it does, it is determined that an oil leak has occurred, the oil screen is output, current is cleared and next is stored in current; otherwise, current is cleared directly and next is stored in current, and the next cycle operation continues.

[0026] A computer-readable storage medium storing computer instructions that, when the computer-readable storage medium is run on a computer, cause the computer to execute the oil leak inspection method based on periodic image acquisition.

[0027] The beneficial effects of this invention are:

[0028] This invention provides a method, system, and medium for oil leak inspection based on periodic image acquisition. By using a pan-tilt unit to increase the detector's monitoring field of view, it enables large-scale periodic inspection of fluid pipeline leaks. This achieves the task of monitoring and detecting micro-leaks in large-area fluid pipelines. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of an oil leak inspection method based on periodic image acquisition according to the present invention.

[0030] Figure 2 This is a structural diagram of the RSU module of the present invention. Detailed Implementation

[0031] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0032] One embodiment of the present invention provides a method for inspecting oil leaks based on periodic image acquisition.

[0033] The camera periodically acquires images using a motorized pan-tilt unit. For the images acquired in the T-th period, an oil leak detection algorithm is used to identify the leak and calculate the oil area percentage matrix for all images within that period, storing it as `current`. The same operation is performed for the T+1-th period, storing the matrix as `next`. The norm difference between `current` and `next` is compared to see if it exceeds a threshold `q`. If it does, an oil leak is detected, the oil image is output, `current` is cleared, and `next` is stored as `current`. Otherwise, `current` is cleared and `next` is stored as `current`. Finally, the operation continues for the next period, setting T=T+1.

[0034] The camera acquires images periodically using a motorized pan-tilt head. The number of images acquired in each period is [number missing]. Every One frame is captured at a time, where α is the horizontal field of view of the camera. This is the time it takes for the camera to rotate one full circle.

[0035] An oil leak detection algorithm is used to identify oil leaks in the images acquired within each period. The oil leak detection algorithm is as follows:

[0036] The U²-Net network model is adopted, consisting of three parts: an encoder, a decoder, and a saliency image fusion module. The encoder consists of six RSU modules, with the first four layers mainly performing feature extraction and dimensionality reduction, and the last two layers expanding the receptive field while maintaining spatial size through dilated convolutions, avoiding semantic information loss caused by oversampling. The decoder mainly implements feature fusion and progressive restoration of spatial resolution. The saliency image fusion module generates multi-scale saliency maps through the sigmoid activation function, upsamples to the size of the original image, and finally further fuses them through concatenation and 1×1 convolution to generate the final saliency prediction map.

[0037] Calculate the matrix of oil area proportion in the image within the calculation period.

[0038] The method for calculating the oil area ratio matrix in the image within a period is as follows:

[0039] ,

[0040]

[0041] In the formula This is the matrix representing the percentage of oil area in the image during the Tth period. For the Tth period, the first The percentage of oil area in the diagram. Let be the number of pixels representing the oil area in the i-th image within the T-th period. This represents the total number of pixels in the entire image.

[0042] Compare the norms of the oil area ratio matrix in two adjacent periods. If the difference exceeds the threshold q, an oil leak is determined to have occurred, an oil leak alarm is output, and the oil area ratio matrix for the current period is updated; otherwise, only the oil area ratio matrix for the current period is updated.

[0043] After updating the oil area percentage matrix for the current period, the above operation is repeated in the next period.

[0044] A second embodiment of the invention, for example Figure 1 As shown, an oil leak inspection method based on periodic image acquisition is provided. The specific implementation method is as follows: An oil leak inspection method based on periodic image acquisition uses a motorized pan-tilt unit to acquire periodic images from a camera. The oil leak detection algorithm is used to identify the images acquired in the T-th period, and the oil area ratio matrix of all images in that period is calculated and stored as `current`. The same operation is performed for the T+1 period, and the matrix is ​​stored as `next`. The norm difference between `current` and `next` is compared to see if it exceeds a threshold `q`. If it does, an oil leak is determined, and an oil leak image is output. `current` is cleared, and `next` is stored as `current`. Otherwise, `current` is directly cleared, and `next` is stored as `current`. Finally, T=T+1 is set to continue the operation for the next period.

[0045] The camera acquires images periodically using a motorized pan-tilt head. The number of images acquired in each period is [number missing]. Every One frame is captured at a time, where α is the horizontal field of view of the camera. This is the time it takes for the camera to rotate one full circle.

[0046] Furthermore, an oil leak detection algorithm is applied to the images acquired in each cycle. The oil leak detection algorithm uses the U²-Net network model, with an overall structure that is a typical encoder-decoder architecture. The encoder consists of six RSU modules, with the first four layers using RSU-7 to RSU-4, mainly for feature extraction and dimensionality reduction; the last two layers use an RSU-4F structure, which expands the receptive field while maintaining spatial size through dilated convolution, avoiding semantic information loss caused by oversampling. The decoder is symmetrical to the encoder, with each layer corresponding to an RSU block, such as... Figure 2 As shown, during the decoding process, the upsampled feature map is concatenated with the features at symmetrical positions in the encoder to achieve feature fusion and progressive restoration of spatial resolution. The output fusion module introduces edge output branches in six stages, De1-De5 and En6, and generates multi-scale saliency maps through 3×3 convolution and the Sigmoid activation function; finally, the outputs of each edge are upsampled to the original image size and further fused through concatenation and 1×1 convolution to generate the final saliency prediction map.

[0047] Furthermore, the oil area ratio matrix of the image within the period is calculated. The calculation method for the oil area ratio matrix of the image within the period is as follows:

[0048] ,

[0049]

[0050] In the formula This is the matrix representing the percentage of oil area in the image during the Tth period. For the Tth period, the first The percentage of oil area in the diagram. Let be the number of pixels representing the oil area in the i-th image within the T-th period. This represents the total number of pixels in the entire image.

[0051] Compare the norms of the oil area ratio matrix in two adjacent periods. If the difference exceeds the threshold q, an oil leak is determined to have occurred, an oil leak alarm is output, and the oil area ratio matrix for the current period is updated; otherwise, only the oil area ratio matrix for the current period is updated.

[0052] After updating the oil area percentage matrix for the current period, the above operation is repeated in the next period.

[0053] The third embodiment of the present invention provides an oil leak inspection system based on periodic image acquisition, applied to the method described, including:

[0054] Image acquisition and recognition module:

[0055] It includes an electric gimbal, an image processing oil leakage detection algorithm model, and a calculation module; the electric gimbal performs periodic image acquisition from the camera, identifies the images acquired in the Tth period using the oil leakage detection algorithm model, and calculates the oil area ratio matrix of all images in the Tth period using the calculation module, storing it as current;

[0056] The images acquired in the T+1 period are identified by the oil leakage detection algorithm model, and the oil area ratio matrix of all images in the T+1 period is calculated by the calculation module and stored as next.

[0057] Judgment and Update Module:

[0058] This is used to compare whether the norm difference between current and next exceeds the threshold q. If it does, it is determined that an oil leak has occurred, the oil screen is output, current is cleared and next is stored in current; otherwise, current is cleared directly and next is stored in current, and the next cycle operation continues.

[0059] In summary, this invention provides a method, system, and medium for oil leak inspection based on periodic image acquisition. By using a pan-tilt unit to increase the detector's monitoring field of view, it enables large-scale periodic inspection of fluid pipeline leaks. This achieves the task of monitoring and detecting micro-leaks in large-area fluid pipelines.

[0060] The above description is merely a specific embodiment of the present invention, providing a detailed description of the invention. Parts not covered herein are conventional techniques. However, the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for inspecting oil leaks based on periodic image acquisition, characterized in that, The camera periodically acquires images using an electric pan-tilt head. The images acquired in the Tth period are identified using an oil leakage detection algorithm model, and the oil area ratio matrix of all images in the Tth period is calculated and stored as current. The images acquired in the (T+1)th period are identified using the oil leakage detection algorithm model, and the oil area ratio matrix of all images in the (T+1)th period is calculated and stored as next. Compare the norm difference between current and next to see if it exceeds the threshold q. If it does, it is determined that an oil leak has occurred, the oil screen is output, current is cleared and next is stored in current; otherwise, current is cleared directly and next is stored in current, and the next cycle operation is continued. The oil leak detection algorithm model adopts the U²-Net network model, which includes an encoder, a decoder, and a significant image fusion module. The encoder consists of 6 RSU modules, with the first four layers used for feature extraction and dimensionality reduction, and the last two layers expanding the receptive field while maintaining the spatial size through dilated convolution. The decoder is used to achieve feature fusion and progressive recovery of spatial resolution; The saliency image fusion module generates multi-scale saliency maps through the Sigmoid activation function, upsamples them to the size of the original image, and then further fuses them through concatenation and 1×1 convolution to generate the final saliency prediction map. The method for calculating the oil area ratio matrix in the image within a period is as follows: ， ， In the formula This is the matrix representing the percentage of oil area in the image during the Tth period. This represents the percentage of oil area in the i-th image within the T-th period. Let be the number of pixels representing the oil area in the i-th image within the T-th period. The total number of pixels in the entire image. This refers to the horizontal field of view of the camera.

2. The oil leak inspection method based on periodic image acquisition according to claim 1, characterized in that, The camera acquires images periodically using a motorized pan-tilt head. The number of images acquired in each period is [number missing]. Every One image was captured at a time, in which This is the time it takes for the camera to rotate one full circle.

3. The oil leak inspection method based on periodic image acquisition according to claim 2, characterized in that, The images acquired in each cycle are identified using an oil leak detection algorithm model.

4. The oil leak inspection method based on periodic image acquisition according to claim 1, characterized in that, Compare the norms of the oil area ratio matrix in two adjacent periods. If the difference exceeds the threshold q, an oil leak is determined to have occurred, an oil leak alarm is output, and the oil area ratio matrix for the current period is updated; otherwise, only the oil area ratio matrix for the current period is updated.

5. The oil leak inspection method based on periodic image acquisition according to claim 4, characterized in that, After updating the oil area percentage matrix for the current period, the above operation is repeated in the next period.

6. An oil leak inspection system based on periodic image acquisition, applied to the method described in any one of claims 1-5, characterized in that, include: Image acquisition and recognition module: It includes an electric gimbal, an image processing oil leakage detection algorithm model, and a calculation module; the electric gimbal performs periodic image acquisition from the camera, identifies the images acquired in the Tth period using the oil leakage detection algorithm model, and calculates the oil area ratio matrix of all images in the Tth period using the calculation module, storing it as current; The images acquired in the T+1 period are identified by the oil leakage detection algorithm model, and the oil area ratio matrix of all images in the T+1 period is calculated by the calculation module and stored as next. Judgment and Update Module: This is used to compare whether the norm difference between current and next exceeds the threshold q. If it does, it is determined that an oil leak has occurred, the oil screen is output, current is cleared and next is stored in current; otherwise, current is cleared directly and next is stored in current, and the next cycle operation continues.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when the computer-readable storage medium is run on a computer, cause the computer to execute the oil leak inspection method based on periodic image acquisition as described in any one of claims 1-5.

Images