CN106601032A - Multi-path terrain integrity detection method based on downward-looking sensor - Google Patents

Abstract

The invention discloses a multi-path terrain integrity detection method based on a downward-looking sensor. The multi-path terrain integrity detection method belongs to the field of avionics systems. In order to convince a pilot that a synthetic visual scene system can provide precise external scenes, a certain degree of integrity monitoring and alarming capabilities is required. The multi-path terrain integrity detection method disclosed by the invention is implemented by detecting consistency of an airborne terrain database and actual terrain data measured by the downward-looking sensor in real time, and sending alarm information to the pilot when the deviation between the airborne terrain database and the actual terrain data exceeds a set threshold value. The multi-path terrain integrity detection method specifically comprises an airborne terrain database sampling algorithm, a sensor measurement data synthetic terrain algorithm, a multi-path detection algorithm and a statistical checking algorithm. The checking can detect horizontal errors and vertical errors of the terrain data simultaneously, the detection precision is not affected by the attitude of an aircraft, the detection precision is high, the real-time performance is good, and the multi-path terrain integrity detection method is of great significance to the flight safety.

Description

A kind of multipath landform integrality detection method based on lower view sensor

Technical field

It is the invention belongs to avionics system field, and in particular in synthetic vision system, a kind of to be based on lower view sensor Multipath landform integrality detection method.

Background technology

In order that pilot believes that synthetic vision system (SVS) can provide accurate external sights, be safe from danger misleading Information, system requirements have certain integrity monitoring and alarm ability.If integrity detection finds airborne profile data base Mismatch with actual landform, display system will send alarm to pilot, point out Synthetic vision unreliable and can not use, with Reduce and provide to pilot harm misleading terrain information.

The purpose of landform integrity detection is that real-time detection airborne profile data base is measured practically with lower view sensor The consistent degree of graphic data, rather than the correctness of airborne profile data base.Its key technology, one is airborne profile database sampling mould The foundation of type, model should be close to echo altimeter measurement characteristicses to greatest extent, and to ensure accuracy of detection, how two be while examining Vertical error and horizontal transfer error are tested, and three is to reduce alarm time, warning information is sent to pilot in time, ensures flight peace Entirely.

The content of the invention

It is an object of the invention in solving synthetic vision system existing landform integrality detection method deficiency, it is proposed that A kind of multipath landform integrality detection method based on lower view sensor.

The multipath landform integrality detection method based on lower view sensor of the present invention, including following step：

Step one：Airborne profile database sampling；

Step 2：Calculate synthesis Terrain Elevation；

Step 3：Counting statistics inspected number；

Step 4：Multipath detected；

Step 5：Hypothesis testing, if the airborne profile data base actual landform data that measured with lower view sensor are differed Cause degree exceeds given threshold, then send warning information to pilot.

It is an advantage of the current invention that：

(1), in existing airborne profile database sampling algorithm, what sampling model assumed echo altimeter measurement mostly is winged Plumb height immediately below machine, this restriction cause sampling error of the aircraft in pitching and roll attitude, and this error Easily occur in the case of taking off and land etc. to landform integrity demands height.The sampling model proposed in the present invention, gram This restriction is taken, no matter which kind of attitude aircraft is in, and keeps high-precision landform integrity detection；

(2) in the present invention, when to airborne profile database sampling, intersecting detection and bilinearity difference have been used, has been protected While card sampling precision, efficiency is improve to greatest extent；

(3) existing landform integrity detection is used mostly single path detection, is only capable of detecting vertical error in topographic database, Particularly when physical features is flat, it is impossible to detect horizontal transfer error in topographic database, multipath inspection used in the present invention The method there is provided horizontal error in a kind of detection topographic database is surveyed, while detecting vertical error and horizontal error, is improve Landform integrity detection ability.

Description of the drawings

Fig. 1 is the flow chart of the present invention；

Fig. 2 is airborne profile database sampling model schematic；

Fig. 3 is synthesis landform instrumentation plan；

Fig. 4 is synthetic vision system landform integrity detection event tree.

Specific embodiment

Below in conjunction with drawings and Examples, the present invention is described in further detail.

The multipath landform integrality detection method based on lower view sensor of the present invention, by multipath detected and statistics Learn algorithm, the consistent degree of the actual landform data that real-time detection airborne profile data base is measured with lower view sensor, when the two is inclined When difference is more than given threshold, warning information is sent to pilot.

The present invention a kind of multipath landform integrality detection method based on lower view sensor, as shown in figure 1, including with Lower step：

Step one：Airborne profile database sampling.

As shown in Fig. 2 in synthetic vision system, with echo altimeter installation site as origin, surveying high comprising radar In meter beam area, ray is sent along echo altimeter antenna direction, carry out intersecting detection with landform, taken nearest with plane distance Point, as the sampled point compared with echo altimeter measured value, and calculates sampled point height value using bilinear interpolation method h_DEM(lat_DGPS(t_i),lon_DGPS(t_i)), and then obtain topographic database terrain profile.

Wherein, ray, and the nearest sampled point of selected distance aircraft are sent along echo altimeter antenna direction, and non-aircraft Vertical sampled point, it is therefore an objective to make sampling model closer to echo altimeter measurement characteristicses.Sampled point height value is calculated, except double Outside linear interpolation algorithm, closest interpolation method and cubic convolution interpolation method are also may be selected, but closest interpolation method precision is low, three times Convoluting interpolation method amount of calculation is excessive, is while meeting the requirement of real-time and precision, the present invention selects bilinear interpolation.

Step 2：Calculate synthesis Terrain Elevation.

As shown in figure 3, synthesis terrain profile is the ground level on more than sea level:

h_synth(t_i)=h_DGPS(t_i)-(h_radalt(t_i)+h_antenna)

Wherein, h_DGPSIt is height value of the aircraft for having DGPS to provide more than sea level；h_antennaGps antenna top with The difference in height of echo altimeter antenna low side；h_radaltIt is the measured value of echo altimeter.

Step 3：Counting statistics inspected number.

(1) echo altimeter is measured absolute error p of synthesis terrain profile and topographic database sampling terrain profile (t_i) be defined as：

p(t_i)=h_SYNT(t_i)-h_DEM(lat_DGPS,lon_DGPS)

Wherein, h_DEM(lat_DGPS,lon_DGPS) be topographic database mesorelief height, h_SYNT(t_i) be landform synthesis it is high Degree.

(2) to absolute error p (t_i) Kalman filtering is carried out, reduce sensor and digital elevation model (Digital Elevation Model, vehicle economy M) specified noise in data the deviation of sensor and DEM is estimated, comprise the following steps that：

1) initialize predictive valueAnd error variance

Wherein,It is 0 according to hypothesis system model,Span be ((15)²,(20)²)；

Calculate Kalman filter gain：

Wherein, K_kIt is Kalman filtering gain,It is the error variance of laststate, H_kIt is unit domain transformation matrix, R_kIt is The measured value of error variance, is constant；

2) update the estimated value of sampled value：

Wherein,It is t_kThe estimated value at moment,It is laststate estimated value, z_kIt is in t_kThe measured value at moment；

3) calculate the error variance of estimated value

Wherein, P_kIt is the error variance of estimated value；

4) calculate predictive value

Wherein,It is t_k+1The predictive value at moment, Φ_kFor unit state-transition matrix,It is t_k+1The prediction at moment is missed Difference variance, Q_kIt is system noise variance, referred to as tuner parameters of wave filter are constant；

5) return 2), repeat above procedure, terminate to system operation.

(3) define statistical test amount：

Wherein, T is statistical test amount, and P is the estimated value of error variance, and N is the number of sampled point,It is through card Absolute error after Kalman Filtering.

Step 4：Multipath detected.

(1) flight path is carried out horizontal-shift by dummy level DEM Transfer Errors, obtains a plurality of flight path, according to step Method in rapid three, calculates multipath statistical test amount：

Wherein, T (m, n) is statistical test amount；

lat_DGPS=lat_DGPS(t_i)+lat_off_mm∈(1,M)；

lon_DGPS=lon_DGPS(t_i)+lon_off_n n∈(1,N)。

(2) statistical test amount T in calculating 1, possesses the offset path of minimum T value, it is considered to be aircraft is in topographic database Current horizontal location, be defined as：

Wherein, min () function returns estimated location,It is the minima of statistic T.

Define T_V(t_i)=T_minFor vertical error inspected number.T_minIt is inspected number minima.

(3) estimated location is defined as topographic database estimated location lon_DBPAnd lat_DBP；Actual position is defined as lon_trueAnd lat_true, calculate horizontal error between the two：

p_H=dist ([lon_DBP,lat_DBP],[lon_true,lat_true])

Wherein, dist () returns 2 points in space of distance；lon_true,lat_trueFor the measurement position of DGPS.

(4) calculated level error statistics inspected number.

Wherein σ_pFor standard deviation.

Step 5：Hypothesis testing is carried out, event tree is as shown in Figure 4.If airborne profile data base is measured with lower view sensor Actual landform data inconsistent degree exceed given threshold, then send warning information to pilot.

(1) two mutually contradictory hypothesis are proposed：

Null hypothesises H₀：System is operated in rated condition or does not find run-time error；

Wherein,It is 0 to obey average, and standard deviation is σ_pNormal distribution；

Alternative hypothesiss H₁：System operation finds mistake；

Wherein,Obedience average is μ_B, standard deviation is σ_pNormal distribution.

(2) according to synthetic vision system as auxiliary, key element, three kinds of different applications of tactics key element, to omitting alarm Probability sets three secure thresholds and determines fault alarm P_FFD, omit alarm P_MDProbability, for three kinds of applications to safe class It is different to require, omit alarm probability and be respectively set as：Less than 10^-3, 10^-4～10^-7,10^-6～10^-9, wherein

P_FFD=P (detect mistake | H₀)·P(H₀)

P_MD=P (it is not detected by mistake | H₁)·P(H₁)

Jing practical flights are tested, and take inspection threshold value T of suitable vertical error detection amount and horizontal error inspected number_VDAnd T_HD Value, meets P_FFDAnd P_MDRequire.

(3) vertical error inspected number and horizontal error inspected number are tested respectively and is made a policy, if being less than step 3rd, vertical check threshold value T and horizontal check threshold value T in step 4_H, then receive H₀, conversely, then refusing H₀。

(4) as refusal H₀When, synthetic vision system sends warning information to pilot, as reception H₀When, then it represents that system is transported Row is normal.

Particular embodiments described above, has been carried out to the purpose of the present invention, technical scheme and beneficial effect further in detail Describe bright, the be should be understood that specific embodiment that the foregoing is only the present invention in detail, be not limited to the present invention, it is all Within the spirit and principles in the present invention, any modification, equivalent substitution and improvements done etc., should be included in the guarantor of the present invention Within the scope of shield.

Claims (2)

1. a kind of multipath landform integrality detection method based on lower view sensor, comprises the following steps：

Step one：Airborne profile database sampling；

In synthetic vision system, with echo altimeter installation site as origin, in echo altimeter beam area, along radar Cathetometer antenna direction sends ray, carries out intersecting detection with landform, takes and plane distance closest approach, as with echo altimeter The sampled point that measured value compares, calculates sampled point height value h_DEM(lat_DGPS(t_i),lon_DGPS(t_i)), wherein, lat_DGPS(t_i) Represent t_iMoment latitude determination value, lon_DGPS(t_i) represent t_iMoment longitude measurement；

Step 2：Calculate synthesis Terrain Elevation；

Synthesis terrain profile is the ground level on more than sea level:

h_synth(t_i)=h_DGPS(t_i)-(h_radalt(t_i)+h_antenna)

Wherein, h_DGPSThe height value for being aircraft more than sea level；h_antennaIt is that gps antenna top is low with echo altimeter antenna The difference in height at end；h_{rada lt}It is the measured value of echo altimeter；

Step 3：Counting statistics inspected number；

(1) obtain the absolute error p (t of echo altimeter measurement synthesis terrain profile and topographic database sampling terrain profile_i)：

p(t_i)=h_SYNT(t_i)-h_DEM(lat_DGPS,lon_DGPS)

Wherein, h_DEM(lat_DGPS,lon_DGPS) be topographic database mesorelief height, h_SYNT(t_i) be landform synthesis height；

(2) to absolute error p (t_i) carry out Kalman filtering；

(3) obtain statistical test amount：

$T=\frac{1}{P}\underset{k=1}{\overset{N}{\Σ}}{\hat{x}}^2\left(t_k\right)$

Wherein, T is statistical test amount, and P is the estimated value of error variance, and N is the number of sampled point,It is to filter through Kalman Absolute error after ripple；

Step 4：Multipath detected；

(1) flight path is carried out horizontal-shift by dummy level DEM Transfer Errors, obtains a plurality of flight path, calculates multipath Statistical test amount：

Wherein, T (m, n) is statistical test amount；M, N represent latitude, longitude side-play amount maximum respectively；

lat_DGPS=lat_DGPS(t_i)+lat_off_mm∈(1,M)；

lon_DGPS=lon_DGPS(t_i)+lon_off_nn∈(1,N)；

(2) set the offset path of minimum T value as aircraft topographic database current horizontal location：

$\[\hat{m},\hat{n}\]=min\left(T\right)$

Wherein, min () function returns estimated location,It is the minima of statistic T；

(3) latitude of estimated location, longitude are the latitude and longitude of topographic database estimated location, are respectively defined as lon_DBP And lat_DBP；The longitude of actual position, latitude are respectively lon_trueAnd lat_true, calculate horizontal error between the two：

p_H=dist ([lon_DBP,lat_DBP],[lon_true,lat_true])

Wherein, dist () returns 2 points in space of distance；

(4) calculated level error statistics inspected number；

$T_H\left(t_i\right)=\frac{1}{{{\mathrm{\σ}}_p}^2}\underset{i=1}{\overset{N}{\Σ}}{p_H}^2\left(t_i\right)$

Wherein：σ_pFor standard deviation；

Step 5：Hypothesis testing is carried out, if the airborne profile data base actual landform data that measured with lower view sensor are differed Cause degree exceeds given threshold, then send warning information to pilot；

(1) two mutually contradictory hypothesis are proposed：

Null hypothesises H₀：Synthetic vision system is operated in rated condition or does not find run-time error；

$H_0:p_0~N(0,{\mathrm{\σ}}_p^2)$

Wherein,It is 0 to obey average, and standard deviation is σ_pNormal distribution；

Alternative hypothesiss H₁：Synthetic vision system operation finds mistake；

$H_1:p_1~N({\mathrm{\μ}}_B,{\mathrm{\σ}}_p^2)$

Wherein,Obedience average is μ_B, standard deviation is σ_pNormal distribution；

(2) determine the fault alarm P of synthetic vision system_FFD, omit alarm P_MDProbability：

P_FFD=P (detect mistake | H₀)·P(H₀)

P_MD=P (it is not detected by mistake | H₁)·P(H₁)

(3) vertical error inspected number and horizontal error inspected number are tested respectively and is made a policy, if less than step 3, step Vertical check threshold value and horizontal check threshold value in rapid four, then receive H₀, conversely, then refusing H₀；

(4) as refusal H₀When, synthetic vision system sends warning information to pilot, as reception H₀When, then it represents that Synthetic vision system System normal operation.

2. a kind of multipath landform integrality detection method based on lower view sensor according to claim 1, described In step 3, (2) are specially：

1) initialize predictive valueAnd error variance It is 0 according to hypothesis system model,Span be ((15)², (20)²)；

Calculate Kalman filter gain：

$K_k=P_k^{-}{H}_k^T{(H_k{P}_k^{-}{H}_k^T+R_k)}^{-1}$

Wherein, K_kIt is Kalman filtering gain,It is the error variance of laststate, H_kIt is unit domain transformation matrix, R_kIt is error The measured value of variance；

2) update the estimated value of sampled value：

${\hat{x}}_k={\hat{x}}_k^{-}+K_k(z_k-H_k{\hat{x}}_k^{-})$

Wherein,It is t_kThe estimated value at moment,It is laststate estimated value, z_kIt is in t_kThe measured value at moment；

3) calculate the error variance of estimated value

$P_k=(I-K_k{H}_k)P_k^{-}$

Wherein, P_kIt is the error variance of estimated value；

4) calculate predictive value

${\hat{x}}_{k+1}^{-}={\mathrm{\Φ}}_k{\hat{x}}_k$

$P_{k+1}^{-}={\mathrm{\Φ}}_k{P}_k{\mathrm{\Φ}}_k^T+Q_k$

Wherein,It is t_k+1The predictive value at moment, Φ_kFor unit state-transition matrix,It is t_k+1The forecast error side at moment Difference, Q_kIt is system noise variance, referred to as tuner parameters of wave filter；

5) return to step 2), until terminate.