A low cloud and fog identification system based on laser wide-area scanning

Abstract

The application discloses a kind of identification systems of low cloud fog based on laser wide-area scanning, comprising the following steps:S1, according to the distance resolution data of laser radar fast wide-area scanning, the visibility information obtained by inversion;S2, according to the historical data training network model of forward scattering visibility meter and laser radar;S3, according to the visibility information, draw horizontal visibility distribution chart;S4, image processing is carried out to the visibility distribution chart, and the development path of low cloud fog is identified;And draw the low-visibility area boundary in the original drawing;S5, according to the low cloud fog development boundary identified, the visibility condition of each channel or road is obtained and output;Through the change process of fog in wide-range scanning port range, obtain all the visibility information in scanning range, and have high space-time resolution, can more fine observation entire sea fog change process, it is favorable to more fine guidance port navigation.

Description

Technical Field

[0001] The present invention relates to the field of laser scanning technology, and in particular to a low cloud and fog recognition system based on laser wide-area scanning. Background Art

[0002] Fast and efficient transportation significantly promotes rapid economic development. In a modern integrated transportation system, highways, ports, and airports serve as core hubs, carrying the heavy burden of transporting massive amounts of people and materials. Safety is crucial for efficient land, sea, and air transportation, but integrated transportation systems are severely affected by weather. For example, high-speed transportation is susceptible to fog, rain, snow, and ice accumulation; maritime transportation, especially ports, is susceptible to fog, strong winds and waves; and air transportation is susceptible to wind shear, fog, and lightning. Heavy fog has the most widespread impact on land, sea, and air transportation. Fog significantly reduces visibility and severely disrupts traffic operations. Due to the complexity and variability of the formation and dissipation of low-lying clouds and fog, and the limited monitoring methods and accuracy, timely forecasts of the formation and dissipation of fog are often impossible. Currently, various transportation systems respond to heavy fog by closing highways or port access, waiting until the fog in all controlled areas has naturally dissipated to within safe visibility before allowing traffic to resume. This not only causes significant inconvenience for commuters but also negatively impacts the efficient operation of the social economy, leading to chain reactions such as increased logistics costs and trade disruptions. In reality, fog does not dissipate in a generalized manner. For example, sea fog often develops and dissipates locally. Ports are generally large, and waiting for the fog to dissipate across the entire port before opening will severely impact port capacity. Therefore, accurately identifying, monitoring, and forecasting the development of low-lying fog is a key challenge in ensuring the rapid operation of transportation.

[0003] Low-lying fog monitoring methods primarily include traditional ground-based monitoring, satellite remote sensing, and cruise monitoring. Traditional ground-based monitoring data is accurate but limited in scope and lags behind. Satellite remote sensing offers a wide range and high temporal resolution, but is susceptible to atmospheric conditions and lacks precision over small areas. Cruise monitoring is flexible and efficient, but requires regular sensor calibration and is unsuitable for long-term, fixed-point monitoring. In contrast, lidar visibility measurement offers the advantages of real-time availability, high accuracy, and wide coverage. It effectively addresses the shortcomings of existing technologies and provides more reliable technical support for the dynamic monitoring of low-lying fog. Summary of the Invention

[0004] Purpose of the invention: In order to overcome the inaccurate prediction of sea fog and thus affect the port's transportation capacity, the present invention provides a low cloud and fog recognition system based on laser wide-area scanning with high temporal and spatial resolution information capability, which can accurately identify sea fog and provide precise location information.

[0005] To achieve the above objectives, the present invention adopts a technical solution of providing a low cloud and fog recognition system based on wide-area laser scanning, comprising the following steps:

[0006] S1, visibility information is obtained by inverting the range-resolved data from the rapid wide-area scanning of the lidar;

[0007] S2, trained based on forward scatter visibility meter and lidar historical data Type network model, and the visibility information inverted in S1 is used through the U-type network model to improve the measurement accuracy;

[0008] S3, drawing a horizontal visibility distribution map based on the visibility information after improving the measurement accuracy; S4, performing threshold segmentation, edge detection, and contour extraction on the obtained horizontal visibility distribution map using a low cloud and fog recognition algorithm to identify the low cloud and fog development path;

[0009] S5: Obtain and output the visibility conditions of each waterway or road based on the identified low cloud and fog development path.

[0010] The visibility information is retrieved from the distance-resolved data of the rapid wide-area scanning of the laser radar in S1 by inverting the radar signal information into the visibility information according to the horizontal visibility inversion algorithm. The horizontal visibility inversion algorithm is:

[0011] ;

[0012] ;

[0013] in, is the atmospheric extinction coefficient, For visibility, is the coefficient, determined by the empirical formula, is the wavelength, and R is the detection distance.

[0014] As a preferred embodiment of the present invention: S2 is specifically:

[0015] The visibility data of the forward scattering visibility meter and the visibility data of the visibility lidar are divided into a training set, a validation set and a test set according to a preset ratio. The visibility data of the forward scattering visibility meter is used as a label and the visibility data of the visibility lidar is used as input data. The visibility data is used to train the Type network model, the model obtained by training is a denoising network.

[0016] As a preferred embodiment of the present invention: The network model training specifically includes the following steps:

[0017] S2-1: The lidar quickly scans a wide area to obtain raw signal data, and then uses the horizontal visibility inversion algorithm to invert the visibility information with distance resolution.

[0018] S2-2, inputting the inverted visibility information and the forward scatter visibility meter visibility information into a U-shaped network model in pairs, wherein the forward scatter visibility meter data serves as a label. The model improves the visibility measurement accuracy of low signal-to-noise ratio by learning the mapping relationship between the input information and the corresponding label;

[0019] S2-3, define the loss function based on the mean absolute error between the denoised lidar visibility result predicted by the model and the forward scattering visibility meter visibility, specifically:

[0020] ;

[0021] Where N is the number of samples; is the model prediction value, is the tag value;

[0022] S2-4, taking the mean absolute error between the uncorrected data and the label as the baseline value, and setting the threshold of the loss function, if the loss value is higher than the preset threshold of the loss function, then updating the U-shaped network model.

[0023] As a preferred embodiment of the present invention, the step S3 specifically comprises: drawing a horizontal visibility distribution map based on the distance-resolved visibility information and the actual port channel position or road surface position.

[0024] As a preferred embodiment of the present invention: S4 specifically comprises: performing threshold segmentation on the obtained horizontal visibility distribution map using the OpenCV image processing library, extracting areas where visibility is lower than the threshold, and then using the edge detection algorithm, namely the Canny algorithm, to extract the boundaries of the low-visibility areas. The discrete edges detected by the Canny algorithm are then connected into continuous closed contours, and the contour geometry information is extracted to form a complete fog boundary.

[0025] As a preferred embodiment of the present invention, the OpenCV image processing specifically includes the following steps:

[0026] S4-1, mapping visibility values ​​to pixel values, pixel values ​​corresponding to visibility values, and the visibility range is dynamically adjusted according to real-time visibility; the highest visibility at any moment is 11 The pixel value is 255, which is the lowest visibility at all times The pixel value is 0;

[0027] S4-2, setting a visibility threshold, extracting areas where visibility is lower than the threshold, and generating a binary image, i.e., a visibility binary map, where low-visibility areas are white and other areas are black;

[0028] The visibility threshold is determined according to traffic regulations, and the pixel value threshold corresponding to the visibility threshold is obtained through the mapping relationship between visibility value and pixel value. ,

[0029] ;

[0030] in, is the visibility threshold;

[0031] S4-3, using the Canny algorithm to detect the boundary of the low-visibility area and generate a binary edge map; performing Canny edge detection on the binary image to extract the boundary of the low-visibility area;

[0032] S4-4, extracting the contours in the edge map and drawing the low visibility area boundary on the original image;

[0033] Use the contour extraction function in the OpenCV image processing library to extract the outer contour in the edge map and draw the red boundary of the low visibility area on the original image;

[0034] S4-5, shows the horizontal visibility distribution map after drawing the boundaries.

[0035] As a preferred embodiment of the present invention, in S5, the method for obtaining the visibility conditions of each waterway or road by identifying the low cloud and fog development path specifically includes the following steps:

[0036] S5-1. Coordinates represent the location of the waterway or road. Load and read the visibility binary image. Use the coordinate order of the image in OpenCV. Ensure that the binary image is a single-channel grayscale image when loading.

[0037] S5-2, traverse all waterway or road locations;

[0038] S5-3. Determine whether the pixel value at the waterway or road coordinate exceeds the pixel value threshold ; For each channel position, check its pixel value in the binary image, if it exceeds the pixel value threshold , then record the position;

[0039] S5-4. Output the course or position coordinates.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] Large-scale dynamic scanning of the fog changes within the port area can obtain visibility information at any point in the sea fog within the scanning range, and more finely observe the entire sea fog change process, which is conducive to providing more detailed guidance for navigation command. In addition, this application has high temporal and spatial resolution information capabilities, can accurately identify sea fog, and make precision corrections through the U-shaped network model, making the measurement data more accurate. Utilizing a wide-area laser system to quickly scan various channel areas of the port, after data processing and manual correction, a low-cloud fog development path identification map and high-precision visibility information with distance resolution are obtained. The navigation visibility threshold is used to determine whether a ship is navigable, and based on the real-time changing low-cloud fog development path identification map information, the port ship navigation can be flexibly and quickly directed, thereby improving the port's transportation capacity. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0043] Figure 2 A training flow chart of a U-shaped network model provided in an embodiment of the present invention;

[0044] Figure 3 OpenCV image processing flow chart provided by the embodiment of the present invention;

[0045] Figure 4 A flowchart of the reasoning of traffic location rules provided by an embodiment of the present invention;

[0046] Figure 5 A low cloud and fog development path identification map provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The present invention is further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these examples are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, modifications of various equivalent forms of the present invention made by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0048] like Figure 1-Figure 5 The following figure shows a low-cloud and fog recognition system based on wide-area laser scanning, which specifically includes the following steps:

[0049] S1, visibility information is obtained by inverting the range-resolved data from the rapid wide-area scanning of the lidar;

[0050] S2, trained based on forward scatter visibility meter and lidar historical data Type network model, and the visibility information inverted in S1 is used through the U-type network model to improve the measurement accuracy;

[0051] S3, drawing a horizontal visibility distribution map based on the visibility information after improving the measurement accuracy; S4, performing threshold segmentation, edge detection, and contour extraction on the obtained horizontal visibility distribution map using a low cloud and fog recognition algorithm to identify the low cloud and fog development path;

[0052] S5: Obtain and output the visibility conditions of each waterway or road based on the identified low cloud and fog development path.

[0053] The visibility information is retrieved from the distance-resolved data of the rapid wide-area scanning of the laser radar in S1 by inverting the radar signal information into the visibility information according to the horizontal visibility inversion algorithm. The horizontal visibility inversion algorithm is:

[0054] ;

[0055] ;

[0056] in, is the atmospheric extinction coefficient, For visibility, is the coefficient, obtained from the empirical formula, is the wavelength, and R is the detection distance.

[0057] In S2, the forward scattering visibility meter and lidar data are used for training. The specific network model is:

[0058] The visibility data of the forward scattering visibility meter and the visibility data of the visibility lidar are divided into a training set, a validation set and a test set according to a preset ratio. The visibility data of the forward scattering visibility meter is used as a label and the visibility data of the visibility lidar is used as input data. The model obtained by training is a denoising network. The denoising network achieves denoising by learning the difference characteristics between forward scattering visibility data and lidar visibility data, thereby improving the measurement accuracy of visibility with low signal-to-noise ratio.

[0059] Among them, the The network model training specifically includes the following steps:

[0060] S2-1: The lidar quickly scans a wide area to obtain raw signal data, and then uses the horizontal visibility inversion algorithm to invert the visibility information with distance resolution.

[0061] S2-2, inputting the inverted visibility information and the forward scatter visibility meter visibility information into a U-shaped network model in pairs, wherein the forward scatter visibility meter data serves as a label. The model improves the visibility measurement accuracy of low signal-to-noise ratio by learning the mapping relationship between the input information and the corresponding label;

[0062] S2-3, define the loss function based on the mean absolute error between the denoised lidar visibility result predicted by the model and the forward scattering visibility meter visibility, specifically:

[0063] ;

[0064] Where N is the number of samples; is the model prediction value, is the tag value;

[0065] S2-4, taking the mean absolute error between the uncorrected data and the label as the baseline value, setting the loss function threshold to 70% of the baseline value, and updating the U-shaped network model if the loss value is higher than the preset threshold of the loss function.

[0066] Drawing the horizontal visibility distribution map in S3 specifically includes drawing the horizontal visibility distribution map according to the visibility information resolved by distance and the actual port channel position or road surface position.

[0067] In S4, the obtained horizontal visibility distribution map is threshold segmented, edge detected, and contour extracted using a low-cloud fog recognition algorithm to identify the low-cloud fog development path. This step specifically includes: using the OpenCV image processing library to threshold segment the obtained horizontal visibility distribution map, extracting areas where visibility is below the threshold, then using the edge detection algorithm, namely the Canny algorithm, to extract the boundaries of the low-visibility areas, then connecting the discrete edges detected by the Canny algorithm into a continuous closed contour, and extracting contour geometry information to form a complete fog boundary. The low-cloud fog recognition algorithm specifically includes the following steps:

[0068] S4-1, mapping visibility values ​​to pixel values, pixel values ​​corresponding to visibility values, and the visibility range is dynamically adjusted according to real-time visibility; the highest visibility at any moment is The pixel value is 255, the lowest visibility at all times The pixel value is 0;

[0069] S4-2, setting a visibility threshold, extracting areas where visibility is lower than the threshold, and generating a binary image, i.e., a visibility binary map, where low-visibility areas are white and other areas are black;

[0070] The visibility threshold is determined according to traffic regulations, and the pixel value threshold corresponding to the visibility threshold is obtained through the mapping relationship between visibility value and pixel value. ,

[0071] ;

[0072] in, is the visibility threshold;

[0073] S4-3, using the Canny algorithm to detect the boundary of the low-visibility area and generate a binary edge map; performing Canny edge detection on the binary image to extract the boundary of the low-visibility area;

[0074] S4-4, extracting the contours in the edge map and drawing the low visibility area boundary on the original image;

[0075] Use the contour extraction function in the OpenCV image processing library to extract the outer contour in the edge map and draw the red boundary of the low visibility area on the original image;

[0076] S4-5, shows the horizontal visibility distribution map after drawing the boundary;

[0077] In S5, the method for obtaining the visibility conditions of each waterway or road by identifying the low cloud and fog development path specifically includes the following steps:

[0078] S5-1. Coordinates represent the location of the waterway or road, and load and read the visibility binary map;

[0079] Use the image coordinate order in OpenCV and ensure single-channel grayscale images when loading binary images.

[0080] S5-2, traverse all waterway or road locations;

[0081] S5-3, determining whether the pixel value at the waterway or road coordinate exceeds a pixel value threshold; for each waterway position, checking its pixel value in the binary image, and if it exceeds the pixel value threshold, recording the position;

[0082] S5-4. Output the course or position coordinates.

[0083] Large-scale dynamic scanning of the changing process of fog within the port area has the function of measuring heavy fog in the entire sea area compared to shore-based in-situ measurements. It can obtain visibility information at any point in the sea fog within the scanning range, and more finely observe the entire sea fog changing process, which is conducive to providing more detailed guidance for navigation command. Satellite observations are easily affected by clouds and convective weather, and the spatiotemporal resolution and accuracy of fog identification are not high. In comparison, the patent of this invention has high spatiotemporal resolution information and is more accurate in identifying sea fog. Traditional lidar visibility algorithms need to be corrected, and the accuracy is generally low. The patent of this invention uses a U-shaped network model and OpenCV image processing to correct the accuracy, making the measurement data more accurate.

[0084] This application uses a wide-area laser system to quickly scan various waterway areas of the port. After the above-mentioned data processing and manual correction, a low cloud and fog development path identification map and high-precision visibility information with distance resolution are obtained. The navigation visibility threshold is used to determine whether the ship can pass through. Based on the real-time changing low cloud and fog development path identification map information, the navigation of port ships can be flexibly and quickly directed, thereby improving the port's transportation capacity.

[0085] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A low cloud and fog recognition system based on wide-area laser scanning, characterized in that: The following steps are involved: S1, visibility information is obtained by inverting the distance-resolved data from the rapid wide-area scanning of the lidar; S2, trained based on forward scatter visibility meter and lidar historical data Type network model, and the visibility information inverted in S1 is used through the U-type network model to improve the measurement accuracy; S3, drawing a horizontal visibility distribution map based on the visibility information with improved measurement accuracy; S4, using a low cloud and fog recognition algorithm to perform threshold segmentation, edge detection, and contour extraction on the obtained horizontal visibility distribution map to identify the low cloud and fog development path; The low cloud and fog recognition algorithm is specifically as follows: S4-1, mapping the visibility value to the pixel value so that the pixel value corresponds to the visibility value, and the visibility range is dynamically adjusted according to the real-time visibility; S4-2, setting a visibility threshold, extracting areas where visibility is below the threshold, and generating a binary image, i.e., a visibility binary map. The pixel value threshold corresponding to the visibility threshold is obtained through the mapping relationship between visibility values ​​and pixel values ​​in S4-1; S4-3, detecting the boundary of the low visibility area and generating a binary edge map; S4-4, extracting the contour in the edge map and drawing the boundary of the low visibility area on the original image; S4-5 shows the horizontal visibility distribution map after drawing the boundary, and the low cloud and fog development path is obtained through this distribution map; S5: Obtain and output the visibility conditions of each waterway or road based on the identified low cloud and fog development path.

2. The low cloud and fog recognition system based on wide-area laser scanning according to claim 1, characterized in that: The visibility information is retrieved from the distance-resolved data of the rapid wide-area scanning of the laser radar in S1 specifically by inverting the radar signal information into visibility information according to a horizontal visibility inversion algorithm. The horizontal visibility inversion algorithm is: ； ； in, is the atmospheric extinction coefficient, For visibility, is the coefficient, obtained from the empirical formula, is the wavelength, R is the detection distance, k is the coefficient related to the lidar wavelength, aerosol refractive index and particle size distribution; Rm is the selected reference distance; S is the correction signal of the distance; Sm=S(Rm) is the signal corresponding to the reference distance; dR' is the differential of the variable R.

3. The low cloud and fog recognition system based on wide-area laser scanning according to claim 1, characterized in that: The S2 is specifically: The visibility data of the forward scattering visibility meter and the visibility data of the visibility lidar are divided into a training set, a validation set and a test set according to a preset ratio. The visibility data of the forward scattering visibility meter is used as a label and the visibility data of the visibility lidar is used as input data. The visibility data is used to train the Type network model, the model obtained by training is a denoising network.

4. The low cloud and fog recognition system based on wide-area laser scanning according to claim 3, characterized in that: described The network model training specifically includes the following steps: S2-1: The lidar quickly scans a wide area to obtain raw signal data, and then uses the horizontal visibility inversion algorithm to invert the visibility information with distance resolution. S2-2, inputting the inverted visibility information and the forward scatter visibility meter visibility information into a U-shaped network model in pairs, wherein the forward scatter visibility meter data serves as a label. The model improves the visibility measurement accuracy of low signal-to-noise ratio by learning the mapping relationship between the input information and the corresponding label; S2-3, define the loss function based on the mean absolute error between the denoised lidar visibility result predicted by the model and the forward scattering visibility meter visibility, specifically: ； Where N is the number of samples; is the model prediction value, is the tag value; S2-4, taking the mean absolute error between the uncorrected data and the label as the baseline value, and setting the threshold of the loss function, if the loss value is higher than the preset threshold of the loss function, then updating the U-shaped network model.

5. The low cloud and fog recognition system based on wide-area laser scanning according to claim 1, characterized in that: The step S3 specifically includes drawing a horizontal visibility distribution map based on the visibility information resolved by distance and the actual port channel position or road surface position.

6. The low cloud and fog recognition system based on wide-area laser scanning according to claim 1, characterized in that: The S4 is specifically as follows: The obtained horizontal visibility distribution map is threshold segmented using the OpenCV image processing library to extract areas with visibility below the threshold. The edge detection algorithm, namely the Canny algorithm, is then used to extract the boundaries of the low-visibility areas. The discrete edges detected by the Canny algorithm are then connected into continuous closed contours, and the contour geometric information is extracted to form a complete fog boundary.

7. The low cloud and fog recognition system based on wide-area laser scanning according to claim 6, characterized in that: In S5, the method for obtaining the visibility conditions of each waterway or road by identifying the low cloud and fog development path specifically includes the following steps: S5-1. Coordinates represent the location of the waterway or road. Load and read the visibility binary image. Use the coordinate order of the image in OpenCV. Ensure that the binary image is a single-channel grayscale image when loading. S5-2, traverse all waterway or road locations; S5-3. Determine whether the pixel value at the waterway or road coordinate exceeds the pixel value threshold ; For each channel position, check its pixel value in the binary image, if it exceeds the pixel value threshold , then record the position; S5-4. Output the course or position coordinates.

Images