Introduction
Related work
Object detection
Speed detection
Methodology
Detection of vehicles
Index | Class | Color |
---|---|---|
0 | Car | Yellow |
1 | Mini_bus | Blue |
2 | Bus | Brown |
3 | Truck | Red |
4 | Tram | Pink |
5 | Trolleybus | Green |
Class | Number of objects | Relation to the total number % |
---|---|---|
Car | 3,518,370 | 81.8 |
Mini_bus | 228,810 | 5.3 |
Bus | 163,430 | 3.8 |
Truck | 184,520 | 4.2 |
Tram | 91,540 | 2.1 |
Trolleybus | 112,690 | 2.6 |
Training of YOLOv3
-
coordinates and positions of the predicted bounding boxes, which should contain the objects;
-
the probability that each bounding box contains an object;
-
the probability that each object within its bounding box belongs to a certain class.
The results of training the neural network
Vehicle tracking
Elimination of the camera distortion
Calculation of the distance
Experimental results and discussions
Counting the vehicles
Data | Time | Intersections | Class | |||||
---|---|---|---|---|---|---|---|---|
Car | Mini Bus | Bus | Truk | Tram | Trolley-bus | |||
18.03.20 | 7:00–9:00 | Komarova-Salyutnaya st. | 5835/5896 | 315/313 | 22/25 | 21/20 | 0/0 | 0/0 |
18.03.20 | 17:00–19:00 | Komarova-Salyutnaya st. | 6912/6767 | 218/227 | 18/19 | 11/11 | 0/0 | 0/0 |
17.03.20 | 7:00–9:00 | Pobedy-Molodogvardeitsev st. | 6587/6677 | 382/364 | 19/21 | 23/23 | 49/48 | 0/0 |
17.03.20 | 17:00–18:00 | Chicherina-Pobedy st. | 6210/6178 | 259/246 | 48/47 | 13/12 | 50/52 | 9/9 |
18.03.20 | 7:00–9:00 | Pobedy-Krasnoznamennaya st. | 4501/4554 | 228/239 | 41/40 | 57/54 | 73/73 | 0/0 |
18.03.20 | 17:00–19:00 | Komsomolskiy-Sverdlovskiy st. | 9865/9501 | 601/602 | 43/41 | 6/6 | 0/0 | 109/102 |
Error (%) | Car | Mini bus | Bus | Truck | Tram | Trolley-bus |
---|---|---|---|---|---|---|
Mean | 1.6 | 3.2 | 13.6 | 2.9 | 2.0 | 3.2 |
Max | 3.6 | 5.0 | 6.4 | 7.6 | 4.0 | 6.4 |
Average vehicle speed
Vehicles | Direction 1 yellow | Direction 2 red | Direction 3 blue | Direction 4 green | ||||
---|---|---|---|---|---|---|---|---|
Real | Detection | Real | Detection | Real | Detection | Real | Detection | |
1 | 48.7 | 48.1 | 59.1 | 60.2 | 34.4 | 35.7 | 31.9 | 31.7 |
2 | 31.8 | 32.3 | 43.6 | 44.2 | 34.2 | 34.0 | 34.7 | 35.1 |
3 | 53.4 | 53.1 | 44.8 | 44.4 | 29.2 | 28.5 | 27.9 | 27.6 |
4 | 40.3 | 41.0 | 40.4 | 40.6 | 40.2 | 40.4 | 40.4 | 40.0 |
5 | 63.7 | 64.3 | 55.3 | 55.3 | 33.1 | 32.9 | 30.0 | 31.5 |
6 | 53.4 | 53.5 | 38.5 | 39.7 | 35.3 | 36.4 | 36.4 | 37.3 |
Max | 0.7 | 1.2 | 1.3 | 1.5 | ||||
Mean | 0.46 | 0.58 | 0.62 | 0.62 |
Time complexity
Software solution
-
frames reading (Process 1);
-
detection and classification of vehicles from the current frame (Process 2);
-
vehicle tracking and counting in all directions of the road junction (Process 3);
-
calculation of the latitude and longitude of the vehicle location (Process 4);
-
calculation of vehicle speeds (Process 5);
-
calculation of metrics related to the amount of harmful substances emitted by each vehicle (Process 6).