Midway rapid calibration method for inertial positioning of unmanned aerial vehicle
By using similarity calculation and pre-stored positioning points calibration methods in drones, the problems of feature matching failure of inertial positioning technology and instability of external positioning signals in geological surveys in unmanned areas are solved, and efficient and accurate positioning calibration is achieved , suitable for harsh environments and multiple flight needs.
Patent Information
- Application Number
- CN202411341634.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-25
- Publication Date
- 2024-11-22
AI Technical Summary
In geological survey of unmanned areas, inertial positioning technology has failed to match feature points and difficult positioning calibration due to complex environments and unstable external positioning signals. The existing technology is inefficient and difficult to adapt to the harsh environment and the needs of multiple flights.
Using the drone's inertial positioning midway fast calibration method based on similarity calculation, through similarity judgment and pre-stored positioning point calibration, avoid feature matching failures and external positioning signal communication failures, and update positioning points and errors using similarity attenuation trends and calibration deviations. value to improve the scope of application and accuracy of positioning calibration.
It effectively avoids feature matching failures, improves the scope and accuracy of positioning calibration, improves the operating efficiency in harsh environments, and is suitable for geological surveys in unmanned areas, especially for efficient positioning calibration of fixed-wing drones.
Smart Images

Figure CN118999625A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a mid-course rapid calibration method for inertial positioning of an unmanned aerial vehicle, and belongs to the technical field of inertial navigation. Background Art
[0002] When conducting geological surveys in uninhabited areas, using drones is a preferred method. However, due to the complex environmental conditions in uninhabited areas, external positioning signals are difficult to apply in real time, so the positioning of drones often needs to be based on inertial positioning. On the other hand, geological surveys often require drones to fly multiple times to achieve repeated surveys from different angles and using different equipment, and inertial positioning technology will have very large cumulative errors during long-term flights, so positioning calibration is very important.
[0003] In the prior art, the calibration of UAV inertial positioning is generally based on a visual solution to match feature points for identification and then estimate the position. For example, the Chinese invention patent with application number CN202110490298.9 discloses a UAV position estimation system and estimation method based on a visual sensor. The digital image is processed by detecting the feature points of the current frame image, etc. to obtain the displacement information of the UAV, which is then fused with the measurement results of the GPS module to finally achieve a higher-precision position estimation.
[0004] This type of method has two key technical points when calibrating the drone's position information. One is that it requires image feature point matching, which is difficult to apply when feature points are difficult to match. The other is that it requires satellite positioning to achieve calibration. Without an external positioning signal, it is difficult to calibrate the inertial sensor during flight.
[0005] Obviously, when applying the existing technology to geological surveys in uninhabited areas, firstly, the environmental conditions are complex and the planned routes often have obvious deviations from the actual flight routes, which leads to poor feature point matching. Therefore, the existing technology public solutions are often stuck in the feature point matching stage, resulting in the inability to perform subsequent positioning. Secondly, the external positioning signal is relatively unstable. In practice, even if the feature point matching is completed, it is often difficult to obtain an effective positioning signal, resulting in positioning calibration failure.
[0006] In addition, since multiple flights are required for multi-dimensional repeated surveys, the more preferred solution is to use fixed-wing drones to increase the speed of a single flight. However, the use of existing technology often means that the drone needs to hover in the air to complete the transmission and reception and calibration of external positioning signals. This inefficient positioning and calibration method is unacceptable for fixed-wing drones. Summary of the invention
[0007] In order to solve the above technical problems, the present invention provides a method for rapid mid-course calibration of inertial positioning of unmanned aerial vehicles. The method for rapid mid-course calibration of inertial positioning of unmanned aerial vehicles can effectively avoid the possibility of feature matching failure based on similarity calculation. At the same time, the calibration method based on pre-stored positioning points can effectively avoid the situation where positioning calibration failure is caused by failure of external positioning signal communication.
[0008] The present invention is achieved through the following technical solutions.
[0009] The invention provides a method for rapid calibration of inertial positioning of an unmanned aerial vehicle, comprising the following steps:
[0010] ① Initial setting: Initially set the error value according to the parameters of the inertial sensor used, and set the calibration waypoint list according to the calibration point set and the itinerary route, with the first calibration waypoint in the waypoint list as the current calibration waypoint;
[0011] ②Enter range judgment: According to the current positioning point, judge whether it has entered the range of the current calibration waypoint plus N times the error value. If it has entered, proceed to the next step;
[0012] ③ Comparison: Get the current aerial image, compare the current aerial image with the aerial image corresponding to the current calibration waypoint and calculate the similarity to get the current similarity;
[0013] ④Similarity decay judgment: judge the size of the current similarity and the previous similarity. If the current similarity is greater than the previous similarity, delete the previous similarity, cache the current similarity as the previous similarity and return the value step ②. If the current similarity is less than the previous similarity, proceed to the next step.
[0014] ⑤Replacement record: Calculate the difference between the current calibration waypoint and the current positioning point, calculate the calibration deviation based on the difference using the previous similarity as the coefficient, and update the positioning point and error value based on the calibration deviation;
[0015] ⑥ Route end judgment: judge whether the current calibration waypoint is the last one in the calibration waypoint list. If not, set the next calibration waypoint as the current calibration waypoint and return to step ②. If yes, exit.
[0016] Therefore, the method based on similarity judgment no longer matches based on image feature points, which can effectively avoid the failure of feature point matching; the method of finding calibration points based on the similarity increase and decrease trend judgment and performing positioning calibration based on the calibration deviation can achieve a larger positioning area, and has a greater tolerance for situations where it is difficult to fly to a better positioning point due to complex environments and unstable routes. It greatly improves the scope of application of positioning calibration and can better adapt to harsh air environments.
[0017] In step ③, the current aerial image and the aerial image corresponding to the current calibration waypoint are first pooled and then compared to calculate the similarity.
[0018] In the pooling operation, the filter size is M×M, the step size is M, and the value range of M is 0.5% to 2% of the number of pixels on the short side of the current aerial image and is rounded up.
[0019] In the step ②, the value of N is 0.7-3.
[0020] In the step ①, the error value is set to 3 to 6 times the parameter error of the inertial sensor used.
[0021] In the step ⑤, the calibration deviation is the multiplication of the previous similarity and the difference, and the previous similarity is first normalized.
[0022] In step ③, the similarity is calculated using a difference value hash algorithm.
[0023] In the step ⑤, the positioning point is updated according to the calibration deviation, and the positioning point obtained by adding the difference to the current calibration waypoint and subtracting the calibration deviation is used as the current positioning point.
[0024] In step ⑤, the error value is updated according to the calibration deviation by multiplying the ratio of the calibration deviation and the difference value by the error value as a new error value, and the error value is replaced by the new error value.
[0025] In the step ①, the calibration waypoint list is set according to the calibration point set and the itinerary route, the itinerary route is expanded by a preset range H to obtain a itinerary route range, the calibration points within the itinerary route range are extracted, and the extracted calibration points are sorted according to the itinerary route direction to obtain a calibration waypoint list.
[0026] The beneficial effects of the present invention are as follows: the method based on similarity calculation can effectively avoid the possibility of feature matching failure, and the method based on pre-stored positioning points for calibration can effectively avoid the situation where positioning calibration failure is caused by failure of external positioning signal communication; the overall method is easy to implement, has high operating efficiency, and acceptable accuracy, and is particularly suitable for the use of drones for geological surveys in uninhabited areas. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 It is a schematic diagram of a process of at least one embodiment of the present invention. DETAILED DESCRIPTION
[0028] The technical solution of the present invention is further described below, but the scope of protection claimed is not limited to the description.
[0029] The first embodiment of the present invention relates to a method for rapid calibration of UAV inertial positioning mid-course, such as Figure 1 As shown, the following steps are included:
[0030] ① Initial setting: Initially set the error value according to the parameters of the inertial sensor used, and set the calibration waypoint list according to the calibration point set and the itinerary route, with the first calibration waypoint in the waypoint list as the current calibration waypoint;
[0031] ②Enter range judgment: According to the current positioning point, judge whether it has entered the range of the current calibration waypoint plus N times the error value. If it has entered, proceed to the next step;
[0032] ③ Comparison: Get the current aerial image, compare the current aerial image with the aerial image corresponding to the current calibration waypoint and calculate the similarity to get the current similarity;
[0033] ④Similarity decay judgment: judge the size of the current similarity and the previous similarity. If the current similarity is greater than the previous similarity, delete the previous similarity, cache the current similarity as the previous similarity and return the value step ②. If the current similarity is less than the previous similarity, proceed to the next step.
[0034] ⑤Replacement record: Calculate the difference between the current calibration waypoint and the current positioning point, calculate the calibration deviation based on the difference using the previous similarity as the coefficient, and update the positioning point and error value based on the calibration deviation;
[0035] ⑥ Route end judgment: judge whether the current calibration waypoint is the last one in the calibration waypoint list. If not, set the next calibration waypoint as the current calibration waypoint and return to step ②. If yes, exit.
[0036] Generally, when a drone is affected by the environment and deviates significantly, the drone itself cannot reach the preset positioning point it should have reached. At this time, since the drone cannot obtain images at the preset positioning point, the feature point matching method is not applicable. However, the method based on similarity calculation, especially the method based on similarity attenuation to determine whether to perform positioning calibration, is sufficient to cope with general degrees of yaw based on simple geometric principles, and the amount of calculation is extremely small. There is no need to consider attitude correction, image projection and other methods that have complex calculation processes and are prone to errors, and the stability and accuracy of the results are guaranteed.
[0037] Generally, the application of the above scheme should have an aircraft to collect relatively accurate positioning information, especially an aircraft with high precision, high stability and relatively complete equipment, to ensure the accuracy of the calibration waypoints. In practice, remote-controlled aircraft or helicopters are generally used. Since multiple flights are required for multi-dimensional repeated surveys in practice, for cost considerations, after the first flight obtains positioning information, it is worth considering using lower-cost drones for subsequent various surveys. Therefore, the above scheme is mainly suitable for lower-cost drones used after the first flight obtains positioning information.
[0038] The second embodiment of the present invention is basically the same as the first embodiment, mainly in the preferred scheme of the initial setting step. In step ①, the error value is set to 3 to 6 times the error of the inertial sensor parameters used. Since the parameters of low-end and mid-end inertial sensors in the existing market are often falsely marked, the purpose of adopting this method is to provide greater initial fault tolerance to ensure the accuracy of positioning calibration to a certain extent.
[0039] Furthermore, in step ①, the calibration waypoint list is set according to the calibration point set and the itinerary route, the itinerary route is expanded by a preset range H to obtain the itinerary route range, the calibration points concentrated within the itinerary route range are taken out, and the taken out calibration points are sorted according to the itinerary route direction to obtain the calibration waypoint list. Generally, H can be set to 2 to 10m. The main purpose is to use as many calibration waypoints as possible to achieve calibration and ensure that the cumulative error of the inertial sensor is kept at a low level.
[0040] The third embodiment of the present invention is basically the same as the first embodiment, mainly in the preferred scheme of the range judgment step. In step ②, N takes a value of 0.7 to 3. Obviously, the value of N directly determines the spatial range of the pre-image similarity calculation before starting the positioning calibration. If the value of N is too large, the amount of calculation is too large, which may easily cause the program to run away. If the value of N is too small, it may cause the positioning calibration to fail to start when the drone yaws seriously. Therefore, based on practice, N should be selected within the range of 0.7 to 3 according to the environmental conditions. If the value of N needs to be set to more than 3, it means that the climate and environmental conditions are very bad, and the drone should not take off. If it takes off forcibly, the drone will have a greater risk of damage.
[0041] The fourth embodiment of the present invention is basically the same as the first embodiment, mainly in the preferred scheme of the comparison step. In step ③, the current aerial image and the aerial image corresponding to the current calibration waypoint are first pooled and then compared to calculate the similarity. The role of pooling is to reduce the amount of calculation while ensuring the accuracy of image similarity calculation to a certain extent.
[0042] Furthermore, in the pooling operation, the filter size is M×M, the step size is M, and the value range of M is 0.5% to 2% of the number of pixels on the short side of the current aerial image and rounded up. Although the filter in this solution is larger than the pooling filter in the prior art, since the subsequent calculation is to calculate the similarity, the role of pooling is to reduce the amount of calculation. Therefore, if the filter is too small, it is easy to cause the accuracy of the similarity calculation to be too low. On the contrary, if the filter is too large, it is easy to cause the amount of similarity calculation to be substantially reduced, but the amount of calculation of the pooling operation itself is greatly increased, which has a significant adverse effect on performance and does not meet the purpose of this solution.
[0043] Furthermore, in step ③, the difference hash algorithm is used to calculate the similarity. A major feature of the difference hash algorithm is that it has high accuracy and is very fast, which is particularly suitable for this solution. It is easy to understand that the smaller the calculation result value of the difference hash algorithm, the higher the similarity, and the value is 0-64. Therefore, for the calculation result Hash value used, the difference hash algorithm calculation value dHash should be inverted, that is, Hash = 64-dHash.
[0044] The fifth embodiment of the present invention is basically the same as the first embodiment, mainly in that the preferred scheme of the replacement recording step is used. In step ⑤, the calibration deviation is the multiplication value of the previous similarity and the difference, and the previous similarity is first normalized. The operation of normalizing the previous similarity is to make the similarity reflect the probability of positioning matching to a certain extent, and its range is consistent with the value range of the probability, both of which are 0 to 1, which greatly reduces the calculation amount of subsequent positioning calibration calculations.
[0045] Preferably, in step ⑤, the positioning point is updated according to the calibration deviation, and the positioning point obtained by adding the difference to the current calibration waypoint and subtracting the calibration deviation is used as the current positioning point. In general, the higher the similarity, the closer the drone is to the calibration waypoint, and therefore the smaller the amount of error correction. Although there is an obvious deviation in theory when using this method, the deviation is within an acceptable range for geological surveys in uninhabited areas. The main method of inertial positioning is to continuously read the values of the inertial sensor and convert them into coordinate differences and then superimpose them on the current positioning point. The current positioning point is essentially a superposition result value, so the above-mentioned replacement operation of the current positioning point is essentially positioning calibration.
[0046] Preferably, in step ⑤, the error value is updated according to the calibration deviation, and the error value is replaced by the new error value by multiplying the ratio of the calibration deviation and the difference value by the error value. Thus, even if the error value is set to be relatively large in the preceding step, after multiple calibrations (i.e., replacing the error value with the new error value), the error value used will be highly close to the true error.
Claims
1. A method for rapid mid-course calibration of UAV inertial positioning. It is characterized in that The following steps are involved: ① Initial setting: Initially set the error value according to the parameters of the inertial sensor used, and set the calibration waypoint list according to the calibration point set and the itinerary route, with the first calibration waypoint in the waypoint list as the current calibration waypoint; ②Enter range judgment: According to the current positioning point, judge whether it has entered the range of the current calibration waypoint plus N times the error value. If it has entered, proceed to the next step; ③ Comparison: Get the current aerial image, compare the current aerial image with the aerial image corresponding to the current calibration waypoint and calculate the similarity to get the current similarity; ④Similarity decay judgment: judge the size of the current similarity and the previous similarity. If the current similarity is greater than the previous similarity, delete the previous similarity, cache the current similarity as the previous similarity and return the value step ②. If the current similarity is less than the previous similarity, proceed to the next step. ⑤Replacement record: Calculate the difference between the current calibration waypoint and the current positioning point, calculate the calibration deviation based on the difference using the previous similarity as the coefficient, and update the positioning point and error value based on the calibration deviation; ⑥ Route end judgment: judge whether the current calibration waypoint is the last one in the calibration waypoint list. If not, set the next calibration waypoint as the current calibration waypoint and return to step ②. If yes, exit.
2. The method for rapid mid-course calibration of inertial positioning of a UAV as claimed in claim 1, It is characterized in that In step ③, the current aerial image and the aerial image corresponding to the current calibration waypoint are first pooled and then compared to calculate the similarity.
3. The method for rapid mid-course calibration of inertial positioning of a UAV as claimed in claim 2, It is characterized in that In the pooling operation, the filter size is M×M, the step size is M, and the value range of M is 0.5% to 2% of the number of pixels on the short side of the current aerial image and is rounded up.
4. The method for rapid mid-course calibration of inertial positioning of a UAV as claimed in claim 1, It is characterized in that In the step ②, the value of N is 0.7-3.
5. The method for rapid mid-course calibration of inertial positioning of a UAV as claimed in claim 1, It is characterized in that In the step ①, the error value is set to 3 to 6 times the parameter error of the inertial sensor used.
6. The method for rapid mid-course calibration of inertial positioning of a UAV as claimed in claim 1, It is characterized in that In the step ⑤, the calibration deviation is the multiplication of the previous similarity and the difference, and the previous similarity is first normalized.
7. The method for rapid mid-course calibration of inertial positioning of a UAV as claimed in claim 1, It is characterized in that In step ③, the similarity is calculated using a difference value hash algorithm.
8. The method for rapid mid-course calibration of inertial positioning of a UAV as claimed in claim 1, It is characterized in that In the step ⑤, the positioning point is updated according to the calibration deviation, and the positioning point obtained by adding the difference to the current calibration waypoint and subtracting the calibration deviation is used as the current positioning point.
9. The method for rapid mid-course calibration of inertial positioning of a UAV as claimed in claim 1, It is characterized in that In step ⑤, the error value is updated according to the calibration deviation by multiplying the ratio of the calibration deviation and the difference value by the error value as a new error value, and the error value is replaced by the new error value.
10. The method for rapid mid-course calibration of inertial positioning of a UAV as claimed in claim 1, It is characterized in that In the step ①, the calibration waypoint list is set according to the calibration point set and the itinerary route, the itinerary route is expanded by a preset range H to obtain a itinerary route range, the calibration points within the itinerary route range are extracted, and the extracted calibration points are sorted according to the itinerary route direction to obtain a calibration waypoint list.
Citation Information
Patent Citations
Unmanned aerial vehicle position estimation system and estimation method based on visual sensor
CN113110556A
Cited By
Defining and testing method for route of unmanned aerial vehicle
CN120101837A
