TIDUE95A April 2018 – April 2020
3.2.2 Test Results
The VOD demo GUI is instrumented with a counting widget that when started, counts frames received from the deviceand counts positive (occupied) detections in each zone. No averaging or smoothing is performed; it is a simple frame count.Each test was allowed to run for at least 30 seconds (approximately 200 frames at 6.25 fps), and the results are recorded in Table 3. Since a large movement such as entering or exiting the vehicle causes a large spike in all zones, occupants were allowed to sit in the zones for several frames before the counting starts.
Table 3. Scenario Config file: od_demo_68xx_4zone.cfg
| TEST (1) (1)(1) | TOTAL FRAMES | Front Left | Front Right | Back Left | Back Right | ERROR (%) |
|---|---|---|---|---|---|---|
| All zones empty | 227 | 0 | 0 | 0 | 0 | 0 |
| Static object | 216 | 0 | 0 | 0 | 0 | 0 |
| Front Left occupied | 243 | 243 | 11 | 21 | 0 | 8.6% |
| Front right occupied | 228 | 0 | 228 | 0 | 0 | 0 |
| Back left occupied | 217 | 0 | 0 | 217 | 0 | 0 |
| Back right occupied | 232 | 0 | 0 | 0 | 232 | 0 |
| All zones occupied | 249 | 249 | 249 | 249 | 249 | 0 |
Test Notes: Detection software that functions well in the lab environment will encounter two obstacles when moving to a vehicle. Both of these difficulties are caused by the confined environment of the vehicle interior, and both cause extra unwanted energy in the heatmap. First, azimuthal leakage of the MVDR algorithm is increased by reflections from the sides of the vehicle. Second, multipath from all interior surfaces creates extra energy that typically co-locates with second row seats. Extra measures need to be taken to reject this extra energy, and are currently under study. Also, both of these conditions are amplified when any large motion occurs, such as when an occupant enters or exits the vehicle, changes seats, bounces in the seat, and so on. This causes widespread energy spikes that must be ignored.
Other tests were also performed with the mmWave sensor mounted in the vehicle. The first test was a child seated in a carseat in one of the zones. Figure 10 shows the child in the carseat, and a capture from the VOD demo GUI during testing.
Another test illustrates an intruder approaching the vehicle. Figure 11 shows the intruder and the corresponding heatmap image. A positive (occupied) detection was not gathered, because the intruder was still outside the defined zones.
