Question about reconstructing the ‘spdist’ by using ‘offset’ and ‘road length’

[Dreamzz5]

idx=2 #print opath
print(gdf.iloc[idx]['opath'])
array([14695, 14715, 14675, 14675, 14675,  8718]))

print(gdf.iloc[idx]['cpath']) #print cpath
array([14695,  7036, 14715, 14717,  8859, 14675,  8718])

print(gdf.iloc[idx]['tpath'].split('|')) #print tpath
['14695,7036,14715', '14715,14717,8859,14675', '14675', '14675', '14675,8718']

print(gdf.iloc[idx]['offset'])#print offset
array([  0.        ,  16.53772351, 154.8306739 , 154.8306739 ,154.8306739 ,  63.57969753])

print(gdf.iloc[idx]['spdist'])#print spdist
array([256.097439  , 462.1053789 ,   0.        ,   0.        ,63.57969753])

print(gdf.iloc[idx]['road_length'])#print road_length
[124.762, 106.248, 50.242, 142.199, 107.313, 143.378, 199.241]

So when I want to use the offset and road length to reconstruct the spdist. For example, tpath 1 '14695,7036,14715'

road1 length - offset1+ road2 length + offset2=
124.762-0+16.53772351+106.248=247.54772351 != 256.097439

[cyang-kth]

It seems the road length you obtained is not correct. How did you get it?

[Dreamzz5]

The original Shapefile was downloaded from OSM for map matching.

[cyang-kth]

From your example, the edge 14675 has a length of 143.378 but its offset is 154.8306739, which is longer than the length. I think the length field you are reporting here is not correct.

[Dreamzz5]

OK, I will check it now.

[Dreamzz5]

Yes, it weirds. I plot the road network and trajectory together and demonstrate that the road id is correct.

fig,ax=plt.subplots(dpi=100)
road_network[road_network.id.isin(eval(gdf.iloc[idx]['cpath']))].plot(ax=ax,edgecolor='b',label='road',alpha=0.8)
gpd.GeoDataFrame(gdf.iloc[idx:idx+1]).set_geometry('pgeom').plot(ax=ax,edgecolor='r',label='trajectory',alpha=0.8)
ax.legend()
ax.axis('off')

[Dreamzz5]

The road network is

[cyang-kth]

If you directly calculate the length from the geometry, what is your result?

[Dreamzz5]

What means about ‘directly calculate’？

[cyang-kth]

The length is calculated from the geometry of the linestring in fmm rather than using the cost information. You could also export the length field of the matching result to check the length inside the program. It may be different from the current one you are using, which comes from the cost field of the shapefile.

[Dreamzz5]

print(gdf.iloc[idx].pgeom.length*111.11*1000) # meter
760.639211613965
print(gdf.iloc[idx]['spdist'].sum()*111.11*1000) # meter
781.78251543

[cyang-kth]

I mean the length of each edge, not the spdist, or pgeom. Pgeom is just a line connecting each matched point, its length is not the same as the spdist.

[Dreamzz5]

OK, I will check it now.

[Dreamzz5]

Yes, you are right. When I transfer the cost to 'linestring', the 'spdist' is matched.

print(road_network[road_network.id.isin(eval(gdf.iloc[idx]['cpath']))].cost.values)
[106.248, 199.241, 107.313, 143.378, 124.762,  50.242, 142.199]

print(road_network[road_network.id.isin(eval(gdf.iloc[idx]['cpath']))].geometry.length.values*111.11*1000)
[109.70259929, 217.65652669, 114.3454652 , 154.83075841,129.85755404,  58.28126324, 151.18537477]

print(road_length2[14695],road_length2[7036],road_length2[14715])
(129.8575540363128, 109.7025992889302, 58.28126324202719)

129.8575540363128-0+109.7025992889302+16.53772351=256.09787683524297， which is almost equal to ‘spdist’(256.097439).

thanks a lot.