A dynamic rendering engine implementation method and system based on multi-source data fusion

Abstract

The application relates to the technical field of intelligent navigation of ships, and provides a dynamic rendering engine implementation method and system based on multi-source data fusion, which comprises the following steps: collecting multi-source data in real time, preprocessing, and obtaining multi-modal data; time alignment and coordinate conversion of the multi-modal data are carried out through space-time registration, prediction is carried out through adaptive Kalman filtering, residual error is calculated according to the predicted data, and sensor anomaly detection is carried out according to the residual error; weighted fusion is carried out according to the sensor state, the fused data is subjected to Mercator projection transformation to obtain rendering coordinates, and the rendering coordinates are subjected to detail level marking; alarm judgment is carried out according to the detail level marking, and rendering sorting is carried out according to the alarm judgment; load detection and rendering parameter adjustment are carried out according to the rendering sorting through a frame rate control module. The application realizes the whole-process optimization from data collection and processing to visual display, improves the navigation precision and reliability, and improves the ship navigation safety.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent navigation of ships, and in particular to a dynamic rendering engine implementation method and system based on multi-source data fusion. BACKGROUND

[0002] The ship integrated navigation system has important application value due to its multi-source data fusion capability, high real-time information processing and strong safety characteristics. Accurate acquisition and efficient management of shipborne navigation information is a key prerequisite for improving ship efficiency, especially in complex electromagnetic environments and without external reference support. With the development of shipborne sensor technology, modern navigation systems can achieve high-precision autonomous navigation through multi-source information fusion and intelligent display and control technology. In existing shipborne navigation solutions, each type of sensor data is usually processed by independent subsystems, such as a satellite navigation system and an inertial navigation system, which are separately run and then simply superimposed and displayed. However, this decentralized architecture has problems such as poor data timeliness and difficulty in system collaboration, and the traditional coarse-grained design of permission management cannot meet the information security requirements in special environments. On the other hand, although commercial navigation devices have a high degree of interface standardization, their open communication protocols are at risk of being interfered with and stolen. Therefore, it is urgent to break through the construction of an autonomous and controllable shipborne navigation information management system in a strong confrontation environment. SUMMARY

[0003] The present application aims to at least solve one of the technical problems in the related art. To this end, the present application provides a dynamic rendering engine implementation method and system based on multi-source data fusion, which realizes the optimization of the whole process from data acquisition, processing to visualization display.

[0004] The present application provides a dynamic rendering engine implementation method based on multi-source data fusion, comprising:

[0005] S1: Real-time acquisition of multi-source data, and pre-processing of the multi-source data to obtain multi-modal data;

[0006] S2: Time alignment and coordinate conversion of the multi-modal data through space-time registration, and prediction of the converted multi-modal data through adaptive Kalman filtering to obtain current prediction data;

[0007] S3: Calculation of residuals according to the current prediction data and the converted multi-modal data, sensor anomaly detection according to the residuals, and obtaining of sensor states;

[0008] S4: Weighted fusion of sensor data according to the sensor states to obtain fusion data;

[0009] S5: The fused data is subjected to Mercator projection transformation to obtain rendering coordinates, and a detail level mark is made according to the rendering coordinates; an alarm judgment is made according to the detail level mark, and rendering sorting is performed according to the alarm judgment;

[0010] S6: Load detection and rendering parameter adjustment are performed according to the rendering sorting by the frame rate control module.

[0011] Further, the S1 step comprises:

[0012] S11: Physical constraint checking is performed on the multi-source data;

[0013] If the physical constraint checking is passed, statistical testing is performed;

[0014] If the physical constraint checking is not passed, an exception is marked;

[0015] S12: Statistical testing is performed by the 3 rule;

[0016] If the statistical testing is passed, multi-modal data is obtained;

[0017] If the statistical testing is not passed, an exception is marked;

[0018] S13: Data compensation and correction are performed on the data marked with the exception to obtain multi-modal data.

[0019] Further, in the S2 step, the state equation of the adaptive Kalman filter is:

[0020]

[0021] wherein, is a state vector at a time point, is a state transition matrix, is a state vector at a time point, is a control input matrix, is a control input vector at a time point, is a process noise at a time point, is an observation vector at a time point, is an observation matrix, is an observation noise at a time point;

[0022] Prediction step:

[0023]

[0024] wherein, is a state vector at a time point,​​​​​​ the prior state estimation at time k, is the posterior state estimation at time k, is the prior covariance matrix at time k, is the posterior covariance matrix at time k, is a process noise covariance matrix;

[0025] update step:

[0026]

[0027] wherein, is the Kalman gain at time k, is the posterior state estimation at time k, is the inverse of a matrix, is an observation noise covariance matrix, is a Huber loss function, is the transpose of a matrix.

[0028] Further, the S3 step comprises:

[0029] S31: calculating the statistical significance of the residual vector by chi-square test, judging the sensor state according to the statistical significance of the residual vector;

[0030] S32: if the sensor is normal, performing Kalman update;

[0031] if the sensor is abnormal, judging whether the sensor has been reset, if the sensor has been reset, re-collecting sensor data and executing the S1 step; if the sensor has not been reset, setting a sensor isolation duration and resetting the sensor within the sensor isolation duration.

[0032] Further, in the S31 step,

[0033] abnormality detection is performed by chi-square test,

[0034] if , the sensor is normal;

[0035] if , the sensor is abnormal;

[0036] wherein, is the residual vector at time k, is a chi-square distribution critical value, is the prior covariance matrix at time k, is an observation noise covariance matrix, This is the transpose of the matrix. It is the inverse of the matrix.

[0037] Furthermore, in step S4,

[0038] The expression for calculating sensor weights is:

[0039]

[0040] in, For the first The fusion weights of individual sensors, For the first Real-time variance of each sensor; Here, M represents the sensor weight attenuation coefficient, and M is the number of sensors participating in the fusion. For the first Real-time variance of each sensor;

[0041] No. The expression for calculating the real-time variance of a sensor is:

[0042]

[0043] in, For the first The observation vector of each sensor, For the first The observation matrix of each sensor, Estimate the current state of the system. The length of the sliding time window. For the Euclidean norm, This refers to the current moment.

[0044] Furthermore, in step S5,

[0045] S51: Perform Mercator projection transformation on the fused data to obtain rendering coordinates;

[0046] S52: Calculate the detail level switching conditions based on the rendering coordinates, select the rendering detail level and mark the detail level according to the detail level switching conditions;

[0047] S53: Decision-making is based on detail level markers, and rendering is sorted based on the decision-making of alarms;

[0048] If an alarm exists, calculate dynamic transparency, which is used for visual differentiation of alarms at the same level; define hierarchical priority, which is used for coverage of alarms across different levels.

[0049] If no warning is detected, the default rendering will be maintained.

[0050] Further, in S6,

[0051] S61: Calculate the rendering load of the current frame by the frame rate control module, and if the system is overloaded, reduce the detail level and adjust the rendering parameters;

[0052] If the system is not overloaded, keep the current rendering parameters;

[0053] S62: Repeat S61 to complete the load detection and rendering parameter adjustment of all frames in turn according to the rendering order.

[0054] The application also provides a dynamic rendering engine implementation system based on multi-source data fusion, which is used to execute the above-mentioned dynamic rendering engine implementation method based on multi-source data fusion, comprising:

[0055] A data acquisition unit, which acquires multi-source data in real time and pre-processes the multi-source data to obtain multi-modal data;

[0056] A prediction unit, which performs time alignment and coordinate conversion on the multi-modal data through space-time registration, and performs prediction on the converted multi-modal data through adaptive Kalman filtering to obtain current prediction data;

[0057] A detection unit, which calculates the residual error according to the current prediction data and the converted multi-modal data, and performs sensor anomaly detection according to the residual error to obtain the sensor state;

[0058] A fusion unit, which weights and fuses the sensor data according to the sensor state to obtain fused data;

[0059] A rendering unit, which performs Mercator projection transformation on the fused data to obtain rendering coordinates, and performs detail level marking according to the rendering coordinates; performs alarm judgment according to the detail level marking, and performs rendering sorting according to the alarm judgment;

[0060] An adjustment unit, which performs load detection and rendering parameter adjustment according to the rendering sorting through a frame rate control module.

[0061] The above one or more technical solutions in the embodiments of the application have at least one of the following technical effects:

[0062] In multi-source data fusion processing, the time and space reference difference problem of sensor data is solved by time and space registration technology, and Huber robust kernel function is introduced in Kalman filter algorithm for data fusion. Based on the real-time measurement variance of each sensor, the fusion weight is dynamically calculated to realize optimal estimation. Through chi-square test, the fault sensor is identified, and dynamic isolation strategy is implemented. The multi-source data after fusion is converted into a unified Mercator projection coordinate system, and the rendering precision level is automatically adjusted according to the target distance. High-precision model is used for short distance, medium-precision model is switched for medium distance, and simplified model is used for long distance. The periodic transparency adjustment algorithm is used for alarm information, and hierarchical rendering is implemented according to the three priority levels of collision warning, boundary warning and general target. Through adaptive frame rate control technology, the rendering cycle is dynamically adjusted according to the system load while ensuring the reference performance. The application supports dynamic rendering of the chart, route planning, equipment state monitoring and other core navigation functions, and can be directly integrated into the existing shipborne navigation system. The transformation cost is low, the compatibility is strong, and the application value for improving the safety and autonomy of ship navigation is important, especially suitable for high-precision and high-reliability scenes such as ocean transportation and special ships.

[0063] Additional aspects and advantages of the application will be set forth in part in the following description, will become apparent from the following description, or will be learned through practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0065] Fig. 1 It is a flowchart of the implementation method of the dynamic rendering engine based on multi-source data fusion provided by the application.

[0066] Fig. 2 It is a structural schematic diagram of the dynamic rendering engine implementation system based on multi-source data fusion provided by the application.

[0067] Reference signs:

[0068] 101, data acquisition unit; 102, prediction unit; 103, detection unit; 104, fusion unit; 105, rendering unit; 106, adjustment unit. DETAILED DESCRIPTION

[0069] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be clearly and completely described below. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the protection scope of the present application. The following embodiments are used to illustrate the present application, but cannot be used to limit the scope of the present application.

[0070] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example" or "some examples" means that the specific features, structures, materials or characteristics described in combination with the embodiment or example are contained in at least one embodiment or example of the embodiments of the present application. In the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any suitable manner in any one or more embodiments or examples. In addition, the different embodiments or examples described in the present application and the features of the different embodiments or examples can be combined and combined by those skilled in the art without contradiction.

[0071] The following will be described in combination with Figs. 1-2 The present application describes a dynamic rendering engine implementation method and system based on multi-source data fusion.

[0072] The navigation system is oriented to a shipborne navigation system composed of a display and control terminal, a data processing server and multiple sensors. In addition to containing basic control and communication modules, the display and control terminal needs to be equipped with an AngularJS framework to realize information visualization, the data processing server needs to be configured with a Python+Qt developed background service, and a Conn and History double-structure database is established. Among them, the high-precision navigation equipment (such as fiber-optic inertial navigation) is directly connected to the data processing server, and the ordinary navigation equipment is connected through the RS-422 or Ethernet interface. After the raw data collected by the navigation equipment is standardized and analyzed, the front-end and back-end data interaction is realized through the intermediate database, in which the inertial measurement unit provides attitude and motion information, the satellite navigation equipment provides absolute position reference, and the radar / AIS equipment provides surrounding situation information.

[0073] The navigation information processing method provided by the application is based on a distributed system architecture, first, through multi-thread concurrent collection, synchronous acquisition of 11 types of sensor data is realized, and data checking mechanism is used to filter abnormal data packets; second, a transaction management strategy of "emptying-inserting-overwriting" is adopted to ensure the atomicity of database operation, sensitive data protection is realized through three-level permission control; finally, comprehensive display of navigation situation is realized on the front-end display and control interface, functions such as dynamic rendering of the chart, route planning, equipment state monitoring and the like are supported, intelligent identification and processing of abnormal conditions are realized through four-level alarm classification, and finally a complete closed-loop system from data acquisition, processing to display is formed.

[0074] The system is initialized and a hardware self-check is performed, and the communication state of the equipment is detected in turn, specifically, the satellite navigation device: verifying the $GPGGA statement of NMEA-0183 protocol received;

[0075] The inertial navigation device: checking the FOG-148 state register (0xAA55 indicates normal);

[0076] The radar: sending PING to test network delay (threshold ≤ 50 ms).

[0077] Software initialization: configuration file loading, reading database connection parameters in SlaveConf.ini, setting sensor sampling rate.

[0078] Database initialization: creating Conn / History library table structure; establishing a database connection pool.

[0079] Real-time acquisition of multi-source data, and preprocessing of the multi-source data, obtaining multi-modal data;

[0080] Time synchronization enhancement, exponential weight dynamic compensation of clock deviation,

[0081] Time synchronization formula:

[0082]

[0083] Among them, is the system unified time reference, is the UTC time provided by the satellite navigation device, is the dynamic clock deviation compensation amount;

[0084] The calculation expression of the dynamic clock deviation compensation amount is:

[0085]

[0086] Among them, is the satellite navigation clock weight, is the deviation of the satellite navigation clock and the system master clock, is the inertial navigation clock weight, The deviation of the inertial navigation clock and the system master clock.

[0087] The calculation expression of the weight is:

[0088]

[0089] Wherein, is the weight of the first clock, is the clock weight attenuation coefficient, is the weight of the first clock source, is the variance of the first clock source, and is an index of sensor stability.

[0090] In some specific embodiments of the present application, .

[0091] As shown in Fig. 1 , a dynamic rendering engine implementation method based on multi-source data fusion comprises:

[0092] S1: Real-time acquisition of multi-source data, and pre-processing of the multi-source data to obtain multi-modal data;

[0093] The multi-source data comprises satellite navigation data, inertial navigation data, radar data and AIS data.

[0094] S11: Physical constraint checking is performed on the multi-source data;

[0095] If the physical constraint checking is passed, statistical testing is performed;

[0096] If the physical constraint checking is not passed, an exception is marked;

[0097] The calculation expression of the physical constraint checking is:

[0098]

[0099] Wherein, is the physical constraint, is the AIS provided speed data, is the attitude change rate, is the height.

[0100] S12: Statistical testing is performed by the 3 rule;

[0101] If the statistical testing is passed, the multi-modal data is obtained;

[0102] If the statistical testing is not passed, an exception is marked;

[0103] The calculation expression is:

[0104] ​

[0105] in, For statistical verification, This is the average value of the active window. For multi-source data, , For the first Data from one data source, Number of data sources;

[0106] In some specific embodiments of the present invention .

[0107] S13: Perform data compensation and correction on the marked outlier data to obtain multimodal data;

[0108] The expression for calculating the standard deviation at time t is:

[0109]

[0110] in, for Standard deviation of time These are weighting coefficients. for Standard deviation of time for Time data.

[0111] A single validation failure is considered a transient anomaly, compensated for by linear interpolation: the calculation expression is: in, For the supplementary data, for Data at any given time for Data at any given time;

[0112] Three consecutive failures constitute a persistent error; switch data sources.

[0113] An abnormal device status indicates a hardware failure, triggering the first alarm and isolating the device.

[0114] S2: Perform time alignment and coordinate transformation on multimodal data through spatiotemporal registration, and predict the transformed multimodal data through adaptive Kalman filtering to obtain the current prediction data;

[0115] Spatiotemporal registration includes time alignment and coordinate transformation.

[0116] The expression for calculating time alignment is:

[0117]

[0118] in, For the aligned time, a local timestamp for the sensor, a fixed delay for the sensor, a current time, an initial capture time for the sensor data packet, a delay time for data from generation to processing, a first dynamic compensation coefficient, a second dynamic compensation coefficient.

[0119] In some embodiments of the present application, , .

[0120] Converting the sensor coordinate system to the world coordinate system, the calculation expression is:

[0121]

[0122] wherein, is the converted world coordinate system, is a fixed displacement vector between the sensor installation position and the origin of the carrier coordinate system, is a yaw angle, is a pitch angle, is a first intermediate matrix, is a second intermediate matrix, is a sensor coordinate system,

[0123] The calculation expression of is:

[0124]

[0125] The calculation expression of is:

[0126]

[0127] The state equation of the adaptive Kalman filter is:

[0128]

[0129] wherein, is a state vector at time, is a state transition matrix, is a state vector at time, is a control input matrix, is a control input vector at time, is a process noise at time, is a measurement noise at time, is a measurement vector at time, is a measurement matrix, is a measurement noise at time, , for a process noise covariance matrix, for an observation vector at time for an observation matrix, for an observation noise at time , for an observation noise covariance matrix, comprising a variable at time , for a three-dimensional coordinate, for a three-dimensional velocity, for a roll angle, for a pitch angle, for a yaw angle;

[0130] a prediction step:

[0131]

[0132] wherein, for a prior state estimate at time for a posterior state estimate at time for a prior covariance matrix at time for a posterior covariance matrix at time for a transpose of a matrix, for a process noise covariance matrix;

[0133] an update step:

[0134]

[0135] wherein, for a Kalman gain at time for a posterior state estimate at time for an observation noise covariance matrix, for a Huber loss function, for an inverse of a matrix;

[0136] a computational expression of the Huber loss function is:

[0137]

[0138] wherein, for a robust kernel threshold value, for an input variable of a function.

[0139] In some specific embodiments of the present invention .

[0140] S3: Calculate the residuals based on the current predicted data and the converted multimodal data, perform sensor anomaly detection based on the residuals, and obtain the sensor status;

[0141] S31: Calculate the statistical significance of the residual vector through the chi-square test, and judge the sensor status based on the statistical significance of the residual vector;

[0142] Anomaly detection is performed using the chi-square test.

[0143] like If so, the sensor is working properly;

[0144] like If so, the sensor is malfunctioning;

[0145] in, for The residual vector at time step, This is the critical value of the chi-square distribution. for The prior covariance matrix at time t, For the observation matrix, To observe the noise covariance matrix;

[0146] S32: If the sensor is working properly, perform a Kalman update;

[0147] If the sensor malfunctions, determine whether the sensor has been reset. If the sensor has been reset, re-acquire sensor data and execute step S1. If the sensor has not been reset, set the sensor isolation time and reset the sensor within the sensor isolation time.

[0148] The expression for calculating the sensor isolation time is:

[0149]

[0150] in, For sensor isolation duration, To find the minimum value function, For the first Real-time variance of individual sensors;

[0151] When a sensor is abnormally marked as faulty due to residual differences, setting an isolation period can temporarily exclude it from data fusion, such as reducing its weight or isolating it directly, to prevent continuous interference from abnormal observations and ensure the accuracy of system state estimation.

[0152] S4: Weight and fuse sensor data according to sensor status to obtain fused data;

[0153] The calculation expression of the sensor weight is:

[0154]

[0155] wherein, is the fusion weight of the i th sensor, is the real-time variance of the i th sensor; is the sensor weight decay coefficient, and M is the number of sensors participating in fusion, is the real-time variance of the i th sensor; The calculation expression of the real-time variance of the i th sensor is:

[0156] The calculation expression of the real-time variance of the i th sensor is:

[0157] wherein,

[0158] is the observation vector of the i th sensor, is the observation matrix of the i th sensor, is the current state estimation of the system, is the length of the sliding time window; is the Euclidean norm. In some specific embodiments of the present application, .

[0159] S5: performing Mercator projection transformation on the fused data to obtain rendering coordinates of the sea, marking a detail level according to the chart coordinates and system performance parameters; performing alarm judgment according to the detail level marking, and performing rendering sorting according to the alarm judgment;

[0160] S51: performing Mercator projection transformation on the fused data to obtain rendering coordinates;

[0161] The calculation expression of the Mercator projection transformation is:

[0162] The calculation expression of the Mercator projection transformation is:

[0163]

[0164] wherein, is the horizontal coordinate of the target point, is the vertical coordinate of the target point; is the target longitude of the target point, is the target latitude of the target point, is the projection center longitude; is the earth radius;

[0165] ​​​​S52: Calculate the detail level switching condition according to the rendering coordinates, select the rendering detail level according to the detail level switching condition, and mark the detail level;

[0166] Adjust the rendering details dynamically according to the distance between the target and the observer, optimize the performance under the premise of ensuring the visual effect. In the near distance, high-precision models are used, such as more triangles and high-definition textures; in the far distance, low-precision models are used, such as simplified geometry and compressed textures; the purpose is to balance the rendering quality and computational load;

[0167] Detail level switching condition:

[0168]

[0169] Wherein, is the level of detail level; is the projection coordinate of the target in the view coordinate system; is the distance threshold; is the high-precision mode; is the medium-precision mode; is the low-precision mode;

[0170] Performance optimization index:

[0171]

[0172] Wherein, is the rendering cost, which is a quantitative performance evaluation index, used to measure the computational resource consumption required for rendering a specific object, is the number of model triangles, used to measure the geometric complexity of the model, is the number of texture pixels, used to measure the texture memory and sampling pressure, is the geometric weight coefficient, used to measure the influence weight of the number of triangles, the typical value is 0.7; is the texture weight coefficient, used to measure the influence weight of the texture resolution, the typical value is 0.3;

[0173] S53: Alarm judgment according to the detail level mark, and rendering sorting according to the alarm judgment;

[0174] If there is an alarm, calculate the dynamic transparency, which is used for visual differentiation of alarms in the same level; define the level priority, which is used for coverage of alarms across levels;

[0175] The calculation expression of dynamic transparency is:

[0176]

[0177] Wherein, is the alarm state transparency; is the flicker frequency, and a typical value is 1 Hz.

[0178] The calculation expression of the hierarchical priority is:

[0179]

[0180] wherein, is the hierarchical priority, and the system renders in the order from = 3 to 1, that is, the regular target → the boundary alert → the collision alert, even if the dynamic transparency is the same, the high-level content will cover the low-level content;

[0181] If there is no alert, the default rendering is kept.

[0182] S6: performing load detection and rendering parameter adjustment according to the rendering order by the frame rate control module;

[0183] S61: calculating the rendering load of the current frame by the frame rate control module, and if the system is overloaded, reducing the detail level and adjusting the rendering parameter.

[0184] The calculation expression of the target frame period is:

[0185]

[0186] wherein, is the target frame period, is the reference period, is the actual time consumption of the last frame, is the load buffer coefficient, is the maximum value function;

[0187] The calculation expression of the system load rate is:

[0188]

[0189] wherein, is the system load rate, and is used for measuring the current frame resource occupancy rate; is the rendering cost of the th model, which comprehensively considers the geometry and texture complexity; is the number of objects rendered in the current frame; is the target frame period.

[0190] If the system is not overloaded, the current rendering parameter is kept.

[0191] If the system is overloaded, the detail level is reduced and the rendering parameter is adjusted, and the adjustment of the rendering parameter includes adjusting the target frame period.

[0192] In some specific embodiments of the present application, , .

[0193] S62: Repeat step S61 to perform load detection and rendering parameter adjustment for all frames in sequence according to the rendering order.

[0194] The system displays equipment status, navigation status, and alarm information through a visual interface, supports interactive operations such as chart zooming and route planning, and maintenance personnel can adjust system parameters through dedicated configuration tools, such as modifying sensor weight coefficients and adjusting alarm thresholds. The system log records the complete operation trajectory and supports post-event audit analysis.

[0195] like Fig. 2 As shown, a dynamic rendering engine implementation system based on multi-source data fusion is used to execute a dynamic rendering engine implementation method based on multi-source data fusion, including:

[0196] The data acquisition unit 101 collects multi-source data in real time and preprocesses the multi-source data to obtain multimodal data;

[0197] Prediction unit 102 performs time alignment and coordinate transformation on multimodal data through spatiotemporal registration, and predicts the transformed multimodal data through adaptive Kalman filtering to obtain the current prediction data;

[0198] The detection unit 103 calculates the residual based on the current predicted data and the converted multimodal data, performs sensor anomaly detection based on the residual, and obtains the sensor status.

[0199] The fusion unit 104 performs weighted fusion of sensor data according to the sensor status to obtain fused data;

[0200] The rendering unit 105 performs Mercator projection transformation on the fused data to obtain chart coordinates, and performs detail level marking based on the chart coordinates and system performance parameters; it then performs alarm judgment based on the detail level marking and performs rendering sorting based on the alarm judgment.

[0201] The adjustment unit 106 performs load detection and adjusts rendering parameters according to the rendering order through the frame rate control module.

[0202] Through the cooperative work of the above-mentioned modules, the problem of time and space reference difference of sensor data is solved in multi-source data fusion processing through space-time registration technology, and Huber robust kernel function is introduced in Kalman filter algorithm for data fusion. Based on the real-time measurement variance of each sensor, the fusion weight is dynamically calculated to realize optimal estimation. Through chi-square test, the fault sensor is identified, and dynamic isolation strategy is implemented. The multi-source data after fusion is converted into a unified Mercator projection coordinate system, the rendering precision level is automatically adjusted according to the target distance, the high-precision model is used for short-distance, the medium-precision model is switched for medium-distance, and the simplified model is used for long-distance. The periodic transparency adjustment algorithm is used for alarm information, and hierarchical rendering is implemented according to the three priority levels of collision warning, boundary warning and general target. Through adaptive frame rate control technology, the rendering cycle is dynamically adjusted according to the system load while ensuring the reference performance. The application supports dynamic rendering of the chart, route planning, equipment state monitoring and other core navigation functions, and can be directly integrated into the existing shipborne navigation system, with low modification cost and strong compatibility, and has important application value for improving the safety and autonomy of ship navigation, especially suitable for high-precision and high-reliability navigation scenes such as ocean transportation and special ships.

[0203] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for implementing a dynamic rendering engine based on multi-source data fusion, characterized in that, Comprise: S1: Real-time acquisition of multi-source data, and preprocessing of multi-source data to obtain multi-modal data; S2: Time alignment and coordinate conversion of multi-modal data through space-time registration, and prediction of converted multi-modal data through adaptive Kalman filtering to obtain current prediction data; The state equation of adaptive Kalman filtering is: wherein is the state vector at time instant is the state transition matrix, is the state vector at time instant is the control input matrix, is the control input vector at time instant is the process noise at time instant is the observation vector at time instant is the observation matrix, is the observation noise at time instant Prediction step: wherein is a prior state estimate at time k, is a posterior state estimate at time k, is a prior covariance matrix at time k, is a posterior covariance matrix at time k, is a process noise covariance matrix; Update step: wherein is the Kalman gain at time k, is the a posteriori state estimate at time k, is the inverse of a matrix, is the observation noise covariance matrix, is the Huber loss function, is the transpose of a matrix; The calculation expression of Huber loss function is: wherein, is a robust kernel threshold, is an input variable of the function; S3: Calculate the residual according to the current prediction data and the converted multi-modal data, and perform sensor anomaly detection according to the residual to obtain the sensor state; S4: Weighted fusion of sensor data according to sensor state to obtain fused data; S5: Mercator projection transformation of fused data to obtain rendering coordinates, and detail level marking according to rendering coordinates; Alarm judgment according to detail level marking, and rendering sorting according to alarm judgment; S6: Load detection and rendering parameter adjustment according to rendering sorting through frame rate control module. 2.The method of claim 1, wherein, The multi-source data includes satellite navigation data, inertial navigation data, radar data and AIS data. 3.The method of claim 1, wherein, S1 step includes: S11: Physical constraint check on multi-source data; If the physical constraint check is passed, statistical test is performed; If the physical constraint check is not passed, mark the anomaly; S12: By 3 statistical tests by rule; If the statistical test is passed, multi-modal data is obtained; If the statistical test is not passed, mark the anomaly; S13: Data compensation and correction of marked abnormal data to obtain multi-modal data.

4. The method of claim 1, wherein, S3 step includes: S31: Calculate the statistical significance of residual vector through chi-square test, and judge the sensor state according to the statistical significance of residual vector; S32: If the sensor is normal, Kalman update is performed; If the sensor is abnormal, it is judged whether the sensor has been reset, if the sensor has been reset, the sensor data is reacquired, and S1 step is executed; If the sensor has not been reset, set the sensor isolation time, and reset the sensor within the sensor isolation time.

5. The method of claim 4, wherein, In S31 step, Abnormal detection is performed through chi-square test, If then the sensor is normal; If then the sensor is abnormal; wherein is the residual vector at time instant is a chi-square distribution threshold, is is the prior covariance matrix at time instant is the observation matrix, is the observation noise covariance matrix, is the transpose of a matrix, is the inverse of a matrix.

6. The method of claim 1, wherein, In S4 step, The calculation expression of sensor weight is: wherein, is the fusion weight of the first sensor, is the real-time variance of the first sensor, is the fusion weight of the first sensor, is the real-time variance of the first sensor; is the sensor weight decay coefficient, M is the number of sensors participating in fusion, is the real-time variance of the first sensor, is the real-time variance of the first sensor. The calculation expression of real-time variance of the first sensor is: ​ wherein, is the observation vector of the th sensor, is the observation matrix of the th sensor, is the current state estimate of the system, is the length of the sliding time window, is the Euclidean norm, is the current time instant.

7. The method of claim 1, wherein, In S5 step, S51: Mercator projection transformation of fused data to obtain rendering coordinates; S52: Calculate detail level switching condition according to rendering coordinates, select rendering detail level and perform detail level marking according to detail level switching condition; S53: Alarm judgment according to detail level marking, and rendering sorting according to alarm judgment; If there is an alarm, calculate the dynamic transparency, which is used for visual differentiation of alarms in the same level; Define level priority, which is used for coverage of cross-level alarms; If there is no alarm, keep the default rendering. 8.The method of claim 1, wherein, In S6 step, S61: Calculate the rendering load of the current frame through the frame rate control module, if the system is overloaded, reduce the detail level and adjust the rendering parameter; If the system is not overloaded, keep the current rendering parameter; S62: Repeat S61 step to complete load detection and rendering parameter adjustment of all frames in turn according to rendering sorting.

9. A multi-source data fusion based dynamic rendering engine implementation system, characterized in that, To execute a multi-source data fusion based dynamic rendering engine implementation method as claimed in any one of claims 1 to 8, comprising: The data acquisition unit collects multi-source data in real time, and pre-processes the multi-source data to obtain multi-modal data; The prediction unit performs time alignment and coordinate conversion on the multi-modal data through space-time registration, and performs prediction on the converted multi-modal data through adaptive Kalman filtering to obtain current prediction data; The detection unit calculates a residual according to the current prediction data and the converted multi-modal data, performs sensor anomaly detection according to the residual, and obtains a sensor state; The fusion unit weights and fuses sensor data according to the sensor state to obtain fused data; The rendering unit performs Mercator projection transformation on the fused data to obtain rendering coordinates, marks a detail level according to the rendering coordinates, judges an alarm according to the detail level, and performs rendering sorting according to the alarm judgment; The adjustment unit performs load detection and rendering parameter adjustment according to the rendering sorting through a frame rate control module.

Images