denoise.py

import pandas as pd
import numpy as np
import os
def smooth(path):
    #path = '41_predict.csv'
    df = pd.read_csv(path)
    df = df.fillna(0)

    x = df['X'].tolist()
    y = df['Y'].tolist()
    vis = df['Visibility'].tolist()

    # Define distance
    pre_dif = []
    for i in range(0,len(x)):
        if i ==0:
            pre_dif.append(0)
        else:
            pre_dif.append(((x[i]-x[i-1])**2+(y[i]-y[i-1])**2)**(1/2)) 


    abnormal = [0]*len(pre_dif)
    X_abn = x
    y_abn = y
    dif_error = 2
    for i in range(len(pre_dif)):
        if i==len(pre_dif):
            abnormal[i]=0
        elif i==len(pre_dif)-1:
            abnormal[i]=0
        elif i==len(pre_dif)-2:
            abnormal[i]=0 
        elif i==len(pre_dif)-3:
            abnormal[i]=0
        elif pre_dif[i] >= 100 and pre_dif[i+1] >= 100:
            if vis[i:i+2] == [1,1]:# and series[i:i+2] == [1,1]:
                abnormal[i] ='bias1'
                X_abn[i] = 0
                y_abn[i] = 0
        elif pre_dif[i] >= 100 and pre_dif[i+2] >= 100:
            if pre_dif[i+1]<dif_error:
                if vis[i:i+3] == [1,1,1]:# and series[i:i+3] == [1,1,1]:
                    abnormal[i:i+2]=['bias2','bias2']
                    X_abn[i:i+2] = [0,0]
                    y_abn[i:i+2] = [0,0]
        elif pre_dif[i] >= 100 and pre_dif[i+3] >= 100:
            if pre_dif[i+1]<dif_error and pre_dif[i+2]<dif_error:
                if vis[i:i+4] == [1,1,1,1]:# and series[i:i+4] == [1,1,1,1]:
                    abnormal[i:i+3]=['bias3','bias3','bias3']
                    X_abn[i:i+3] = [0,0,0]
                    y_abn[i:i+3] = [0,0,0]
        elif pre_dif[i] >= 100 and pre_dif[i+4] >= 100:
            if pre_dif[i+1]<dif_error and pre_dif[i+2]<dif_error and pre_dif[i+3]<dif_error:
                if vis[i:i+5] == [1,1,1,1,1]:# and series[i:i+5] == [1,1,1,1,1]:
                    abnormal[i:i+4]=['bias4','bias4','bias4','bias4']
                    X_abn[i:i+4] = [0,0,0,0]
                    y_abn[i:i+4] = [0,0,0,0]


    # # II. Poly line check

    x_test = X_abn
    y_test = y_abn


    vis2 = [1] * len(df)
    for i in range(len(df)):
        if x_test[i] ==0 and y_test[i] ==0:
            vis2[i] = 0

    fuc2 = [0]*len(df)
    fuc1 = [0]*len(df)
    fuc0 = [0]*len(df)
    x_ck_bf = [0]*len(df)
    y_ck_bf = [0]*len(df)
    bf_dis = [0]*len(df)
    x_ck_af = [0]*len(df)
    y_ck_af = [0]*len(df)
    af_dis = [0]*len(df)

    for i in range(1,len(df)-7):
        if i==154:
            print(df.iloc[i:i+7])
            print(vis2[i:i+7])
            print('sum(vis2[i:i+7]) : {}'.format(sum(vis2[i:i+7])))
        if sum(vis2[i:i+7])>=2:
            vis_window = np.array(vis2[i:i+7])
            loc = np.where(vis_window==1)
            for k in loc:
                x_ar = np.array(x_test)[i+k]
                y_ar = np.array(y_test)[i+k]
            f1 = np.polyfit(x_ar, y_ar, 2)
            p1 = np.poly1d(f1)
            fuc2[i]=f1[0]
            fuc1[i]=f1[1]
            fuc0[i]=f1[2]
            
            if vis[i+7]==1:
                y_check_af=p1(x_test[i+7])
                x_ck_af[i+7]=x_test[i+7]
                y_ck_af[i+7]=y_check_af
                af_dis[i+7]=abs(y_check_af-y_test[i+7])
            elif vis[i+7]==0:
                x_ck_af[i+7]='NA'
                y_ck_af[i+7]='NA'
            if vis[i-1]==1:
                y_check_bf=p1(x_test[i-1])
                x_ck_bf[i-1]=x_test[i-1]
                y_ck_bf[i-1]=y_check_bf
                bf_dis[i-1]=abs(y_check_bf-y_test[i-1])
            elif vis[i-1]==0:
                x_ck_bf[i-1]='NA'
                y_ck_bf[i-1]='NA'



    # # III. 2nd Denoise

    x_test_2nd = X_abn
    y_test_2nd = y_abn
    abnormal2 = abnormal

    for i in range(len(df)):
        if af_dis[i]>30 and vis2[i]==1:
            if bf_dis[i]>30 and vis2[i]==1:
                x_test_2nd[i]=0
                y_test_2nd[i]=0
                abnormal2[i]='2bias1'
            elif bf_dis[i+1]>30 and vis2[i+1]==1:
                if af_dis[i+1]<30:
                    x_test_2nd[i:i+2]=[0,0]
                    y_test_2nd[i:i+2]=[0,0]
                    abnormal2[i:i+2]=['2bias2','2bias2']
            elif bf_dis[i+2]>30 and vis2[i+1:i+3]==[1,1]:
                if af_dis[i+1]<30 and af_dis[i+2]<30:
                    x_test_2nd[i:i+3]=[0,0,0]
                    y_test_2nd[i:i+3]=[0,0,0]
                    abnormal2[i:i+3]=['2bias3','2bias3','2bias3']
            elif bf_dis[i+3]>30 and vis2[i+1:i+4]==[1,1,1]:
                if af_dis[i+1]<30 and af_dis[i+2]<30 and af_dis[i+3]<30:
                    x_test_2nd[i:i+4]=[0,0,0,0]
                    y_test_2nd[i:i+4]=[0,0,0,0]
                    abnormal2[i:i+4]=['2bias4','2bias4','2bias4','2bias4']
        if i + 4 < len(df):
            if bf_dis[i+4]>30 and vis2[i+1:i+5]==[1,1,1,1]:
                if af_dis[i+1]<30 and af_dis[i+2]<30 and af_dis[i+3]<30 and af_dis[i+4]<30:
                    x_test_2nd[i:i+5]=[0,0,0,0,0]
                    y_test_2nd[i:i+5]=[0,0,0,0,0]
                    abnormal2[i:i+5]=['2bias5','2bias5','2bias5','2bias5','2bias5']
        if i + 5 < len(df):
            if bf_dis[i+5]>30 and vis2[i+1:i+6]==[1,1,1,1,1]:
                if af_dis[i+1]<30 and af_dis[i+2]<30 and af_dis[i+3]<30 and af_dis[i+4]<30 and af_dis[i+5]<30:
                    x_test_2nd[i:i+6]=[0,0,0,0,0,0]
                    y_test_2nd[i:i+6]=[0,0,0,0,0,0]
                    abnormal2[i:i+6]=['2bias6','2bias6','2bias6','2bias6','2bias6','2bias6']
                
        elif af_dis[i]>1000 and vis2[i]==1:
            x_test_2nd[i]=0
            y_test_2nd[i]=0
            abnormal2[i]='2bias1'
        
        elif bf_dis[i]>1000 and vis2[i]==1:
            x_test_2nd[i]=0
            y_test_2nd[i]=0
            abnormal2[i]='2bias1'


    # # IV. Compensate

    vis3 = [1] * len(df)
    for i in range(len(df)):
        if x_test_2nd[i] ==0 and y_test_2nd[i] ==0:
            # print('frame " {}'.format(i))
            vis3[i] = 0

    f2 = fuc2
    f1 = fuc1
    f0 = fuc0
    x_sm = x_test_2nd
    y_sm = y_test_2nd
    comp_ft = [0] * len(df)
    comp_bk = [0] * len(df)
    for i in range(len(vis3)):
        if af_dis[i]!=0 and bf_dis[i]!=0 and af_dis[i]<5 and bf_dis[i]<5:
            if sum(vis3[i-7:i])!=7: # front side compensate
                print(vis3[i-7:i])
                for k in range(5):
                    if vis3[i-7+k:i-4+k]==[1,0,1]:
                        x_ev = (x_sm[i-7+k]+x_sm[i-5+k])/2
                        y_ev = f2[i-7]*x_ev*x_ev + f1[i-7]*x_ev + f0[i-7]
                        x_sm[i-6+k]=x_ev
                        y_sm[i-6+k]=y_ev
                        
                        vis3[i-7+k:i-4+k]=[1,1,1]
                for k in range(4):
                    if vis3[i-7+k:i-3+k]==[1,0,0,1]:
                        for j in range(1,3):
                            x_ev = ((x_sm[i-4+k]-x_sm[i-7+k])/3)*j+x_sm[i-7+k]
                            y_ev = f2[i-7]*x_ev*x_ev + f1[i-7]*x_ev + f0[i-7]
                            x_sm[i-7+k+j]=x_ev
                            y_sm[i-7+k+j]=y_ev
                        
                        vis3[i-7+k:i-3+k]=[1,1,1,1]
                for k in range(3):
                    if vis3[i-7+k:i-2+k]==[1,0,0,0,1]:
                        for j in range(1,4):
                            x_ev = ((x_sm[i-3+k]-x_sm[i-7+k])/4)*j+x_sm[i-7+k]
                            y_ev = f2[i-7]*x_ev*x_ev + f1[i-7]*x_ev + f0[i-7]
                            x_sm[i-7+k+j]=x_ev
                            y_sm[i-7+k+j]=y_ev
                            
                        vis3[i-7+k:i-2+k]=[1,1,1,1,1]
                for k in range(2):
                    if vis3[i-7+k:i-1+k]==[1,0,0,0,0,1]:
                        for j in range(1,5):
                            x_ev = ((x_sm[i-2+k]-x_sm[i-7+k])/5)*j+x_sm[i-7+k]
                            y_ev = f2[i-7]*x_ev*x_ev + f1[i-7]*x_ev + f0[i-7]
                            x_sm[i-7+k+j]=x_ev
                            y_sm[i-7+k+j]=y_ev
                        
                        vis3[i-7+k:i-1+k]=[1,1,1,1,1,1]
                for k in range(1):
                    if vis3[i-7+k:i+k]==[1,0,0,0,0,0,1]:
                        for j in range(1,6):
                            x_ev = ((x_sm[i-1+k]-x_sm[i-7+k])/6)*j+x_sm[i-7+k]
                            y_ev = f2[i-7]*x_ev*x_ev + f1[i-7]*x_ev + f0[i-7]
                            x_sm[i-7+k+j]=x_ev
                            y_sm[i-7+k+j]=y_ev
                        
                        vis3[i-7+k:i+k]=[1,1,1,1,1,1,1]
                        
        
            if sum(vis3[i+1:i+8])!=7: # back side compensate
                print(vis3[i+1:i+8])
                for k in range(5):
                    if vis3[i+1+k:i+4+k]==[1,0,1]:
                        x_ev = (x_sm[i+1+k]+x_sm[i+3+k])/2
                        y_ev = f2[i+1]*x_ev*x_ev+f1[i+1]*x_ev+f0[i+1]
                        x_sm[i+2+k]=x_ev
                        y_sm[i+2+k]=y_ev
                        
                        vis3[i+1+k:i+4+k]=[1,1,1]
                for k in range(4):   
                    if vis3[i+1+k:i+5+k]==[1,0,0,1]:
                        for j in range(1,3):
                            x_ev = ((x_sm[i+4+k]-x_sm[i+1+k])/3)*j+x_sm[i+1+k]
                            y_ev = f2[i+1]*x_ev*x_ev+f1[i+1]*x_ev+f0[i+1]
                            x_sm[i+1+k+j]=x_ev
                            y_sm[i+1+k+j]=y_ev
                        
                        vis3[i+1+k:i+5+k]=[1,1,1,1]
                for k in range(3):   
                    if vis3[i+1+k:i+6+k]==[1,0,0,0,1]:
                        for j in range(1,4):
                            x_ev = ((x_sm[i+5+k]-x_sm[i+1+k])/4)*j+x_sm[i+1+k]
                            y_ev = f2[i+1]*x_ev*x_ev+f1[i+1]*x_ev+f0[i+1]
                            x_sm[i+1+k+j]=x_ev
                            y_sm[i+1+k+j]=y_ev
                        
                        vis3[i+1+k:i+6+k]=[1,1,1,1,1]
                for k in range(2):  
                    if vis3[i+1+k:i+7+k]==[1,0,0,0,0,1]:
                        for j in range(1,5):
                            x_ev = ((x_sm[i+6+k]-x_sm[i+1+k])/5)*j+x_sm[i+1+k]
                            y_ev = f2[i+1]*x_ev*x_ev+f1[i+1]*x_ev+f0[i+1]
                            x_sm[i+1+k+j]=x_ev
                            y_sm[i+1+k+j]=y_ev
                        
                        vis3[i+1+k:i+7+k]=[1,1,1,1,1,1]
        
                for k in range(1):  
                    if vis3[i+1+k:i+8+k]==[1,0,0,0,0,0,1]:
                        for j in range(1,5):
                            x_ev = ((x_sm[i+7+k]-x_sm[i+1+k])/6)*j+x_sm[i+1+k]
                            y_ev = f2[i+1]*x_ev*x_ev+f1[i+1]*x_ev+f0[i+1]
                            x_sm[i+1+k+j]=x_ev
                            y_sm[i+1+k+j]=y_ev
                        
                        vis3[i+1+k:i+8+k]=[1,1,1,1,1,1,1]


    # # V. 2nd Compensate

    vis4 = [1] * len(df)
    for i in range(len(df)):
        if x_sm[i] ==0 and y_sm[i] ==0:
            vis4[i] = 0

    mis1 = []
    mis2 = []
    mis3 = []
    mis4 = []
    mis5 = []
    for i in range(len(vis4)):
        if i == 0:
            mis1.append(0)
        elif vis4[i-1:i+2] == [1,0,1]:
            mis1.append(1)
        elif i == len(vis4):
            mis1.append(0)
        else:
            mis1.append(0)
    for i in range(len(vis4)):
        if i == 0:
            mis2.append(0)
        elif vis4[i-1:i+3] == [1,0,0,1]:
            mis2.append(1)
        elif i == len(vis4)-1:
            mis2.append(0)
        elif i == len(vis4):
            mis2.append(0)
        else:
            mis2.append(0)
    for i in range(len(vis4)):
        if i == 0:
            mis3.append(0)
        elif vis4[i-1:i+4] == [1,0,0,0,1]:
            mis3.append(1)
        elif i == len(vis4)-2:
            mis3.append(0)
        elif i == len(vis4)-1:
            mis3.append(0)
        elif i == len(vis4):
            mis3.append(0)
        else:
            mis3.append(0)
    for i in range(len(vis4)):
        if i == 0:
            mis4.append(0)
        elif vis4[i-1:i+5] == [1,0,0,0,0,1]:
            mis4.append(1)
        elif i == len(vis4)-3:
            mis4.append(0)
        elif i == len(vis4)-2:
            mis4.append(0)
        elif i == len(vis4)-1:
            mis4.append(0)
        elif i == len(vis4):
            mis4.append(0)
        else:
            mis4.append(0)
    for i in range(len(vis4)):
        if i == 0:
            mis5.append(0)
        elif vis4[i-1:i+6] == [1,0,0,0,0,0,1]:
            mis5.append(1)
        elif i == len(vis4)-4:
            mis5.append(0)
        elif i == len(vis4)-3:
            mis5.append(0)
        elif i == len(vis4)-2:
            mis5.append(0)
        elif i == len(vis4)-1:
            mis5.append(0)
        elif i == len(vis4):
            mis5.append(0)
        else:
            mis5.append(0)


    x_sm2 = x_sm
    y_sm2 = y_sm


    mis1_X = []
    mis1_y = []
    for i in range(len(mis1)):
        if i == 0 or i == 1 or i ==2 or i ==len(mis1) or i ==len(mis1)-1 or i ==len(mis1)-2:
            mis1_X.append(x_sm2[i])
            mis1_y.append(y_sm2[i])        
        elif mis1[i] == 0:
            mis1_X.append(x_sm2[i])
            mis1_y.append(y_sm2[i])
        elif mis1[i] ==1:
            miss_point = i

            num_X = [x_sm2[miss_point-1],x_sm2[miss_point+1]]
            num_y = [y_sm2[miss_point-1],y_sm2[miss_point+1]]
            x_mis1 = np.array(num_X) 
            y_mis1 = np.array(num_y)
            

            f1 = np.polyfit(x_mis1, y_mis1, 1)
            p1 = np.poly1d(f1)
            yvals = p1(x)

            insert_X = (x_sm2[miss_point-1]+x_sm2[miss_point+1])/2
            insert_y = np.polyval(f1, insert_X)
            mis1_X.append(insert_X)
            mis1_y.append(insert_y)
        else:
            mis1_X.append(x_sm2[i])
            mis1_y.append(y_sm2[i])

    mis2_X = []
    mis2_y = []
    for i in range(len(mis2)):
        if i == 0 or i == 1 or i ==2 or i ==len(mis2) or i ==len(mis2)-1 or i ==len(mis2)-2:
            mis2_X.append(mis1_X[i])
            mis2_y.append(mis1_y[i])        
        
        elif mis2[i] == 0 and mis2[i-1]==0:
            mis2_X.append(mis1_X[i])
            mis2_y.append(mis1_y[i])    

        elif mis2[i] ==1:
            miss_point = i
            print(len(mis1_X))
            if mis1_X[miss_point-3]!=0 and mis1_X[miss_point-2]!=0 and mis1_X[miss_point-1]!=0 and mis1_X[miss_point+2]!=0 and mis1_X[miss_point+3]!=0 and mis1_X[miss_point+4]!=0:
                num_X = [mis1_X[miss_point-3],mis1_X[miss_point-2],mis1_X[miss_point-1],mis1_X[miss_point+2],mis1_X[miss_point+3],mis1_X[miss_point+4]]
                num_y = [mis1_y[miss_point-3],mis1_y[miss_point-2],mis1_y[miss_point-1],mis1_y[miss_point+2],mis1_y[miss_point+3],mis1_y[miss_point+4]]
                x_mis2 = np.array(num_X) 
                y_mis2 = np.array(num_y)

                f1 = np.polyfit(x_mis2, y_mis2, 2)
                p1 = np.poly1d(f1)
                yvals = p1(x)

                for j in range(1,3):
                    insert_X = ((mis1_X[miss_point+2]-mis1_X[miss_point-1])/3)*j+mis1_X[miss_point-1]
                    insert_y = np.polyval(f1, insert_X)
                    mis2_X.append(insert_X)
                    mis2_y.append(insert_y)
            else:
                mis2_X.append(mis1_X[i])
                mis2_y.append(mis1_y[i])
                mis2_X.append(mis1_X[i+1])
                mis2_y.append(mis1_y[i+1])


    mis3_X = []
    mis3_y = []
    for i in range(len(mis3)):
        if i == 0 or i == 1 or i ==2 or i ==len(mis3) or i ==len(mis3)-1 or i ==len(mis3)-2:
            mis3_X.append(mis2_X[i])
            mis3_y.append(mis2_y[i])        
     
        elif mis3[i-2:i+1] == [0,0,0]:
            mis3_X.append(mis2_X[i])
            mis3_y.append(mis2_y[i])    

        elif mis3[i] ==1:
            miss_point = i
            if mis2_X[miss_point-3]!=0 and mis2_X[miss_point-2]!=0 and mis2_X[miss_point-1]!=0 and mis2_X[miss_point+3]!=0 and mis2_X[miss_point+4]!=0 and mis1_X[miss_point+5]!=0:
                num_X = [mis2_X[miss_point-3],mis2_X[miss_point-2],mis2_X[miss_point-1],mis2_X[miss_point+3],mis2_X[miss_point+4],mis2_X[miss_point+5]]
                num_y = [mis2_y[miss_point-3],mis2_y[miss_point-2],mis2_y[miss_point-1],mis2_y[miss_point+3],mis2_y[miss_point+4],mis2_y[miss_point+5]]
                x_mis3 = np.array(num_X) 
                y_mis3 = np.array(num_y)

                f1 = np.polyfit(x_mis3, y_mis3, 2)
                p1 = np.poly1d(f1)
                yvals = p1(x)

                for j in range(1,4):
                    insert_X = ((mis2_X[miss_point+3]-mis2_X[miss_point-1])/4)*j+mis2_X[miss_point-1]
                    insert_y = np.polyval(f1, insert_X)
                    mis3_X.append(insert_X)
                    mis3_y.append(insert_y)
            else:
                mis3_X.append(mis2_X[i])
                mis3_y.append(mis2_y[i])
                mis3_X.append(mis2_X[i+1])
                mis3_y.append(mis2_y[i+1])
                mis3_X.append(mis2_X[i+2])
                mis3_y.append(mis2_y[i+2])
            


    mis4_X = []
    mis4_y = []
    for i in range(len(mis4)):
        if i == 0 or i == 1 or i ==2 or i ==len(mis4) or i ==len(mis4)-1 or i ==len(mis4)-2:
            mis4_X.append(mis3_X[i])
            mis4_y.append(mis3_y[i])           
        elif mis4[i-3:i+1] == [0,0,0,0]:
            mis4_X.append(mis3_X[i])
            mis4_y.append(mis3_y[i])    

        elif mis4[i] ==1:
            miss_point = i
            if mis3_X[miss_point-3]!=0 and mis3_X[miss_point-2]!=0 and mis3_X[miss_point-1]!=0 and mis3_X[miss_point+4]!=0 and mis3_X[miss_point+5]!=0 and mis3_X[miss_point+6]!=0:
                num_X = [mis3_X[miss_point-3],mis3_X[miss_point-2],mis3_X[miss_point-1],mis3_X[miss_point+4],mis3_X[miss_point+5],mis3_X[miss_point+6]]
                num_y = [mis3_y[miss_point-3],mis3_y[miss_point-2],mis3_y[miss_point-1],mis3_y[miss_point+4],mis3_y[miss_point+5],mis3_y[miss_point+6]]
                x_mis4 = np.array(num_X) 
                y_mis4 = np.array(num_y)

                f1 = np.polyfit(x_mis4, y_mis4, 2)
                p1 = np.poly1d(f1)
                yvals = p1(x)


                for j in range(1,5):
                    insert_X = ((mis3_X[miss_point+4]-mis3_X[miss_point-1])/5)*j+mis3_X[miss_point-1]
                    insert_y = np.polyval(f1, insert_X)
                    mis4_X.append(insert_X)
                    mis4_y.append(insert_y)
            else:
                mis4_X.append(mis3_X[i])
                mis4_y.append(mis3_y[i])
                mis4_X.append(mis3_X[i+1])
                mis4_y.append(mis3_y[i+1])
                mis4_X.append(mis3_X[i+2])
                mis4_y.append(mis3_y[i+2])
                mis4_X.append(mis3_X[i+3])
                mis4_y.append(mis3_y[i+3])



    mis5_X = []
    mis5_y = []
    for i in range(len(mis5)):
        if i == 0 or i == 1 or i ==2 or i==3 or i ==len(mis5) or i ==len(mis5)-1 or i ==len(mis5)-2:
            mis5_X.append(mis4_X[i])
            mis5_y.append(mis4_y[i])           
        elif mis5[i-4:i+1] == [0,0,0,0,0]:
            mis5_X.append(mis4_X[i])
            mis5_y.append(mis4_y[i])    

        elif mis5[i] ==1:
            miss_point = i
            if mis4_X[miss_point-3]!=0 and mis4_X[miss_point-2]!=0 and mis4_X[miss_point-1]!=0 and mis4_X[miss_point+5]!=0 and mis4_X[miss_point+6]!=0 and mis4_X[miss_point+7]!=0:
                num_X = [mis4_X[miss_point-3],mis4_X[miss_point-2],mis4_X[miss_point-1],mis4_X[miss_point+5],mis4_X[miss_point+6],mis4_X[miss_point+7]]
                num_y = [mis4_y[miss_point-3],mis4_y[miss_point-2],mis4_y[miss_point-1],mis4_y[miss_point+5],mis4_y[miss_point+6],mis4_y[miss_point+7]]
                x_mis5 = np.array(num_X) 
                y_mis5 = np.array(num_y)

                f1 = np.polyfit(x_mis5, y_mis5, 2)
                p1 = np.poly1d(f1)
                yvals = p1(x)


                for j in range(1,6):
                    insert_X = ((mis4_X[miss_point+5]-mis4_X[miss_point-1])/6)*j+mis4_X[miss_point-1]
                    insert_y = np.polyval(f1, insert_X)
                    mis5_X.append(insert_X)
                    mis5_y.append(insert_y)
            else:
                mis5_X.append(mis4_X[i])
                mis5_y.append(mis4_y[i])
                mis5_X.append(mis4_X[i+1])
                mis5_y.append(mis4_y[i+1])
                mis5_X.append(mis4_X[i+2])
                mis5_y.append(mis4_y[i+2])
                mis5_X.append(mis4_X[i+3])
                mis5_y.append(mis4_y[i+3])
                mis5_X.append(mis4_X[i+4])
                mis5_y.append(mis4_y[i+4])

    df['X'] = mis5_X
    df['Y'] = mis5_y

    df.to_csv(os.path.join(rootdir, 'pred_result', os.path.basename(path)),index=False)

rootdir = os.getcwd()

for i in os.listdir(os.path.join(rootdir, 'pred_result')):
    if i[-4:]=='.csv':
        smooth(os.path.join(rootdir, 'pred_result', i))