Final testing
The following is a real-life test that was created to recognise traffic light signals, the camera was tested during different adverse conditions using actual traffic lights. The camera was held away from the traffic light (attached to a car mirror), while the car was moving further away from the traffic light to see how from how far the camera can detect the green lights in the extreme weather conditions. The figures below the table are the output on the web app while testing.
Table 1: Device testing results
Figure 1 – Output on the web app at night-time
Figure 2 – Output on the web app at daylight
Figure 3 – Output on the web app (Dust at daylight)
Figure 4 - Output on the web app (Brightness level 25% at night-time)
Figure 5 – Output on the web app (Rain at daylight)
The following table shows the tests done with different traffic lights, and with different backgrounds for accuracy.
Table 2: Testing result with actual traffic lights
The following table shows that the device has been tested with coloured circles made in a word document.
Table 3: Testing result with geometric shape (Circle)
To conclude, the trucks used at loading/unloading facilities are roughly 19m maximum in length within the UK. Table 1 shows that the device was able to recognise the green light at a distance of 19m, as notifications were sent, except when there was dust on the camera lens. Therefore, the lens needs to be clean to recognise the green light. Also, as the camera did not have a case when this testing was carried out, so for camera protection; it was tested from inside the car by facing it to the side mirror from inside the car. Snow and fog are extreme conditions that will be tested later.
Table 2 provides information of how many times the camera was tested with an actual traffic light and how many times it was able to detect the green signals. As shown at night-time, it was a successful detection, it always detects the signal accurately. However, at daytime it was harder, it only managed to detect 2 out of 5. The 2 that were detected did not have a noisy background, the other 3 they had trees in the background which made it hard. While the camera was detecting, notifications were being sent even though the light was still red. Background noises such as trees would not be a problem at most warehouses, as the traffic light are placed on a plain wall. This testing shows that the device works better at night-time as it has less visual noise. The device was also tested with audio noise at both night-time and daytime. That did not affect the detection, and the device was working fine.
Due to COVID-19 the testing with limited using an actual traffic light. So, using different coloured and sized circles on word document was used for testing. The camera was able to detect the green light accurately in both night and daytime as shown in table 3. There were no background noises, therefore the detection was successful.
Comments