Fig. 1. Diagram of lidar scanning angle
Fig. 2. Overlapping vehicles and close pedestrians
Fig. 3. Point cloud of road surface
Fig. 4. Ground removal
Fig. 5. Location between pedestrians and vehicles. (a) Ray diagram; (b) top view
Fig. 6. Diagram of angle
Fig. 7. Diagram of lidar segmentation
Fig. 8. Experimental platform. (a) Electric control car; (b) electric control equipment
Fig. 9. Principle of Bounding Box
Fig. 10. Ground removal. (a) Least square method; (b) RANSAC algorithm
Fig. 11. Local cluster comparison. (a) Point cloud map; (b) traditional Euclidean clustering algorithm; (c) improved Euclidean clustering algorithm
Fig. 12. Global cluster comparison. (a) Original point cloud; (b) traditional Euclidean clustering algorithm; (c) improved Euclidean clustering algorithm
Algorithm | Positive detection /times | False detection /times | Missed detection /times |
---|
Least square method | 3 | 3 | 1 | RANSAC algorithm | 6 | 1 | 0 |
|
Table 1. Comparison of ground segmentation
Algorithm | Positive detection /times | False detection /times | Positive detection rate /% |
---|
Traditional Euclidean clustering | 103 | 62 | 62.42 | Improved Euclidean clustering | 147 | 18 | 89.09 |
|
Table 2. Vehicles parked on roadside(165 vehicles)
Algorithm | Positivedetection /times | Falsedetection /times | Misseddetection /times | Positive detectionrate /% |
---|
Traditional Euclidean clustering | 103 | 62 | 19 | 66.37 | Improved Euclidean clustering | 147 | 18 | 22 | 76.10 |
|
Table 3. Pedestrian (113 pedestrians)
Algorithm | Positive detection /times | False detection /times | Positive detection rate /% |
---|
Traditional Euclidean clustering | 26 | 7 | 78.78 | Improved Euclidean clustering | 30 | 3 | 90.90 |
|
Table 4. Mobile vehicles including non-motorized vehicles (33 vehicles)