traj-dist-rs 1.0.0-rc.5

High-performance trajectory distance & similarity measures in Rust with Python bindings
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194

"""
This file is copied from TrajCl repo: https://github.com/changyanchuan/TrajCL/blob/master/utils/edwp.py
"""

# EDwP python implementation
# Paper: Indexing and Matching Trajectories under Inconsistent Sampling Rates

# This module is implemented by following the original author provided java implementation [1].
# I have tried two different python implementations [2,3] on github.
# Neither has the same outputs as the authorized java implementation with the same inputs.
# Thus, I implemented this. -- yc
# [1]. https://www.cse.iitd.ac.in/~sayan/publications.html
# [2]. https://github.com/Rena7ssance/search-trajectory-kBCT/blob/master/distance/edwp.py
# [3]. https://github.com/skairunner/TrajectoryInspector/blob/master/EDwP/edwp.py


import math
import sys
from itertools import tee

import numpy as np

F64MAX = sys.float_info.max


class Matrix:
    def __init__(self, nrow, ncol):
        self.nrow = nrow
        self.ncol = ncol
        self.value = np.zeros([nrow, ncol])
        self.delta = np.zeros([nrow, ncol])
        self.col_edits = np.zeros([nrow, ncol, 2])
        self.row_edits = np.zeros([nrow, ncol, 2])

        self.value[0, 1:] = F64MAX
        self.value[1:, 0] = F64MAX

    def add(self, i, j, val, col_edit, row_edit):
        self.value[i, j] = val
        self.col_edits[i, j] = col_edit
        self.row_edits[i, j] = row_edit

    def score(self):
        return self.value[-1, -1]


def _pairwise(iterable):
    a, b = tee(iterable)
    next(b, None)
    return zip(a, b)


def _distance(p, q):
    # p, q: list
    return math.sqrt((p[0] - q[0]) ** 2 + (p[1] - q[1]) ** 2)


def _line_map(p1, p2, p):
    l2 = (p1[0] - p2[0]) ** 2 + (p1[1] - p2[1]) ** 2
    if l2 == 0:
        return p
    t = ((p[0] - p1[0]) * (p2[0] - p1[0]) + (p[1] - p1[1]) * (p2[1] - p1[1])) / l2
    if t < 0:
        return p1
    elif t > 1:
        return p2
    else:
        return [p1[0] + t * (p2[0] - p1[0]), p1[1] + t * (p2[1] - p1[1])]


def edwp(t1, t2):
    # t1, t2 : list of 2D list = [[x,y], [], ...]

    t1_len = len(t1)
    t2_len = len(t2)

    t1_edge_length = [_distance(p1, p2) for p1, p2 in _pairwise(t1)]
    t2_edge_length = [_distance(p1, p2) for p1, p2 in _pairwise(t2)]

    matrix = Matrix(t1_len, t2_len)

    total_length = sum(t1_edge_length) + sum(t2_edge_length)

    for i in range(1, matrix.nrow):
        for j in range(1, matrix.ncol):
            row_delta = col_delta = F64MAX
            row_converge1 = row_converge2 = col_converge1 = col_converge2 = F64MAX
            row_spatial_score = col_spatial_score = F64MAX

            t1_insert = t2_insert = None
            t1_edit = t2_edit = None

            if i > 1:
                t1_edit = matrix.row_edits[i - 1, j]
                t2_edit = matrix.col_edits[i - 1, j]
                prev_point_edge = _distance(t1_edit, t1[i - 1])
                t2_insert = _line_map(t2_edit, t2[j], t1[i - 1])
                row_edit_distance = _distance(t2_insert, t1[i - 1])
                row_edit_edge = _distance(t2_edit, t2_insert)

                row_converge1 = (row_edit_edge + prev_point_edge) / total_length
                row_converge2 = (
                    _distance(t2[j], t2_insert) + t1_edge_length[i - 1]
                ) / total_length

                row_delta = (
                    matrix.value[i - 1, j]
                    - matrix.delta[i - 1, j]
                    + (row_edit_distance + _distance(t1_edit, t2_edit)) * row_converge1
                )
                row_spatial_score = (
                    row_delta
                    + (row_edit_distance + _distance(t2[j], t1[i])) * row_converge2
                )

            if j > 1:
                t1_edit = matrix.row_edits[i, j - 1]
                t2_edit = matrix.col_edits[i, j - 1]
                if t1_edit is None:
                    break
                prev_point_edge = _distance(t2_edit, t2[j - 1])
                t1_insert = _line_map(t1_edit, t1[i], t2[j - 1])
                col_edit_distance = _distance(t1_insert, t2[j - 1])
                col_edit_edge = _distance(t1_edit, t1_insert)

                col_converge1 = (col_edit_edge + prev_point_edge) / total_length
                col_converge2 = (
                    _distance(t1[i], t1_insert) + t2_edge_length[j - 1]
                ) / total_length

                col_delta = (
                    matrix.value[i, j - 1]
                    - matrix.delta[i, j - 1]
                    + (col_edit_distance + _distance(t1_edit, t2_edit)) * col_converge1
                )
                col_spatial_score = (
                    col_delta
                    + (col_edit_distance + _distance(t1[i], t2[j])) * col_converge2
                )

            diag_coverage = (
                t1_edge_length[i - 1] + t2_edge_length[j - 1]
            ) / total_length
            sub_score = (
                _distance(t2[j], t1[i]) + _distance(t2[j - 1], t1[i - 1])
            ) * diag_coverage
            diag_score = matrix.value[i - 1, j - 1] + sub_score

            if diag_score <= col_spatial_score and diag_score <= row_spatial_score:
                matrix.add(i, j, diag_score, t2[j - 1], t1[i - 1])
                matrix.delta[i, j] = diag_score - matrix.value[i - 1, j - 1]
            elif col_spatial_score < row_spatial_score or (
                col_spatial_score == row_spatial_score and t2_len > t1_len
            ):
                matrix.add(i, j, col_spatial_score, t2[j - 1], t1_insert)
                matrix.delta[i, j] = col_spatial_score - col_delta
            else:
                matrix.add(i, j, row_spatial_score, t2_insert, t1[i - 1])
                matrix.delta[i, j] = row_spatial_score - row_delta

    return matrix.score()


if __name__ == "__main__":
    import time

    _time = time.time()
    t1 = [
        (48038, 420644),
        (48057, 420665),
        (48140, 420818),
        (48196, 420800),
        (48201, 420798),
        (48210, 420794),
        (48328, 420790),
        (48335, 420787),
        (48336, 420788),
    ]
    t2 = [
        (47376, 420478),
        (48126, 420826),
        (48140, 420816),
        (48191, 420756),
        (48215, 420747),
        (48283, 420761),
        (48351, 420782),
        (48351, 420787),
        (48342, 420788),
    ]
    _time = time.time()

    for _ in range(10000):
        rtn = edwp(t1, t2)
    print(time.time() - _time)