Ambiguity resolved precise point positioning with GPS and BeiDou

This paper focuses on the contribution of the global positioning system (GPS) and BeiDou navigation satellite system (BDS) observations to precise point positioning (PPP) ambiguity resolution (AR). A GPS + BDS fractional cycle bias (FCB) estimation method and a PPP AR model were developed using integrated GPS and BDS observations. For FCB estimation, the GPS + BDS combined PPP float solutions of the globally distributed IGS MGEX were first performed. When integrating GPS observations, the BDS ambiguities can be precisely estimated with less than four tracked BDS satellites. The FCBs of both GPS and BDS satellites can then be estimated from these precise ambiguities. For the GPS + BDS combined AR, one GPS and one BDS IGSO or MEO satellite were first chosen as the reference satellite for GPS and BDS, respectively, to form inner-system single-differenced ambiguities. The single-differenced GPS and BDS ambiguities were then fused by partial ambiguity resolution to increase the possibility of fixing a subset of decorrelated ambiguities with high confidence. To verify the correctness of the FCB estimation and the effectiveness of the GPS + BDS PPP AR, data recorded from about 75 IGS MGEX stations during the period of DOY 123-151 (May 3 to May 31) in 2015 were used for validation. Data were processed with three strategies: BDS-only AR, GPS-only AR and GPS + BDS AR. Numerous experimental results show that the time to first fix (TTFF) is longer than 6 h for the BDS AR in general and that the fixing rate is usually less than 35 % for both static and kinematic PPP. An average TTFF of 21.7 min and 33.6 min together with a fixing rate of 98.6 and 97.0 % in static and kinematic PPP, respectively, can be achieved for GPS-only ambiguity fixing. For the combined GPS + BDS AR, the average TTFF can be shortened to 16.9 min and 24.6 min and the fixing rate can be increased to 99.5 and 99.0 % in static and kinematic PPP, respectively. Results also show that GPS + BDS PPP AR outperforms single-system PPP AR in terms of convergence time and position accuracy.     

BDS/GPS精密单点定位收敛时间与定位精度的比较

采用武汉大学卫星导航定位技术研究中心发布的北斗精密卫星轨道和钟差,在TriP 2.0软件的基础上实现了BDS PPP定位算法,并利用大量实测数据进行了BDS/GPS静态PPP和动态PPP浮点解试验。结果表明,BDS静态PPP的收敛时间约为80min,动态PPP的收敛时间为100min;对于3h的观测数据,静态PPP收敛后定位精度优于5cm,动态PPP收敛后水平方向优于8cm,高程方向约12cm;与GPS PPP类似,东分量上定位精度较北分量稍差。当前由于BDS的全球跟踪站有限,精密轨道和钟差精度不如GPS,因此BDS PPP的收敛时间较GPS长,但收敛后可实现厘米至分米级的绝对定位。     

Initial results of precise orbit and clock determination for COMPASS navigation satellite system

The development of the COMPASS satellite system is introduced, and the regional tracking network and data availability are described. The precise orbit determination strategy of COMPASS satellites is presented. Data of June 2012 are processed. The obtained orbits are evaluated by analysis of post-fit residuals, orbit overlap comparison and SLR (satellite laser ranging) validation. The RMS (root mean square) values of post-fit residuals for one month’s data are smaller than 2.0 cm for ionosphere-free phase measurements and 2.6 m for ionosphere-free code observations. The 48-h orbit overlap comparison shows that the RMS values of differences in the radial component are much smaller than 10 cm and those of the cross-track component are smaller than 20 cm. The SLR validation shows that the overall RMS of observed minus computed residuals is 68.5 cm for G01 and 10.8 cm for I03. The static and kinematic PPP solutions are produced to further evaluate the accuracy of COMPASS orbit and clock products. The static daily COMPASS PPP solutions achieve an accuracy of better than 1 cm in horizontal and 3 cm in vertical. The accuracy of the COMPASS kinematic PPP solutions is within 1–2 cm in the horizontal and 4–7 cm in the vertical. In addition, we find that the COMPASS kinematic solutions are generally better than the GPS ones for the selected location. Furthermore, the COMPASS/GPS combinations significantly improve the accuracy of GPS only PPP solutions. The RMS values are basically smaller than 1 cm in the horizontal components and 3–4 cm in the vertical component.     

基于非差观测量的BDS+GPS近实时钟差估计

卫星钟差的难预测性是影响实时高精度定位的重要因素之一。为快速获得高精度位置或对流层等信息，在非差观测模型的基础上，本文提出了一种延迟量约1 h的近实时钟差估计策略，该策略主要包含超快速轨道解算和钟差估计两部分。经验证，预报部分第2~5 h的GPS轨道三维平均精度为3.85 cm，BDS GEO和IGSO+MEO轨道三维平均精度分别为81.4和21.74 cm。基于超快速轨道可获得近实时钟差精度GPS为0.054 ns，BDS为0.12 ns。最后通过BDS+GPS静态PPP试验验证了轨道和钟差的可用性。     

Integrating GPS and BDS to shorten the initialization time for ambiguity-fixed PPP

The main challenge of ambiguity resolution in precise point positioning (PPP) is that it requires 30 min or more to succeed in the first fixing of ambiguities. With the full operation of the BeiDou (BDS) satellite system in East Asia, it is worthwhile to investigate the performance of GPS + BDS PPP ambiguity resolution, especially the improvements of the initial fixing time and ambiguity-fixing rate compared to GPS-only solutions. We estimated the wide- and narrow-lane fractional-cycle biases (FCBs) for BDS with a regional network, and PPP ambiguity resolution was carried out at each station to assess the contribution of BDS. The across-satellite single-difference (ASSD) GPS + BDS combined ambiguity-fixed PPP model was used, in which the ASSD is applied within each system. We used a two-day data set from 48 stations. For kinematic PPP, the percentage of fixing within 10 min for GPS only (Model A) is 17.6 %, when adding IGSO and MEO of BDS (Model B), the percentage improves significantly to 42.8 %, whereas it is only 23.2 % if GEO is added (Model C) due to the low precision of GEO orbits. For static PPP, the fixing percentage is 32.9, 53.3 and 28.0 % for Model A, B and C, respectively. In order to overcome the limitation of the poor precision of GEO satellites, we also used a small network of 10 stations to analyze the contribution of GEO satellites to kinematic PPP. We took advantage of the fact that for stations of a small network the GEO satellites appear at almost the same direction, such that the GEO orbit error can be absorbed by its FCB estimates. The results show that the percentage of fixing improves from 39.5 to 57.7 % by adding GEO satellites.     

Multi-GNSS precise point positioning (MGPPP) using raw observations

A joint-processing model for multi-GNSS (GPS, GLONASS, BDS and GALILEO) precise point positioning (PPP) is proposed, in which raw code and phase observations are used. In the proposed model, inter-system biases (ISBs) and GLONASS code inter-frequency biases (IFBs) are carefully considered, among which GLONASS code IFBs are modeled as a linear function of frequency numbers. To get the full rank function model, the unknowns are re-parameterized and the estimable slant ionospheric delays and ISBs/IFBs are derived and estimated simultaneously. One month of data in April, 2015 from 32 stations of the International GNSS Service (IGS) Multi-GNSS Experiment (MGEX) tracking network have been used to validate the proposed model. Preliminary results show that RMS values of the positioning errors (with respect to external double-difference solutions) for static/kinematic solutions (four systems) are 6.2 mm/2.1 cm (north), 6.0 mm/2.2 cm (east) and 9.3 mm/4.9 cm (up). One-day stabilities of the estimated ISBs described by STD values are 0.36 and 0.38 ns, for GLONASS and BDS, respectively. Significant ISB jumps are identified between adjacent days for all stations, which are caused by the different satellite clock datums in different days and for different systems. Unlike ISBs, the estimated GLONASS code IFBs are quite stable for all stations, with an average STD of 0.04 ns over a month. Single-difference experiment of short baseline shows that PPP ionospheric delays are more precise than traditional leveling ionospheric delays.     

北斗/GPS实时精密卫星钟差融合解算模型及精度分析

实时GNSS精密单点定位(PPP)技术必须使用实时的高精度卫星精密轨道和钟差。本文研究了精密卫星钟差融合解算模型及策略,并利用滤波算法实现了北斗/GPS实时精密卫星钟差融合估计算法。仿真实时试验结果显示:获得的北斗/GPS实时钟差与GFZ事后多GNSS精密钟差(GBM)的标准差在0.15 ns左右;使用该钟差进行GPS动态PPP试验,收敛后水平精度优于5 cm,高程精度优于10 cm;使用仿真实时钟差进行的北斗动态PPP与使用GFZ事后多GNSS精密钟差开展的试验相比精度相当,可实现分米级定位。     

Performance assessment of single- and dual-frequency BeiDou/GPS single-epoch kinematic positioning

The first results of the short baseline single-epoch kinematic positioning based on dual-frequency real BeiDou/GPS data are presented. The performance of the BeiDou/GPS single-epoch positioning is demonstrated in both static and kinematic modes and compared with corresponding GPS-only performance. It is shown that the availability and reliability of the single-frequency BeiDou/GPS and dual-frequency BeiDou single-epoch kinematic positioning are comparable to those of the dual-frequency GPS. The fixed rate and reliability of ambiguity resolution for the single- and dual-frequency BeiDou/GPS are remarkably improved as compared to that of GPS-only, especially in case of high cutoff elevations. For positioning accuracy with fixed ambiguities, the BeiDou/GPS single-epoch solutions are improved by 23 and 4 % relative to the GPS-only case for two short baseline tests of 8 km, respectively. These results reveal that dual-frequency BeiDou real-time kinematic (RTK) is already applicable in Asia–Pacific areas and that single-frequency BeiDou/GPS RTK is also achievable but only with initialization of several seconds. More promisingly, the dual-frequency BeiDou/GPS RTK can overcome the difficulties with GPS-only RTK under the challenging conditions assuming, of course, that the additional BeiDou satellites are visible.     

估计接收机差分码偏差的GPS/BDS非组合精密单点定位模型

在传统多系统非差非组合精密单点定位（precise point positioning，PPP）模型中，电离层延迟会吸收部分接收机码硬件延迟，其估计值可能为负数。提出了一种估计接收机差分码偏差（differential code bias，DCB）参数的GPS（Global Positioning System）/BDS（BeiDou Navigation Satellite System）非组合PPP模型，将每个系统第1个频率上的接收机码硬件延迟约束为零，对接收机DCB进行参数估计，达到了分离电离层延迟和接收机码硬件延迟的目的，降低了接收机钟差和电离层延迟的相关程度。利用4个多星座实验（multi-GNSS experiment，MGEX）跟踪站的GPS/BDS数据进行了静态和动态PPP试验，结果表明，与不估计DCB参数的PPP模型相比，采用估计DCB参数PPP模型后，静态模式下定位精度和收敛速度平均提高了29.3%和29.8%，动态模式下定位精度和收敛速度平均提高了15.7%和21.6%。     

北斗区域导航系统的PPP精度分析

北斗卫星导航系统的开放运行为其在高精度领域的应用提供了可能,系统精密单点定位性能受到了极大关注。本文首先介绍了北斗区域导航系统的星座和BDS/GPS跟踪网,分析了基于国内布站定轨的北斗卫星精密轨道和钟差精度。在此基础上研究了北斗区域导航系统静态、动态精密单点定位精度,并与GPS定位结果进行比较。实测算例表明:北斗精密单点定位可以实现静态厘米级、动态分米级的定位精度,达到目前GPS精密单点定位水平。     

北斗三频数据的两个无电离层组合轨道钟差估计及其应用

本文针对全球连续监测评估系统（iGMAS）和国际多系统GNSS试验计划（MGEX）两个观测网接收到不同频率北斗卫星数据的情况,提出了一种北斗卫星（BDS）3个频率（B1I、B2I、B3I）的两种无电离层组合（B1/B3和B1/B2）数据精密定轨（POD）和钟差估计（CE）方法。该方法可以统一处理上述两个观测网收到的北斗二代（BDS-2）,北斗三代试验系统（BDS-3e）和北斗三代全球系统（BDS-3g）3个频率的观测数据,并在一次程序运行中对所有北斗卫星进行联合处理,可有效提高一次运行的数据使用率,从而提高参数估计精度。采集了多天iGMAS、MGEX的GPS和BDS数据进行试验。结果表明,对BDS-3e+BDS-2+GPS联合定轨时,采用三频两组合方法后由于增强了观测几何,BDS轨道重叠RMS为15.9 cm,比传统双频法定轨精度提高11.3%。新方法引入了与卫星端3个频率相关的码偏差,该量多天估计结果稳定,证明了模型和方法可靠。将新方法用于BDS-3g+BDS-3e+BDS-2+GPS联合定轨,6颗BDS-3g的MEO卫星轨道重叠RMS为14.5 cm,钟差重叠RMS为0.43 ns,与BDS-3e的15.1 cm和0.49 ns相当。开展了北斗卫星精密单点定位（PPP）试验,结果显示增加了BDS-3g的6颗MEO的精密轨道和钟差后,测站定位精度水平为39.6 mm,天顶为37.8 mm,比仅用BDS-2和BDS-3e卫星定位精度提高了11.1%。     

Multi-GNSS phase delay estimation and PPP ambiguity resolution: GPS,BDS, GLONASS,Galileo

This paper focuses on the precise point positioning (PPP) ambiguity resolution (AR) using the observations acquired from four systems: GPS, BDS, GLONASS, and Galileo (GCRE). A GCRE four-system uncalibrated phase delay (UPD) estimation model and multi-GNSS undifferenced PPP AR method were developed in order to utilize the observations from all systems. For UPD estimation, the GCRE-combined PPP solutions of the globally distributed MGEX and IGS stations are performed to obtain four-system float ambiguities and then UPDs of GCRE satellites can be precisely estimated from these ambiguities. The quality of UPD products in terms of temporal stability and residual distributions is investigated for GPS, BDS, GLONASS, and Galileo satellites, respectively. The BDS satellite-induced code biases were corrected for GEO, IGSO, and MEO satellites before the UPD estimation. The UPD results of global and regional networks were also evaluated for Galileo and BDS, respectively. As a result of the frequency-division multiple-access strategy of GLONASS, the UPD estimation was performed using a network of homogeneous receivers including three commonly used GNSS receivers (TRIMBLE NETR9, JAVAD TRE_G3TH DELTA, and LEICA). Data recorded from 140 MGEX and IGS stations for a 30-day period in January in 2017 were used to validate the proposed GCRE UPD estimation and multi-GNSS dual-frequency PPP AR. Our results show that GCRE four-system PPP AR enables the fastest time to first fix (TTFF) solutions and the highest accuracy for all three coordinate components compared to the single and dual system. An average TTFF of 9.21 min with \(7{^{\circ }}\) cutoff elevation angle can be achieved for GCRE PPP AR, which is much shorter than that of GPS (18.07 min), GR (12.10 min), GE (15.36 min) and GC (13.21 min). With observations length of 10 min, the positioning accuracy of the GCRE fixed solution is 1.84, 1.11, and 1.53 cm, while the GPS-only result is 2.25, 1.29, and 9.73 cm for the east, north, and vertical components, respectively. When the cutoff elevation angle is increased to \(30{^{\circ }}\), the GPS-only PPP AR results are very unreliable, while 13.44 min of TTFF is still achievable for GCRE four-system solutions.     

Real-time clock offset prediction with an improved model

The GPS orbit precision of the IGS ultra-rapid predicted (IGU-P) products has been remarkably improved since 2007. However, the satellite clock offsets of the IGU-P products have not shown sufficient high-quality prediction to achieve sub-decimeter precision in real-time precise point positioning (RTPPP), being at the level of 1–3 ns (30–90 cm) RMS in recent years. An improved prediction model for satellite clocks is proposed in order to enhance the precision of predicted clock offsets. First, the proposed prediction model adds a few cyclic terms to absorb the periodic effects, and a time adaptive function is used to adjust the weight of the observation in the prediction model. Second, initial deviations of the predictions are reduced by using a recomputed constant term. The simulation results have shown that the proposed prediction model can give a better performance than the IGU-P clock products and can achieve precision better than 0.55 ns (16.5 cm) in real-time predictions. In addition, the RTPPP method was chosen to test the efficiency of the new model for real-time static and kinematic positioning. The numerical examples using the data set of 140 IGS stations show that the static RTPPP precision based on the proposed clock model has been improved about 22.8 and 41.5 % in the east and height components compared to the IGU-P clock products, while the precisions in the north components are the equal. The kinematic example using three IGS stations shows that the kinematic RTPPP precision based on the proposed clock model has improved about 30, 72 and 44 % in the east, north and height components.     

BDS不同轨道卫星精密单点定位性能分析

为了分析北斗不同轨道卫星对定位结果的影响,从而更好地利用我国自主研发的北斗卫星导航系统。该文采用亚太地区7个MGEX测站12d观测数据,进行静态、后处理动态和模拟实时动态3种模式的精密单点定位实验。实验结果表明,在北斗3类轨道卫星等权的情况下,倾斜地球同步轨道(IGSO)卫星对定位结果贡献最大;北斗两类轨道卫星组合中,IGSO+MEO组合定位精度最高,其静态精密单点定位(PPP)在E、N、U方向的RMS分别为0.62、0.39、3.71cm,后处理动态和模拟实时动态PPP的RMS为分米级;北斗各类轨道卫星与GPS组合定位中,GPS+IGSO+MEO组合定位结果收敛速度最快,收敛时间为26.30min。     

北斗三频精密单点定位模型比较及定位性能分析

在Trip软件的基础上实现了北斗三频无电离层两两组合、三频消电离层组合和三频非组合精密单点定位（precise point positioning,PPP）算法。利用12个陆态网观测站的北斗三频观测数据对3种三频PPP定位模型及传统的双频无电离层组合PPP模型的定位性能进行分析。试验结果表明,对大多数测站,3种三频PPP模型静态定位精度水平方向优于1 cm,高程方向优于2 cm,动态定位精度水平方向优于4 cm,高程方向优于6 cm;3种三频PPP模型静态收敛时间约为120 min,动态收敛时间约180 min;相比于传统的双频PPP模型,三频PPP模型的定位精度有所提高,其中,三频非组合模型静态单天解RMS在水平方向和高程方向分别提高36.1%和6.3%,动态单天解RMS在水平方向和高程方向分别提高9.1%和6.3%。     

BDS/GPS组合精密单点定位精度分析与评价

精密单点定位（precise point positioning,PPP）已经广泛应用于许多领域,如测绘、交通、导航、地震监测等。近些年来,随着卫星数量的增多,多系统组合呈现越来越明显的趋势。利用全球MGEX（Multi-GNSS Experiment）网数据研究了BDS（BeiDou navigation satellite system）/GPS（global positioning system）组合精密单点定位技术,并与BDS单系统和GPS单系统进行了对比。结果表明,在静态定位中,BDS PPP在E、N、U方向的均方根误差分别为4.35 cm、3.01 cm、6.40 cm;GPS PPP在E、N、U方向的均方根误差分别为1.21 cm、0.48 cm、1.79 cm;BDS/GPS组合PPP在E、N、U方向的均方根误差分别为1.21 cm、0.50 cm、1.87 cm。在动态定位中,BDS PPP外符合精度水平方向优于10 cm,高程方向优于15 cm;GPS PPP和BDS/GPS组合PPP的外符合精度水平方向均优于5 cm,高程方向均优于8 cm。另外,无论是在静态还是动态的PPP中,组合系统相对于单系统,能大大缩短收敛时间,减少定位结果抖动,尤其是相对于BDS PPP来说,优势更为明显。     

Spatial–temporal characteristic of BDS phase delays and PPP ambiguity resolution with GEO/IGSO/MEO satellites

GPS precise point positioning (PPP) ambiguity resolution (AR) can improve the positioning accuracy and shorten the convergence time. However, for the BeiDou Satellite Navigation System (BDS), the problems of satellite-induced code bias, imperfections in the error models and the inadequate accuracy of orbit products limit the applications of the BDS PPP AR system, which requires more than 6 h to achieve the first ambiguity-fixed solution. In this study, the accuracy of a wide-lane (WL) uncalibrated phase delay (UPD) is improved after careful consideration of the code bias and multipath. Meanwhile, the accuracy of the BDS float ambiguity is also improved by multi-GNSS fusion and improved precise orbit and clock products, which are critical for high-quality narrow-lane (NL) UPD estimations. With three tracking networks of different scales, including Hong Kong, the Crustal Movement Observation Network of China (CMONOC) and the multi-GNSS experiment (MGEX) networks, the spatial–temporal characteristics of WL and NL UPDs for BDS GEO/IGSO/MEO satellites are analyzed, and the PPP AR is performed. Numerous results show that WL and NL UPDs with a standard deviation (STD) of less than 0.15 cycles can be achieved for BDS GEO satellites, while a STD of less than 0.1 cycles can be obtained for IGSO and MEO satellites. With the precise UPD estimation, for the first time, the BDS PPP rapid ambiguity resolution for GEO/IGSO/MEO satellites is achieved. We found that the average time to first fix (TTFF) of the BDS PPP AR is shortened significantly, to approximately 40 min for Hong Kong and the CMONOC, while the TTFF was 57.4 min for the MGEX networks. With ambiguity resolution, the accuracy of the daily BDS PPP in the east, north and vertical directions improves from 1.74 cm, 1.08 cm, and 5.52 cm to 0.72 cm, 0.54 cm, and 3.21 cm for the Hong Kong network, 2.24 cm, 2.31 cm, and 5.64 cm to 1.18 cm, 0.79 cm, and 3.30 cm for the CMONOC, and 2.71 cm, 1.80 cm, and 6.00 cm to 1.58 cm, 1.15 cm, and 4.33 cm for the MGEX networks. Significant improvement is also achieved for kinematic PPP, with improvements of 40.41%, 34.33% and 37.17% in the east, north and vertical directions for the MGEX networks, respectively.     

iGMAS GNSS钟差产品综合与评估

高精度卫星导航定位离不开精密钟差改正信息,钟差产品综合与评估是产品综合与服务中心(ISC)的重要任务之一。本文首先介绍了ISC钟差产品综合策略和抗差估计方法,重点阐述了非线性系统误差处理、钟差参考基准统一及残余线性偏差补偿;然后,对比逐历元对准方法和"三步"校准法两种不同综合产品的精度和稳定性,并对iGMAS近两年的钟差产品进行了评估和详细分析;最后,为验证综合产品的精度和一致性,采用5个测站的静态数据进行了PPP测试。计算结果表明:综合钟差的RMS最优且稳定,可作为评估分析中心产品精度的参考解。此外,综合产品北斗单系统静态PPP的结果优于多数分析中心,位置精度在4cm以内。     

PPP/PPP-RTK新进展与北斗/GNSS PPP定位性能比较

首先简要回顾了精密单点定位（PPP）技术在最近几年的发展现状,重点总结了高采样率钟差实时快速估计、多系统组合PPP模糊度固定、多频GNSS PPP模型及其模糊度固定、PPP快速初始化、PPP-RTK等若干热点方向的最新研究进展。在此基础上,利用目前四大卫星导航系统（GPS、GLONASS、Galileo、北斗）最新的实际观测数据,全面比较分析了各系统及多系统组合PPP定位性能,重点给出了北斗二号+北斗三号PPP浮点解和固定解的定位精度、收敛时间和首次固定时间。结果表明：我国北斗导航卫星系统已经可以实现与其他导航卫星系统基本相当的PPP定位性能。北斗二号+北斗三号组合PPP的收敛时间/首次固定时间20~30 min;静态解的东、北、天方向定位精度在毫米到厘米级;动态解水平方向约5 cm,高程方向约7 cm;多系统组合可显著提高PPP定位精度、收敛时间和首次固定时间：固定解定位精度比浮点解在东、北、天方向分别提升了14.8%、12.0%和12.8%;相比单GPS,多系统组合PPP浮点解的收敛时间和固定解首次固定时间分别缩短了36.5%和40.4%。     

GPS/BDS三频信号频间时钟偏差估计与定位验证

在三频GNSS应用中,受精密产品频率基准不一致的影响,会引入系统性偏差,即频间时钟偏差（IFCB）。本文首先通过对IFGF组合观测值进行历元差分,利用全球分布的80个MGEX观测站及中国区域内100个连续运行参考站,在2021年年积日（DOY）153—160 d的实测数据,进行了IFCB的估计并分析了其时变特性;然后将IFCB的估计结果运用到非差非组合PPP中。结果表明：GPS BLOCK Ⅱ-F的IFCB较大,幅值可达14 cm,GPS BLOCK Ⅲ与BDS的IFCB则较小,一般不超过5 cm。在定位验证中,经过IFCB改正后,GPS/BDS-2/BDS-3-IGSO在第3频点L5、B2I、B2a的相位残差分别减小了59.54%、26.31%、10.98%。其中,动态定位的GPS、BDS-2/BDS-3-IGSO、GPS/BDS-2/BDS-3-IGSO 3种方案的点位精度分别提升了56.55%、29.16%、20.72%,改善效果显著。