旋转门压缩算法(SDT)的Go实现
Posted bkzy
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了旋转门压缩算法(SDT)的Go实现相关的知识,希望对你有一定的参考价值。
原理
实质上,是计算门轴到新数据点之间线段的斜率,门轴既是线段的零点。由于直线段的公式为:
x = kt+b //k为斜率,b为0点,t为时间轴,x为数据大小
上下门轴点的计算方法为:
ub=x0+ΔE //上门轴,x0为存储点的数值
db=x0-ΔE //下门轴,x0为存储点的数值
上下门斜率的计算方法为:
uk=(xt-ub)/Δt //上门斜率
dk=(xt-db)/Δt //下门斜率
在t0点,门关闭,上门斜率无线小,下门斜率无限大。随着时间轴的右伸,上门斜率逐渐变大(单向,只能变大),下门斜率逐渐变小(单向,只能变小)。当上门斜率大于等于下门斜率时,保存前一个数据点。
具体流程为:
算法的实现
package models
import (
"time"
)
//旋转门压缩结构体
type SdtDoor struct
DeltaE float64 //门初始宽度
LastHisV float64 //上一次存储的数据值
LastHisT time.Time //上一次存储的数据时间戳
LastRealV float64 //上一次的实时数据值
LastRealT time.Time //上一次实时数据的时间戳
MaxIntervalSec int64 //数据存储最大间隔秒数
isinit bool //初始化状态
closed bool //初始关门状态
uk float64 //上斜率
dk float64 //下斜率
ub float64 //上零点
db float64 //下零点
//新建旋转门压缩实体
func NewSdtDoor(deltaE float64, maxsec int64) *SdtDoor
sdt := &SdtDoorDeltaE: deltaE, MaxIntervalSec: maxsec, closed: true, isinit: true
return sdt
//旋转门过滤器
func (sdt *SdtDoor) Filter(pointV float64, pointT time.Time) bool
if sdt.DeltaE == 0 //门宽度为零
return true //保存每一个数据
save := false //过滤检查结果
if sdt.isinit //初始化状态
sdt.isinit = false
save = true
else
deltaT := float64(pointT.Sub(sdt.LastHisT).Milliseconds())
//fmt.Printf("时间差:%fms\\n", deltaT)
if sdt.closed //开门第一个点
if deltaT > 0
sdt.closed = false
sdt.uk = (pointV - sdt.ub) / deltaT
sdt.dk = (pointV - sdt.db) / deltaT
else
uk := (pointV - sdt.ub) / deltaT
dk := (pointV - sdt.db) / deltaT
if uk > sdt.uk //上斜率只保存增大的
sdt.uk = uk
if dk < sdt.dk //下斜率只保存减小的
sdt.dk = dk
if sdt.dk <= sdt.uk //下斜率小于等于上斜率,触发保存
save = true
if !save
if deltaT/1000.0 > float64(sdt.MaxIntervalSec) //已经长时间没有触发保存
save = true
if save
sdt.LastHisV = sdt.LastRealV
sdt.LastHisT = sdt.LastRealT
sdt.ub = sdt.LastHisV + sdt.DeltaE
sdt.db = sdt.LastHisV - sdt.DeltaE
sdt.closed = true
sdt.LastRealV = pointV
sdt.LastRealT = pointT
return save
测试用例
func TestSdt(t *testing.T)
deltaE := 0.005 //门宽,即压缩精度
datas := []struct
tstamp int64 //时间戳,UNIX秒
sinv float64 //正玄波值
1651800000, 0,
1651800001, 0.0998334166468282,
1651800002, 0.198669330795061,
1651800003, 0.29552020666134,
1651800004, 0.389418342308651,
1651800005, 0.479425538604203,
1651800006, 0.564642473395035,
1651800007, 0.644217687237691,
1651800008, 0.717356090899523,
1651800009, 0.783326909627483,
1651800010, 0.841470984807897,
1651800011, 0.891207360061435,
1651800012, 0.932039085967226,
1651800013, 0.963558185417193,
1651800014, 0.98544972998846,
1651800015, 0.997494986604054,
1651800016, 0.999573603041505,
1651800017, 0.991664810452469,
1651800018, 0.973847630878195,
1651800019, 0.946300087687414,
1651800020, 0.909297426825682,
1651800021, 0.863209366648874,
1651800022, 0.80849640381959,
1651800023, 0.74570521217672,
1651800024, 0.675463180551151,
1651800025, 0.598472144103956,
1651800026, 0.515501371821464,
1651800027, 0.427379880233829,
1651800028, 0.334988150155904,
1651800029, 0.239249329213982,
1651800030, 0.141120008059866,
1651800031, 0.0415806624332896,
1651800032, -0.058374143427581,
1651800033, -0.15774569414325,
1651800034, -0.255541102026833,
1651800035, -0.350783227689621,
1651800036, -0.442520443294854,
1651800037, -0.529836140908495,
1651800038, -0.61185789094272,
1651800039, -0.687766159183975,
1651800040, -0.756802495307929,
1651800041, -0.818277111064411,
1651800042, -0.871575772413589,
1651800043, -0.916165936749456,
1651800044, -0.951602073889517,
1651800045, -0.977530117665097,
1651800046, -0.993691003633465,
1651800047, -0.999923257564101,
1651800048, -0.99616460883584,
1651800049, -0.982452612624332,
1651800050, -0.958924274663138,
1651800051, -0.925814682327731,
1651800052, -0.883454655720152,
1651800053, -0.8322674422239,
1651800054, -0.772764487555985,
1651800055, -0.705540325570389,
1651800056, -0.631266637872319,
1651800057, -0.550685542597635,
1651800058, -0.464602179413754,
1651800059, -0.373876664830233,
1651800060, -0.279415498198922,
1651800061, -0.182162504272092,
1651800062, -0.0830894028174929,
1651800063, 0.0168139004843542,
1651800064, 0.116549204850497,
1651800065, 0.215119988087819,
1651800066, 0.311541363513382,
1651800067, 0.404849920616602,
1651800068, 0.494113351138612,
1651800069, 0.578439764388203,
1651800070, 0.656986598718792,
1651800071, 0.72896904012588,
1651800072, 0.793667863849156,
1651800073, 0.850436620628567,
1651800074, 0.898708095811629,
1651800075, 0.937999976774741,
1651800076, 0.967919672031488,
1651800077, 0.988168233877001,
1651800078, 0.998543345374605,
1651800079, 0.998941341839772,
1651800080, 0.989358246623381,
1651800081, 0.969889810845085,
1651800082, 0.940730556679771,
1651800083, 0.902171833756291,
1651800084, 0.854598908088278,
1651800085, 0.798487112623487,
1651800086, 0.73439709787411,
1651800087, 0.662969230082178,
1651800088, 0.584917192891757,
1651800089, 0.50102085645788,
1651800090, 0.412118485241752,
1651800091, 0.319098362349347,
1651800092, 0.222889914100242,
1651800093, 0.124454423507058,
1651800094, 0.0247754254533542,
sdt := NewSdtDoor(deltaE, 100)
for i, dt := range datas
t := time.Unix(dt.tstamp, 0)
save := sdt.Filter(dt.sinv, t)
//fmt.Printf("%d uk=%f,dk=%f,diff=%f,save=%t\\n", i, sdt.uk, sdt.dk, sdt.uk-sdt.dk, save)
if save
if i == 0
fmt.Println(dt)
else
if i != 1 //第一个数前面已经输出过了,不再重复
fmt.Println(datas[i-1])
测试结果
1651800000 0
1651800001 0.0998334166468282
1651800005 0.479425538604203
1651800008 0.717356090899523
1651800011 0.891207360061435
1651800013 0.963558185417193
1651800015 0.997494986604054
1651800017 0.991664810452469
1651800019 0.946300087687414
1651800021 0.863209366648874
1651800024 0.675463180551151
1651800027 0.427379880233829
1651800032 -0.058374143427581
1651800036 -0.442520443294854
1651800039 -0.687766159183975
1651800042 -0.871575772413589
1651800044 -0.951602073889517
1651800046 -0.993691003633465
1651800048 -0.99616460883584
1651800050 -0.958924274663138
1651800052 -0.883454655720152
1651800055 -0.705540325570389
1651800058 -0.464602179413754
1651800062 -0.0830894028174929
1651800067 0.404849920616602
1651800070 0.656986598718792
1651800073 0.850436620628567
1651800075 0.937999976774741
1651800077 0.988168233877001
1651800079 0.998941341839772
1651800081 0.969889810845085
1651800083 0.902171833756291
1651800086 0.73439709787411
1651800089 0.50102085645788
1651800093 0.124454423507058
对上述测试数据绘图
参考文献
SDT旋转门压缩算法MFC图形测试
数据压缩算法:旋转门算法(SDT)的C#实现
以上是关于旋转门压缩算法(SDT)的Go实现的主要内容,如果未能解决你的问题,请参考以下文章
go路由httprouter中的压缩字典树算法图解及c++实现
算法题LeetCode-硬币划分问题-(动态规划斜率优化空间压缩)