GNSS data quality prediction method and device based on fisheye camera, and medium

Abstract

The invention provides a GNSS data quality prediction method and device based on a fisheye camera, and a medium, and belongs to the technical field of data processing, and the method comprises the steps: determining the satellite orbit information of a future time period with a to-be-evaluated observation station as a reference point based on ephemeris prediction; according to the fisheye sky image shot at the to-be-evaluated observation station and the satellite orbit information, determining satellite feature data of a future time period and the occlusion porosity of the future time period; respectively inputting the occlusion porosity of the future time period and at least part of satellite feature data of the future time period into a general prediction model and an observation station prediction model corresponding to the observation station to be evaluated for data quality evaluation, and obtaining a first data quality evaluation index and a second data quality evaluation index; and determining a data quality prediction result based on the first data quality evaluation index and the second data quality evaluation index. A general prediction model and an observation station exclusive prediction model are introduced to carry out double-model evaluation, and accurate prediction of GNSS observation data quality is realized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, device and medium for predicting GNSS data quality based on a fisheye camera. Background Technology

[0002] As the application of Global Navigation Satellite System (GNSS) continues to deepen in fields such as traffic navigation, geographic information collection, disaster monitoring, and autonomous driving, the demand for high-precision GNSS positioning in complex environments such as cities, mountains, and forests is becoming increasingly urgent. Ensuring the quality of GNSS observation data is the core prerequisite for achieving high-precision positioning. Therefore, there is an urgent need for efficient and accurate GNSS observation data quality prediction schemes to meet the practical application needs such as site selection and real-time observation quality assessment in complex scenarios.

[0003] Traditional GNSS data quality analysis methods primarily rely on long-term field observation data, assessing data quality through statistical analysis or long-term smoothing. These methods struggle to reflect actual data quality in short-term field observation environments, exhibiting slow response times and failing to meet the demands of rapid site selection and field assessment. Summary of the Invention

[0004] This invention provides a GNSS data quality prediction method, device, and medium based on a fisheye camera, to solve the problems of slow response speed, difficulty in timely reflecting actual data quality, and inability to meet the needs of rapid site selection and on-site assessment in short-term field observation environments.

[0005] This invention provides a method for predicting the quality of GNSS data based on a fisheye camera, comprising the following steps: Based on ephemeris forecasts, satellite orbit information for future periods is determined with the observation station to be evaluated as a reference point; Satellite characteristic data and occlusion porosity for future periods are determined based on fisheye sky images taken at the observation station to be evaluated and the satellite orbit information. The occlusion porosity of the future time period and at least part of the satellite feature data of the future time period are input into a general prediction model for data quality assessment, and a first data quality assessment index output by the general prediction model is obtained. The occlusion porosity of the future time period and at least part of the satellite feature data of the future time period are input into the observation station prediction model corresponding to the observation station to be evaluated for data quality assessment, and a second data quality assessment index output by the observation station prediction model is obtained. The data quality prediction result is determined based on the first data quality assessment index and the second data quality assessment index.

[0006] The GNSS data quality prediction method based on a fisheye camera provided by this invention, wherein determining the data quality prediction result based on the first data quality assessment index and the second data quality assessment index includes: Determine the first weight value corresponding to the first data quality assessment indicator and the second weight value corresponding to the second data quality assessment indicator; Determine the first product value of the first data quality assessment index and the first weight value; Determine the second product value of the second data quality assessment index and the second weight value; The sum of the first product value and the second product value is used as the data quality prediction result; Wherein, both the first data quality assessment indicator and the second data quality assessment indicator include one or a combination of the following: Carrier-to-noise ratio, pseudorange multipath; The satellite feature data includes one or a combination of the following: Satellite elevation angle and azimuth angle.

[0007] The GNSS data quality prediction method based on fisheye cameras provided by this invention, wherein the general prediction model is trained through the following steps: A first training set is determined, which includes a first combination of samples and data quality evaluation index labels corresponding to the first combination of samples. The first combination of samples includes a first satellite elevation angle and a first occlusion porosity. The first satellite elevation angle is a satellite elevation angle determined based on GNSS observation data under various occlusion environments. The first occlusion porosity is an occlusion porosity determined by image recognition of a first fisheye image. The first fisheye image is a fisheye image captured synchronously with GNSS observation data. The first combined sample is input into the initial prediction model to obtain the first prediction data quality evaluation index output by the initial prediction model. The first loss value is determined based on the data quality assessment index label corresponding to the first combined sample and the first predicted data quality assessment index. The trainable parameters in the initial prediction model are updated based on the first loss value; The observation station prediction model is trained through the following steps: A second training set is determined, which includes a second combined sample and data quality assessment index labels corresponding to the second combined sample. The second combined sample includes a second satellite elevation angle, a first azimuth angle, and a second occlusion porosity. The second satellite elevation angle is a satellite elevation angle determined based on short-time GNSS observation data. The first azimuth angle is an azimuth angle determined based on short-time GNSS observation data. The second occlusion porosity is an occlusion porosity determined by image recognition of a second fisheye image. The second fisheye image is a fisheye image that is synchronously captured with the short-time GNSS observation data at the observation station to be evaluated. The second combined sample is input into the initial observation station prediction model to obtain the second prediction data quality evaluation index output by the initial observation station prediction model. The second loss value is determined based on the data quality assessment index label corresponding to the second combined sample and the second predicted data quality assessment index. The trainable parameters in the initial observation station prediction model are updated based on the second loss value.

[0008] The GNSS data quality prediction method based on a fisheye camera provided by this invention, wherein the satellite orbit information includes the satellite position, and the occlusion porosity for future time periods is obtained through the following steps: The satellite position is projected onto the fisheye sky image to obtain the satellite pixel coordinates; Based on the satellite pixel coordinates, the satellite position is determined to be within the visible sky area, and the occlusion porosity is defined as 1. Based on the satellite pixel coordinates, the satellite position is determined to be in a building obstruction area, and the obstruction porosity is defined as -1; Based on the satellite pixel coordinates, the satellite position is determined to be in a tree-shaded area, and the shading porosity is defined as a first ratio; Wherein, the first ratio is the ratio of visible sky pixels to total pixels in the first region, and the first region is the region defined by the satellite pixel coordinates as the center and a radius of 100 pixels.

[0009] The present invention provides a GNSS data quality prediction method based on a fisheye camera, wherein the satellite position includes the satellite elevation angle and azimuth angle, and the satellite position is projected onto the fisheye sky image to obtain the satellite pixel coordinates, including: Obtain the projection model of the fisheye camera; Based on the projection model, the satellite elevation angle and azimuth angle are converted into pixel coordinates to obtain the satellite pixel coordinates.

[0010] The GNSS data quality prediction method based on a fisheye camera provided by this invention further includes: The total number of theoretically visible satellites and the total number of actually visible satellites are determined based on the occlusion type determined by the satellite pixel coordinates and the satellite feature data for future time periods. The data integrity rate is determined based on the ratio of the actual total number of visible satellites to the theoretical total number of visible satellites.

[0011] The GNSS data quality prediction method based on a fisheye camera provided by this invention further includes: The design matrix is ​​determined based on the unit vector of each first-type satellite relative to the observation station to be evaluated. The first-type satellite is a satellite whose elevation angle is higher than the elevation angle threshold determined by the observation station to be evaluated and whose occlusion type is visible or conditionally visible. Determine the cofactor matrix based on the design matrix; The position accuracy factor is determined based on the elements in the cofactor matrix.

[0012] This invention provides a GNSS data quality prediction device based on a fisheye camera, comprising the following modules: The first determination module is used to determine the satellite orbit information for a future period based on ephemeris forecasts, with the observation station to be evaluated as the reference point. The second determining module is used to determine satellite characteristic data and occlusion porosity for future periods based on fisheye sky images taken at the observation station to be evaluated and the satellite orbit information. The first evaluation module is used to input the occlusion porosity of the future time period and at least part of the satellite feature data of the future time period into a general prediction model to evaluate the data quality and obtain the first data quality evaluation index output by the general prediction model. The second evaluation module is used to input the occlusion porosity of the future time period and at least part of the satellite feature data of the future time period into the observation station prediction model corresponding to the observation station to be evaluated for data quality evaluation, and obtain the second data quality evaluation index output by the observation station prediction model. The prediction module is used to determine the data quality prediction result based on the first data quality assessment index and the second data quality assessment index.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the GNSS data quality prediction method based on a fisheye camera as described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the GNSS data quality prediction method based on a fisheye camera as described above.

[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the GNSS data quality prediction method based on a fisheye camera as described above.

[0016] This invention provides a method, apparatus, electronic device, and storage medium for predicting GNSS data quality based on a fisheye camera. By combining satellite orbit information from ephemeris predictions and occlusion information from fisheye sky images, and simultaneously introducing a general prediction model and an observation station-specific prediction model for dual-model evaluation, this invention retains the adaptability of the general model to various scenarios while also taking into account the accuracy of the observation station-specific model for specific environments. Ultimately, it achieves efficient and accurate prediction of GNSS observation data quality, solving the problems of traditional methods relying on long-term observations and slow response. It can meet the needs of rapid site selection and real-time quality assessment in complex environments. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is one of the flowcharts of the GNSS data quality prediction method based on a fisheye camera provided by the present invention; Figure 2 This is a flowchart illustrating the process of determining data quality prediction results based on a first data quality assessment index and a second data quality assessment index, provided by the present invention. Figure 3 This is the second flowchart of the GNSS data quality prediction method based on fisheye camera provided by the present invention; Figure 4 This is a schematic diagram of the image recognition result provided by the image recognition algorithm of this invention; Figure 5 This is a schematic diagram of the fitting effect of SNR1 of the G01 satellite provided by the present invention; Figure 6 This is a schematic diagram of the fitting effect of SNR2 of the G01 satellite provided by the present invention; Figure 7 This is a schematic diagram of the fitting effect of MP1 of the C28 satellite provided by the present invention; Figure 8 This is a schematic diagram of the fitting effect of MP2 of the C28 satellite provided by the present invention; Figure 9 This is a schematic block diagram of the GNSS data quality prediction device based on a fisheye camera provided by the present invention; Figure 10 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] It should be noted that in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. The terms "upper," "lower," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly, for example, as a fixed connection, a detachable connection, or an integral connection; a mechanical connection or an electrical connection; a direct connection or an indirect connection through an intermediate medium; or a connection within two elements. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0021] The terms "first," "second," etc., used in this invention are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0022] The following is combined with Figures 1-10 This invention describes a GNSS data quality prediction method, device, and medium based on a fisheye camera, which aims to solve the problems of slow response speed, difficulty in timely reflecting actual data quality, and inability to meet the needs of rapid site selection and on-site assessment in short-term field observation environments.

[0023] Figure 1 This is one of the flowcharts illustrating the GNSS data quality prediction method based on a fisheye camera provided by the present invention, such as... Figure 1 As shown, including but not limited to the following steps: Step 101: Determine the satellite orbit information for future periods based on ephemeris forecasts, using the observation station to be evaluated as a reference point.

[0024] Ephemeris forecasts are obtained through official service interfaces of satellite navigation systems and dedicated ephemeris data receiving equipment. These forecasts contain basic information such as satellite orbital parameters. Subsequently, using the geographical location (such as latitude, longitude, and elevation) of the observation station to be evaluated as a reference point, and combining the satellite orbital parameters from the ephemeris forecasts, orbital calculation algorithms are used to calculate the orbital information of each satellite relative to the observation station in future time periods (such as the next 24 hours, the next 7 days, etc.). The satellite orbital information may include data such as the satellite's real-time position, trajectory, elevation angle change trend, and azimuth angle change trend, providing a basis for subsequent data quality prediction based on the spatial distribution of satellites.

[0025] Step 102: Determine satellite characteristic data and occlusion porosity for future periods based on fisheye sky images taken at the observation station to be evaluated and satellite orbit information.

[0026] A fisheye camera is deployed at the observation station to be evaluated. This fisheye camera captures a panoramic fisheye sky image above the station, providing a complete view of the visible sky area, building obstructions, tree obstructions, and other environmental occlusions around the station. On one hand, based on the satellite orbit information obtained in step 101, satellite feature data for each satellite relative to the observation station to be evaluated is extracted for future time periods. This satellite feature data reflects the satellite's spatial orientation attributes. On the other hand, the satellite's spatial position from its orbit information is projected onto the fisheye sky image. Image recognition and analysis techniques are used to determine the occlusion porosity corresponding to the satellite's projected position. This occlusion porosity quantifies the degree of obstruction along the satellite signal propagation path. Specifically, a deep learning algorithm combining color and edge feature extraction is first used to divide the fisheye sky image into regions, distinguishing the visible sky area, building obstruction areas, and tree obstruction areas. Then, the satellite position is mapped to the corresponding region, thereby determining the occlusion porosity of the satellite's position for the future time period.

[0027] Step 103: Input the occlusion porosity of the future time period and at least part of the satellite feature data of the future time period into the general prediction model for data quality assessment, and obtain the first data quality assessment index output by the general prediction model.

[0028] A pre-trained general prediction model is generated based on GNSS observation data under various occlusion environments and synchronous fisheye images, enabling it to assess the quality of GNSS data in general scenarios. The occlusion porosity data for future periods obtained in step 102, along with at least some satellite feature data, are input into this general prediction model. The model processes the input feature data based on its internal computational logic and outputs a corresponding first data quality assessment index. This index can preliminarily reflect the GNSS data quality level of the observation station to be evaluated in the future period.

[0029] Step 104: Input the occlusion porosity of the future time period and at least part of the satellite feature data of the future time period into the observation station prediction model corresponding to the observation station to be evaluated for data quality assessment, and obtain the second data quality assessment index output by the observation station prediction model.

[0030] For each observation station to be evaluated, a dedicated prediction model is pre-trained. This model is based on short-term GNSS observation data and synchronous fisheye images collected on-site at the observation station and is adapted to the specific environmental characteristics of that station. The occlusion porosity for future periods obtained in step 102, along with at least some satellite feature data, is input into the prediction model. The prediction model then performs calculations on the input data in conjunction with the environmental characteristics of the observation station, outputting a second data quality assessment index. This index accurately reflects the GNSS data quality under the specific environment of the observation station.

[0031] Step 105: Determine the data quality prediction result based on the first data quality assessment index and the second data quality assessment index.

[0032] By integrating two indicators using a pre-defined fusion strategy, the predicted GNSS data quality for the future time period of the observation station under evaluation is finally determined. The fusion strategy can be flexibly selected according to actual application needs, as long as it can integrate the evaluation advantages of the two models and output reliable prediction results.

[0033] In this embodiment, by combining satellite orbit information from ephemeris predictions and occlusion information from fisheye sky images, and simultaneously introducing a general prediction model and an observation station-specific prediction model for dual-model evaluation, the general model's adaptability to various scenarios is preserved, while the observation station-specific model's accuracy for specific environments is also taken into account. Ultimately, efficient and accurate prediction of GNSS observation data quality is achieved, solving the problems of traditional methods relying on long-term observations and slow response. This approach can meet the needs of rapid site selection and real-time quality assessment in complex environments.

[0034] In some embodiments, such as Figure 2 As shown, the data quality prediction results are determined based on the first data quality assessment index and the second data quality assessment index, including: Step 201: Determine the first weight value corresponding to the first data quality assessment indicator and the second weight value corresponding to the second data quality assessment indicator; Step 202: Determine the first product value of the first data quality assessment index and the first weight value; Step 203: Determine the second product value of the second data quality assessment index and the second weight value; Step 204: The sum of the first product value and the second product value is used as the data quality prediction result.

[0035] In this embodiment, weight values ​​can be configured for the first and second data quality assessment indicators based on factors such as the actual application scenario, the fitting effect of the basic model, and the environmental complexity of the observation station to be evaluated. These weight values ​​are designated as the first weight value and the second weight value. The first weight value corresponds to the output of the general prediction model, and the second weight value corresponds to the output of the observation station's prediction model. The range of the two weight values ​​is typically [0, 1], and their sum can be set to 1 or other reasonable values ​​according to actual needs. For example, in scenarios where the general prediction model has a good fitting effect, the first weight value can be configured as 0.8 and the second weight value as 0.2. If the environment of the observation station to be evaluated has strong specificity, and the targeted advantage of the observation station's prediction model is more obvious, the second weight value can be appropriately increased, such as setting both the first and second weight values ​​to 0.5, thus balancing the universality of the general model and the accuracy of the specific model.

[0036] By fusing the evaluation metrics of the two models through weighted summation, the proportion of the general prediction model and the observation station prediction model in the final result can be flexibly adjusted. This approach ensures the universality and stability of the prediction results by relying on the general model, while also enhancing the adaptability to specific environments by leveraging the observation station's dedicated model. This effectively improves the accuracy and reliability of the data quality prediction results. Furthermore, this fusion method is simple to operate, has high computational efficiency, and can meet the needs of real-time evaluation.

[0037] In some embodiments, data quality prediction results include: SNRform=ω1⋅SNRbaxe+(1-ω1)SNRcorr; MPfinal=ω2⋅MPbase+(1-ω2)MPcorr; Wherein, SNRfinal, SNRbase, and SNRcorr are the carrier-to-noise ratio (CNR) in the data quality prediction result, the CNR in the first data quality assessment metric, and the CNR in the second data quality assessment metric, respectively. MPfinal, MPbase, and MPcorr are the pseudorange multipath in the data quality prediction result, the pseudorange multipath in the first data quality assessment metric, and the pseudorange multipath in the second data quality assessment metric, respectively. ω1 and ω2 represent the weighting coefficients of the CNR and pseudorange multipath, which are set to 0.8 and 0.5, respectively, based on the fitting effect of the base model.

[0038] In some embodiments, the carrier-to-noise ratio can be directly extracted from the observation data, and the pseudorange multipath is calculated as follows: ; in, Representing different signal frequencies, It is a pseudorange multipath combination, where P and L are the pseudorange and phase observations, and f is the frequency of the corresponding signal. After obtaining the pseudorange multipath combination for each epoch, the average value of the multipath in the sliding window is calculated using the moving average method. At this point, the pseudorange multipath value for a certain epoch is the pseudorange multipath combination value of the current epoch minus the average value of the window.

[0039] In some embodiments, both the first data quality assessment metric and the second data quality assessment metric include one or a combination of the following: Carrier-to-noise ratio, pseudorange multipath; Satellite signature data includes one or a combination of the following: Satellite elevation angle and azimuth angle.

[0040] The carrier-to-noise ratio (CNR) is a key metric that directly reflects the ratio of GNSS satellite signal strength to noise. It is a core indicator for assessing signal quality and can be directly extracted from GNSS observation files. In the evaluation process of general prediction models and station-based prediction models, the CNR reflects the stability of the satellite signal; a higher CNR value generally indicates lower signal interference and better observation data quality.

[0041] Pseudorange multipath: This metric measures the impact of multipath errors caused by GNSS signals reflected from surrounding obstacles on pseudorange observations. It is calculated by combining pseudorange and phase observations. Specifically, the combined pseudorange multipath value is first calculated, and then the trend term is eliminated using a moving average method to obtain the final pseudorange multipath value. The magnitude of this metric directly relates to the accuracy of the positioning results; a smaller value indicates weaker multipath interference.

[0042] Satellite elevation angle: refers to the angle between the satellite and the observation plane of the station to be evaluated. It can be calculated based on the satellite orbit information predicted by ephemeris. This indicator directly affects the propagation path length of the satellite signal and the probability of being blocked. Generally, the higher the elevation angle, the lower the possibility of the signal being blocked by ground obstacles.

[0043] Azimuth: refers to the horizontal angle between the satellite and the north direction of the observation station being evaluated. It can reflect the horizontal distribution of the satellite in the sky above the observation station. Combining the azimuth angle can further refine the impact of obstruction at different azimuths of the station on the satellite signal.

[0044] In this embodiment, by clearly defining the data quality assessment indicators and the specific types of satellite feature data, standardized and quantifiable input features and output indicators are provided for the evaluation process of the dual models. This ensures the standardization and repeatability of model operations, and allows for flexible combination of indicator types according to the needs of different complex scenarios. This improves the adaptability and practicality of the data quality prediction scheme, and also makes the interpretation of prediction results more targeted and professional.

[0045] In some embodiments, the general prediction model is trained through the following steps: A first training set is determined. The first training set includes a first combination of samples and data quality evaluation index labels corresponding to the first combination of samples. The first combination of samples includes a first satellite elevation angle and a first occlusion porosity. The first satellite elevation angle is a satellite elevation angle determined based on GNSS observation data under various occlusion environments. The first occlusion porosity is the occlusion porosity determined by image recognition of a first fisheye image. The first fisheye image is a fisheye image captured synchronously with the GNSS observation data.

[0046] Specifically, a first training set needs to be constructed for training a general prediction model. This training set contains a large number of first combination samples and data quality evaluation index labels corresponding to each first combination sample.

[0047] For the first sample set, its core components are the first satellite elevation angle and the first occlusion porosity. The first satellite elevation angle is determined based on GNSS observation data under various occlusion environments. Specifically, long-term GNSS observation data can be collected from multiple observation stations in different occlusion scenarios such as cities, forest areas, mountains, and open plains. Then, based on the GNSS data processing algorithm, the elevation angle of each satellite relative to the corresponding station within each observation period is calculated, i.e., the first satellite elevation angle. The first occlusion porosity is determined after image recognition of the first fisheye image. The first fisheye image is a panoramic sky image of the station taken synchronously with the GNSS observation data. After acquiring the image, a deep learning recognition algorithm combined with color and edge feature extraction can be used to first divide the image into regions, distinguishing the visible sky area, building occlusion area, and tree occlusion area. Then, the satellite position is projected onto the fisheye image, and the occlusion porosity at the corresponding position is calculated according to preset rules, i.e., the first occlusion porosity.

[0048] For the data quality assessment index label, the true index value of GNSS data quality for the time period corresponding to the first combination sample should be selected. Specifically, it can include a combination of one or more indices such as carrier-to-noise ratio and pseudorange multipath. The carrier-to-noise ratio can be directly extracted from the GNSS observation file, and the pseudorange multipath can be calculated by combining pseudorange observations and phase observations with the moving average method.

[0049] In practice, data can be collected from five or more stations with different occlusion types. For example, one observation station can be set up in densely populated urban areas, dense forest areas, mountainous areas, semi-occluded suburbs, and open plains. Each station can collect at least one day of GNSS observation data and synchronous fisheye images. Then, a sufficient number of first combination samples and corresponding labels can be extracted from them to form the first training set covering multiple scenarios.

[0050] The first combination of samples is input into the initial prediction model to obtain the first prediction data quality evaluation index output by the initial prediction model.

[0051] In some embodiments, a random forest model is selected as the initial prediction model (other machine learning models such as neural networks and support vector machines can also be selected as needed). The first combination sample (first satellite elevation angle + first occlusion porosity) in the first training set is input into the initial prediction model one by one. The initial prediction model will process the input feature data according to its own network structure or algorithm logic and output the corresponding first prediction data quality evaluation index. This index is the model's preliminary evaluation result of the sample data quality.

[0052] The first loss value is determined based on the data quality assessment index label corresponding to the first combined sample and the first predicted data quality assessment index.

[0053] To measure the deviation between the output of the initial prediction model and the actual situation, a first loss value needs to be calculated. Specifically, the data quality assessment index label (true value) corresponding to the first combination sample can be compared with the first prediction data quality assessment index (predicted value) output by the initial prediction model. An appropriate loss function is selected based on the model type to calculate the deviation. For example, for a regression model, the mean squared error loss function or the mean absolute error loss function can be used; for a classification model, the cross-entropy loss function can be used. The result obtained through the loss function calculation is the first loss value. The magnitude of the first loss value directly reflects the accuracy of the model's current prediction results; a larger value indicates a greater deviation.

[0054] The trainable parameters in the initial prediction model are updated based on the first loss value.

[0055] After obtaining the first loss value, the trainable parameters in the initial prediction model are iteratively updated along the direction of decreasing loss value using parameter optimization algorithms such as gradient descent. For example, in a random forest model, parameters such as the number of decision trees, the depth of the decision trees, and the feature selection strategy can be adjusted; in a neural network model, the weights and biases of each layer can be updated. After each parameter update, the first combination of samples needs to be input into the model again and the loss value calculated. This iterative process is repeated until the first loss value decreases to a preset threshold, or the number of model iterations reaches a set upper limit. At this point, the trained initial prediction model becomes a general prediction model that can be used for evaluation in general scenarios.

[0056] In this embodiment, a general prediction model is trained based on sample data from multiple scenarios and types of occlusion environments. This enables the model to have universal evaluation capabilities for different GNSS observation scenarios. With satellite elevation angle and occlusion porosity as core input features, it can accurately capture the impact of satellite spatial orientation and environmental occlusion on data quality. At the same time, the model's prediction accuracy is ensured through loss-value-driven parameter iteration updates, laying a solid foundation for high-precision data quality prediction by combining it with station-specific models.

[0057] In some embodiments, the observation station prediction model is trained through the following steps: A second training set is determined, which includes a second set of combined samples and corresponding data quality assessment index labels. The second set of combined samples includes a second satellite elevation angle, a first azimuth angle, and a second occlusion porosity. The second satellite elevation angle is determined based on short-time GNSS observation data, the first azimuth angle is determined based on short-time GNSS observation data, and the second occlusion porosity is determined by image recognition of the second fisheye image. The second fisheye image is a fisheye image that is synchronously captured at the observation station to be evaluated with the short-time GNSS observation data.

[0058] A second training set needs to be constructed for training the observation station prediction model. This training set contains several second combination samples and data quality assessment index labels corresponding to each second combination sample.

[0059] For the second sample set, its core components are the second satellite elevation angle, the first azimuth angle, and the second occlusion porosity. The second satellite elevation angle and the first azimuth angle are determined based on short-term GNSS observation data from the station to be evaluated. Specifically, short-term GNSS observation data of 15-30 minutes can be collected on-site at the station (long-term data is sufficient for training). Then, a GNSS data calculation algorithm is used to calculate the elevation angle (i.e., the second satellite elevation angle) and azimuth angle (i.e., the first azimuth angle) of each satellite relative to the station within this short observation period. The second occlusion porosity is determined after image recognition of the second fisheye image. The second fisheye image is a panoramic fisheye image of the sky taken simultaneously with the aforementioned short-term GNSS observation data at the station to be evaluated. The same recognition algorithm as the first fisheye image (deep learning recognition combined with color and edge feature extraction) can be used to complete image region segmentation and satellite projection position matching, thereby calculating the occlusion porosity (i.e., the second occlusion porosity) at the corresponding position.

[0060] For the data quality assessment index labels, the actual GNSS data quality index values ​​within the short-term observation period corresponding to the second combination of samples should be selected. Specifically, they can include a combination of one or more indicators such as carrier-to-noise ratio, pseudorange multipath, data integrity rate, and position accuracy factor. The acquisition and calculation methods are the same as those for the data quality assessment index labels. That is, the carrier-to-noise ratio is directly extracted from the observation file, the pseudorange multipath is obtained by combining pseudorange and phase observation values ​​with the moving average method, and the data integrity rate and position accuracy factor are calculated by combining satellite orbit and obstruction conditions.

[0061] In practice, short-term data collection can be completed at the observation station to be evaluated, and then a sufficient number of second-group samples and corresponding labels can be extracted from them to form a second training set that fits the actual environment of the station.

[0062] The second combined sample is input into the initial observation station prediction model to obtain the second prediction data quality evaluation index output by the initial observation station prediction model.

[0063] A suitable initial observation station prediction model is selected. In this embodiment, a function fitting model is preferred (lightweight neural networks, gradient boosting trees, etc. can also be selected as needed). The second combination samples (second satellite elevation angle + first azimuth angle + second occlusion porosity) in the second training set are input one by one into the initial observation station prediction model. The initial observation station prediction model will process the input specific feature data according to its own algorithm logic and output the corresponding second prediction data quality evaluation index. This index is the model's preliminary evaluation result of the data quality under the specific scenario of the station.

[0064] The second loss value is determined based on the data quality assessment index label corresponding to the second combined sample and the second predicted data quality assessment index.

[0065] The trainable parameters in the initial observation station prediction model are updated based on the second loss value.

[0066] In this embodiment, a dedicated observation station prediction model is trained based on short-term field data of the observation station to be evaluated. This allows the model to accurately adapt to the unique environment of the station without relying on long-term observation data. By introducing satellite azimuth as an additional input feature, it can more meticulously capture the impact of different azimuth occlusions on data quality. At the same time, the loss-value-driven parameter iteration update ensures the model's accuracy in evaluating the data quality of the station. When used in conjunction with a general prediction model, it can effectively compensate for the prediction bias of the general model in specific scenarios, and significantly improve the accuracy and relevance of the overall data quality prediction.

[0067] In some embodiments, satellites are classified into three types of obstructed areas: sky, trees, and buildings, based on the obstruction porosity. For each type of obstructed area, a function fitting algorithm is used, with satellite elevation angle (ele), azimuth angle (azi), and obstruction porosity as independent variables, and actual observed carrier-to-noise ratio and pseudorange multipath as dependent variables, to perform regression analysis and obtain the functional relationship and coefficients of the initial observation station prediction model.

[0068] In some embodiments, the initial observation prediction model for each type of occlusion region has the following functional forms for the carrier-to-noise ratio and pseudorange multipath: SNRατ=a·ele+b·azi+c·porosity; MFam=a+b·porosity+c·ele+d·ele 2 ; Where a, b, c, and d represent model coefficients, SNRατ is a function of carrier-to-noise ratio, and MFam is a function of pseudorange multipath.

[0069] In some embodiments, satellite orbital information includes satellite position, and the occlusion porosity for future periods is obtained through the following steps: The satellite position is projected onto the fisheye sky image to obtain the satellite pixel coordinates.

[0070] The satellite's position relative to the observation station to be evaluated is obtained within a future time period. This position reflects the satellite's spatial orientation in the sky. Subsequently, a spatial coordinate projection algorithm is used to map the satellite's actual spatial position onto the two-dimensional pixel plane of the fisheye sky image already captured by the observation station to be evaluated. This yields the satellite's pixel coordinates in the fisheye image, which can accurately pinpoint the satellite's specific location in the image, providing a basis for subsequent determination of the type of occlusion area.

[0071] The satellite's position is determined to be within the visible sky area based on the satellite pixel coordinates, and the occlusion porosity is defined as 1.

[0072] The satellite's location is determined to be within a building-obscured area based on its pixel coordinates, with the obscuration porosity defined as -1.

[0073] The satellite's location is determined to be within a tree-shaded area based on its pixel coordinates, and the shading porosity is defined as the first ratio.

[0074] The first ratio is the ratio of visible sky pixels to total pixels in the first region, and the first region is a region defined by satellite pixel coordinates as the center and a radius of 100 pixels.

[0075] In this embodiment, after obtaining the satellite pixel coordinates, the region type of the satellite pixel coordinates is determined based on the region identification results of the fisheye sky image (the visible sky region, building occlusion region, and tree occlusion region have been divided through deep learning combined with color and edge feature extraction algorithms). Then, the corresponding occlusion porosity is configured for different regions according to preset rules, as follows: If the satellite pixel coordinates are determined to be in the visible area of ​​the sky, it means that the satellite signal propagation path is unobstructed and the signal can directly reach the observation station to be evaluated. At this time, the obstruction porosity corresponding to the satellite is defined as 1, which characterizes the ideal observation conditions without obstruction. If the satellite pixel coordinates are determined to be in a building-blocked area, it means that the satellite signal will be completely blocked or severely blocked by the building, and the observation station cannot effectively receive the satellite signal. In this case, the occlusion porosity corresponding to the satellite is defined as -1 to characterize the severe observation conditions of complete blockage. If the satellite pixel coordinates indicate that the area is obscured by trees, the satellite signal may penetrate through gaps in the tree branches and leaves (i.e., sky gaps). Therefore, it is necessary to quantify the transparency of these gaps. Specifically, a circular area with a radius of 100 pixels is defined as the first region, centered on the satellite pixel coordinates. The number of visible sky pixels within this first region is counted, along with the total number of pixels in the region. The ratio between these two values ​​(i.e., the first ratio) is then defined as the satellite's obstruction porosity at this point. A higher ratio indicates a larger gap in the tree-obscured area, and thus a stronger satellite signal penetration capability.

[0076] In practical applications, the rationality of this porosity definition can be verified in specific scenarios. For example, at an urban forest observation station, if the pixel coordinates of a satellite fall into the area blocked by trees, and there are 300 visible pixels in the sky within a 100-pixel radius area, with a total of 1000 pixels, then the first ratio is 0.3, which means that the porosity of the satellite is 0.3, reflecting the actual degree to which the satellite signal is blocked by trees.

[0077] By projecting satellite positions onto fisheye images, precise matching between satellite spatial positions and ground-based occlusion environments was achieved. Furthermore, differentiated porosity definition rules were established for different types of occlusion areas. This enabled rapid identification of unobstructed and fully occluded scenes, and quantified the porosity transparency of tree-occluded areas through pixel proportions. This made the occlusion porosity index more closely reflect the actual observation environment, providing high-precision environmental occlusion feature data for subsequent model evaluation and effectively improving the accuracy of data quality predictions.

[0078] In some embodiments, the satellite position includes the satellite elevation angle and azimuth angle. Projecting the satellite position onto a fisheye sky image to obtain the satellite pixel coordinates includes: Obtain the projection model of the fisheye camera; Based on the projection model, the satellite elevation angle and azimuth angle are converted into pixel coordinates to obtain the satellite pixel coordinates.

[0079] In this embodiment, by first acquiring the fisheye camera's dedicated projection model and then performing coordinate transformation, the problem of satellite position projection deviation caused by fisheye camera imaging distortion is solved. This achieves accurate mapping of satellite elevation angle and azimuth angle to fisheye image pixel coordinates, providing a high-precision coordinate basis for accurately determining the type of obstruction area where the satellite is located and calculating obstruction porosity. This further improves the reliability of the obstruction porosity index, thereby ensuring the accuracy of the entire GNSS data quality prediction process.

[0080] Specifically, the elevation angle (ele) and azimuth angle (azi) of the satellite are first obtained, and the projection model of the fisheye camera is obtained through calibration. Finally, the elevation angle ele and azi angle azi are converted into pixel coordinates (x, y), and the calculation expression is as follows: ; ; ; Where (xc, yc) is the image center. K is the focal length, and k1, k2, k3, k4, and k5 are camera distortion parameters.

[0081] In some embodiments, the GNSS data quality prediction method based on fisheye cameras further includes: The total number of theoretically visible satellites and the total number of actually visible satellites are determined based on the occlusion type determined by satellite pixel coordinates and satellite feature data for future time periods. The data integrity rate is determined by the ratio of the actual number of visible satellites to the theoretical number of visible satellites.

[0082] Among them, data integrity rate: This indicator is calculated based on satellite trajectories predicted by ephemeris and the obstruction situation of the station to complete the satellite visibility judgment. It reflects the proportion of satellite observation data that the station to be evaluated can effectively receive in the future period. The higher the data integrity rate, the stronger the continuity and availability of the observation data.

[0083] Specifically, based on satellite pixel coordinates, if a satellite is located in a visible area of ​​the sky, the occlusion type is visible; if a satellite is located in a building occlusion area, the occlusion type is invisible; if a satellite is located in a tree occlusion area, the occlusion type is conditionally visible, then visibility marking is completed for all satellites in all epochs of future time periods, forming a "satellite-epoch" visibility matrix.

[0084] Since satellite feature data includes the elevation angle of each satellite at the corresponding time, the total number of theoretically visible satellites in all epochs of the future time period can be determined by combining satellite feature data of the future time period. That is, the total number of satellites with elevation angles higher than the minimum threshold when occlusion is not considered, and the total number of actually visible satellites, that is, the total number of satellites marked as "visible" or "conditionally visible" after the above occlusion determination.

[0085] The ratio of the actual number of visible satellites to the theoretical number of visible satellites is then used as the data integrity rate.

[0086] In some embodiments, the GNSS data quality prediction method based on fisheye cameras further includes: The design matrix is ​​determined based on the unit vector of each Category I satellite relative to the observation station to be evaluated. Category I satellites are those whose elevation angle is higher than the elevation angle threshold determined by the observation station to be evaluated and whose occlusion type is visible or conditionally visible. Determine the cofactor matrix based on the design matrix; The position dilution of precision (PDOP) is determined based on the elements in the cofactor matrix. PDOP is calculated based on the spatial geometric distribution of satellites after assessing satellite visibility using ephemeris predictions and station obstruction. It reflects the impact of satellite constellation configuration on positioning accuracy; a smaller PDOP value indicates a better satellite geometric distribution and higher accuracy and reliability of the positioning results.

[0087] In this context, the unit vector of the first type of satellite relative to the observation station to be evaluated is represented by (l,m,n), where l=ΔX / r, m=ΔY / r, n=ΔZ / r, r is the spatial distance from the satellite to the station, and ΔX, ΔY, and ΔZ are the differences between the geocentric coordinates of the first type of satellite and the geocentric coordinates of the observation station to be evaluated.

[0088] The components l, m, n of the unit line-of-sight vector are used as the first 3 columns of the design matrix A to construct the design matrix, and the cofactor matrix is ​​determined based on the relationship between the cofactor matrix and the design matrix.

[0089] Wherein, the cofactor matrix Q = (A T A) -1 A T It is the transpose of A, (A T A) -1 It is A T The inverse matrix of A is used to calculate the position accuracy factor using the following formula.

[0090] GDOP=Q 11 +Q 22 +Q 33 +Q44 Q 11 +Q 22 +Q 33 +Q 44 Q is the cofactor of the X, Y, Z coordinates of the observation station to be evaluated. 44 This is the receiver clock offset cofactor.

[0091] In one embodiment, such as Figure 3 As shown, the GNSS data quality prediction method based on fisheye cameras includes: Step 301: Obtain historical GNSS observation data and corresponding fisheye images; Step 302: Calculate satellite elevation angle, obstruction porosity, and data quality assessment indicator labels; Step 303: Train to obtain a general prediction model; Step 304: Obtain short-time GNSS observation data and corresponding fisheye images of the observation station to be evaluated; Step 305: Calculate the satellite elevation angle, azimuth angle, obstruction porosity, and data quality assessment index labels; Step 306: Train the observation station prediction model; Step 307: Based on the ephemeris forecast and fisheye image, call the general prediction model and the observatory prediction model respectively; Step 308: Output the data quality prediction results.

[0092] In this embodiment, a general random forest basic prediction model, i.e., the initial prediction model, is first constructed. This model uses satellite elevation angle and occlusion porosity accurately calculated from fisheye images as input features, and corresponding GNSS data quality indicators such as carrier-to-noise ratio (CNR) and pseudorange multipath as training labels for model training. Subsequently, at the location of the new observation station to be evaluated, short-term GNSS observation data and synchronous fisheye images are used, with satellite elevation angle, azimuth angle, and porosity as parameters, to fit functions to the CNR and pseudorange multipath, obtaining the observation station prediction model. This provides data quality prediction for the entire satellite and the entire day. Based on ephemeris predictions of satellite trajectories and combined with static occlusion information provided by the station's fisheye images, the elevation angle, azimuth angle, and porosity of future satellite sequences are calculated. Finally, the above features are simultaneously input into both the general prediction model and the observation station prediction model, and the output prediction values ​​of the two models are weighted and fused to achieve the prediction of conventional GNSS data quality indicators.

[0093] This method, through the fusion architecture of "general prediction model + observation station prediction model", enables the prediction of data quality for the whole day at a new observation station using only a small amount of short-term observation data, which significantly improves the efficiency of GNSS station site selection and quality assessment in complex obstruction environments such as urban areas and forest areas.

[0094] To verify the effectiveness of the method, this embodiment sets up a total of 6 observation stations, and the almanac data of the first five observation stations are used to build the basic model. Figure 4 This shows a schematic diagram illustrating the results of the image recognition algorithm, such as... Figure 4 As shown, it can better preserve the porosity of trees. Figure 5 and Figure 6 The fitting results for SNR1 and SNR2 of the G01 satellite are shown, such as... Figure 5 and Figure 6 As shown, the regression coefficients The values ​​are 0.9282 and 0.9440 respectively, with average errors of 1.5763 and 0.7406 respectively. The regression coefficients for most satellites are above 0.75, indicating that the basic model has excellent fitting ability for the carrier-to-noise ratio. Figure 7 and Figure 8 The fitting results for MP1 and MP2 of the C28 satellite are shown, such as... Figure 7 and Figure 8 As shown, the regression coefficients The coefficients are 0.6713 and 0.7344 respectively. The regression coefficients of most satellites are around 0.5. This is related to the strong randomness and environmental specificity of multipath error itself, but the model can still capture its basic trend. To improve prediction accuracy, this embodiment constructs an initial observation station prediction model based on 15 minutes of on-site observation data, divided into regions, as shown in Table 1. This model integrates a general prediction model with the initial observation station prediction model to predict data quality for the new observation station. The results are shown in Table 2. The prediction errors for SNR1 and SNR2 are stable at around 0.2, demonstrating extremely high accuracy. The prediction errors for MP1 and MP2 are approximately 20 cm, demonstrating practical application value in complex environments. Furthermore, satellite visibility is assessed based on ephemeris predictions and occlusion information provided by the fisheye camera. Data integrity rate and PDOP value are calculated. The prediction deviation for data integrity rate is 5.99%, and the prediction deviation for PDOP is 0.02, both results being relatively accurate.

[0095] Table 1

[0096] Table 2

[0097] It should be noted that the data quality prediction device provided by the present invention can execute the GNSS data quality prediction method based on fisheye camera of any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0098] like Figure 9 As shown, the GNSS data quality prediction device based on a fisheye camera provided by the present invention includes: The first determining module 901 is used to determine the satellite orbit information for a future period based on the ephemeris forecast, with the observation station to be evaluated as the reference point; The second determining module 902 is used to determine satellite characteristic data and occlusion porosity for future periods based on fisheye sky images taken at the observation station to be evaluated and satellite orbit information. The first evaluation module 903 is used to input the occlusion porosity of future time periods and at least part of the satellite characteristic data of future time periods into the general prediction model for data quality evaluation, and obtain the first data quality evaluation index output by the general prediction model. The second evaluation module 904 is used to input the occlusion porosity of the future time period and at least part of the satellite characteristic data of the future time period into the observation station prediction model corresponding to the observation station to be evaluated for data quality evaluation, and obtain the second data quality evaluation index output by the observation station prediction model. The prediction module 905 is used to determine the data quality prediction result based on the first data quality assessment index and the second data quality assessment index.

[0099] In this embodiment, by combining satellite orbit information from ephemeris predictions and occlusion information from fisheye sky images, and simultaneously introducing a general prediction model and an observation station-specific prediction model for dual-model evaluation, the general model's adaptability to various scenarios is preserved, while the observation station-specific model's accuracy for specific environments is also taken into account. Ultimately, efficient and accurate prediction of GNSS observation data quality is achieved, solving the problems of traditional methods relying on long-term observations and slow response. This approach can meet the needs of rapid site selection and real-time quality assessment in complex environments.

[0100] Figure 10 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 10As shown, the electronic device may include: a processor 1010, a communication interface 1020, a memory 1030, and a communication bus 1040, wherein the processor 1010, the communication interface 1020, and the memory 1030 communicate with each other through the communication bus 1040. The processor 1010 can call logic instructions in the memory 1030 to execute a GNSS data quality prediction method based on a fisheye camera. This method includes: determining satellite orbit information for a future period, using the observation station to be evaluated as a reference point, based on ephemeris forecasts; determining satellite feature data and occlusion porosity for the future period based on fisheye sky images taken at the observation station to be evaluated and the satellite orbit information; inputting the occlusion porosity and at least a portion of the satellite feature data for the future period into a general prediction model for data quality assessment, obtaining a first data quality assessment index output by the general prediction model; inputting the occlusion porosity and at least a portion of the satellite feature data for the future period into an observation station prediction model corresponding to the observation station to be evaluated for data quality assessment, obtaining a second data quality assessment index output by the observation station prediction model; and determining the data quality prediction result based on the first and second data quality assessment indices.

[0101] Furthermore, the logical instructions in the aforementioned memory 1030 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0102] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer is able to execute the GNSS data quality prediction method based on a fisheye camera provided in the above embodiments. The method includes: determining satellite orbit information for a future period based on ephemeris forecasts, with the observation station to be evaluated as a reference point; determining satellite feature data and occlusion porosity for a future period based on fisheye sky images taken at the observation station to be evaluated and the satellite orbit information; inputting the occlusion porosity and at least part of the satellite feature data for a future period into a general prediction model for data quality evaluation, and obtaining a first data quality evaluation index output by the general prediction model; inputting the occlusion porosity and at least part of the satellite feature data for a future period into an observation station prediction model corresponding to the observation station to be evaluated for data quality evaluation, and obtaining a second data quality evaluation index output by the observation station prediction model; and determining a data quality prediction result based on the first data quality evaluation index and the second data quality evaluation index.

[0103] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program is implemented to perform the GNSS data quality prediction method based on a fisheye camera provided in the above embodiments. The method includes: determining satellite orbit information for a future period based on ephemeris forecasts, with the observation station to be evaluated as a reference point; determining satellite feature data and occlusion porosity for the future period based on fisheye sky images taken at the observation station to be evaluated and the satellite orbit information; inputting the occlusion porosity and at least a portion of the satellite feature data for the future period into a general prediction model for data quality evaluation, and obtaining a first data quality evaluation index output by the general prediction model; inputting the occlusion porosity and at least a portion of the satellite feature data for the future period into an observation station prediction model corresponding to the observation station to be evaluated for data quality evaluation, and obtaining a second data quality evaluation index output by the observation station prediction model; and determining a data quality prediction result based on the first data quality evaluation index and the second data quality evaluation index.

[0104] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0105] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the quality of GNSS data based on a fisheye camera, characterized in that, include: Based on ephemeris forecasts, satellite orbit information for future periods is determined with the observation station to be evaluated as a reference point; Satellite characteristic data and occlusion porosity for future periods are determined based on fisheye sky images taken at the observation station to be evaluated and the satellite orbit information. The occlusion porosity of the future time period and at least part of the satellite feature data of the future time period are input into a general prediction model for data quality assessment, and a first data quality assessment index output by the general prediction model is obtained. The occlusion porosity of the future time period and at least part of the satellite feature data of the future time period are input into the observation station prediction model corresponding to the observation station to be evaluated for data quality assessment, and a second data quality assessment index output by the observation station prediction model is obtained. The data quality prediction result is determined based on the first data quality assessment index and the second data quality assessment index.

2. The GNSS data quality prediction method based on a fisheye camera according to claim 1, characterized in that, The step of determining the data quality prediction result based on the first data quality assessment index and the second data quality assessment index includes: Determine the first weight value corresponding to the first data quality assessment indicator and the second weight value corresponding to the second data quality assessment indicator; Determine the first product value of the first data quality assessment index and the first weight value; Determine the second product value of the second data quality assessment index and the second weight value; The sum of the first product value and the second product value is used as the data quality prediction result; wherein, both the first data quality assessment index and the second data quality assessment index include one or a combination of the following: carrier-to-noise ratio, pseudorange multipath; The satellite feature data includes one or a combination of the following: satellite elevation angle and azimuth angle.

3. The GNSS data quality prediction method based on a fisheye camera according to claim 1, characterized in that, The general prediction model is trained through the following steps: A first training set is determined, which includes a first combination of samples and data quality evaluation index labels corresponding to the first combination of samples. The first combination of samples includes a first satellite elevation angle and a first occlusion porosity. The first satellite elevation angle is a satellite elevation angle determined based on GNSS observation data under various occlusion environments. The first occlusion porosity is an occlusion porosity determined by image recognition of a first fisheye image. The first fisheye image is a fisheye image captured synchronously with GNSS observation data. The first combined sample is input into the initial prediction model to obtain the first prediction data quality evaluation index output by the initial prediction model. The first loss value is determined based on the data quality assessment index label corresponding to the first combined sample and the first predicted data quality assessment index. The trainable parameters in the initial prediction model are updated based on the first loss value; The observation station prediction model is trained through the following steps: A second training set is determined, which includes a second combined sample and data quality assessment index labels corresponding to the second combined sample. The second combined sample includes a second satellite elevation angle, a first azimuth angle, and a second occlusion porosity. The second satellite elevation angle is a satellite elevation angle determined based on short-time GNSS observation data. The first azimuth angle is an azimuth angle determined based on short-time GNSS observation data. The second occlusion porosity is an occlusion porosity determined by image recognition of a second fisheye image. The second fisheye image is a fisheye image that is synchronously captured with the short-time GNSS observation data at the observation station to be evaluated. The second combined sample is input into the initial observation station prediction model to obtain the second prediction data quality evaluation index output by the initial observation station prediction model. The second loss value is determined based on the data quality assessment index label corresponding to the second combined sample and the second predicted data quality assessment index. The trainable parameters in the initial observation station prediction model are updated based on the second loss value.

4. The GNSS data quality prediction method based on a fisheye camera according to any one of claims 1 to 3, characterized in that, The satellite orbit information includes the satellite position, and the future occlusion porosity is obtained through the following steps: The satellite position is projected onto the fisheye sky image to obtain the satellite pixel coordinates; Based on the satellite pixel coordinates, the satellite position is determined to be within the visible sky area, and the occlusion porosity is defined as 1. Based on the satellite pixel coordinates, the satellite position is determined to be in a building obstruction area, and the obstruction porosity is defined as -1; Based on the satellite pixel coordinates, the satellite position is determined to be in a tree-shaded area, and the shading porosity is defined as a first ratio; Wherein, the first ratio is the ratio of visible sky pixels to total pixels in the first region, and the first region is the region defined by the satellite pixel coordinates as the center and a radius of 100 pixels.

5. The GNSS data quality prediction method based on a fisheye camera according to claim 4, characterized in that, The satellite position includes the satellite elevation angle and azimuth angle. Projecting the satellite position onto the fisheye sky image yields the satellite pixel coordinates, including: Obtain the projection model of the fisheye camera; Based on the projection model, the satellite elevation angle and azimuth angle are converted into pixel coordinates to obtain the satellite pixel coordinates.

6. The GNSS data quality prediction method based on a fisheye camera according to claim 4, characterized in that, The GNSS data quality prediction method based on fisheye cameras also includes: The total number of theoretically visible satellites and the total number of actually visible satellites are determined based on the occlusion type determined by the satellite pixel coordinates and the satellite feature data for future time periods. The data integrity rate is determined based on the ratio of the actual total number of visible satellites to the theoretical total number of visible satellites.

7. The GNSS data quality prediction method based on a fisheye camera according to claim 4, characterized in that, The GNSS data quality prediction method based on fisheye cameras also includes: The design matrix is ​​determined based on the unit vector of each first-class satellite relative to the observation station to be evaluated. The first-class satellites are those whose elevation angle is higher than the elevation angle threshold determined by the observation station to be evaluated and whose occlusion type is visible or conditionally visible. Determine the cofactor matrix based on the design matrix; The position accuracy factor is determined based on the elements in the cofactor matrix.

8. A GNSS data quality prediction device based on a fisheye camera, characterized in that, include: The first determination module is used to determine the satellite orbit information for a future period based on ephemeris forecasts, with the observation station to be evaluated as the reference point. The second determining module is used to determine satellite characteristic data and occlusion porosity for future periods based on fisheye sky images taken at the observation station to be evaluated and the satellite orbit information. The first evaluation module is used to input the occlusion porosity of the future time period and at least part of the satellite feature data of the future time period into a general prediction model to evaluate the data quality and obtain the first data quality evaluation index output by the general prediction model. The second evaluation module is used to input the occlusion porosity of the future time period and at least part of the satellite feature data of the future time period into the observation station prediction model corresponding to the observation station to be evaluated for data quality evaluation, and obtain the second data quality evaluation index output by the observation station prediction model. The prediction module is used to determine the data quality prediction result based on the first data quality assessment index and the second data quality assessment index.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the GNSS data quality prediction method based on a fisheye camera as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the GNSS data quality prediction method based on a fisheye camera as described in any one of claims 1 to 7.

Images