*2.2. Vehicles*

Seventeen (17) cars in total were included in the study. Fourteen of them were from different manufacturers while the remaining three were from the same manufacturer. Most of those cars are new vehicles (12 of them were produced in 2020), while the other five ranged from 2014 to 2019. Each vehicle has TSRS which is based on a camera located at the front windshield inside the vehicle. The vehicles were, for the sole purpose of the research, lent to us by their official distributors for their respective manufacturing brand in Croatia. It is worth noting that the technical aspects of the system were not provided to us. Since the aim of the study is to detect the differences between existing TSRS and not to determine which car brand has the "best" TSRS, brand names are in code.

#### *2.3. Scenarios and Test Procedure*

In order to test the accuracy of TSRS, nine scenarios were created. The first one was a control scenario in which signs did not have any graphical changes. In the second scenario, the red outline of the signs was covered with black paper. In the third and fourth scenarios, minor changes on symbols were made. For example, a piece of black paper was placed so that 60 km/h speed limit looks similar to 80 km/h or black lines were placed on the prohibition of overtaking signs. In the fifth, sixth and seventh scenarios, a half of the signs' face was covered with black paper. The difference between the scenarios was in the orientation of the coverage. In the eighth scenario, the last digit (0) on the speed limit sign was covered with white paper. On the two prohibition of overtaking signs, the symbols of cars were covered with white paper while on the third prohibition of overtaking sign, the symbols were mirrored. In the last scenario, symbols were covered with graffiti. The visual presentation of signs in each scenario is shown in Figure 2.

**Figure 2.** *Cont.*

**Figure 2.** Visual presentation of signs used in each scenario.

Tests were conducted during daytime between 9:00 a.m. and 3:00 p.m. in the period of three days. During all three days, weather conditions were almost the same, i.e., normal (sunny) weather. Before the start, we held a "practice" run with each vehicle in order to check if the TSRS were functioning properly. During the test, each vehicle passed through the test track once per each scenario. Alongside the driver, the researcher was present in the vehicle during tests, controlling and recording if the signs were or were not displayed to the driver and, if yes, what was displayed. Only one vehicle was permitted on the testing track during each run and other traffic was not present. Because of the road configuration (campus site), the driving speed was between 40 and 60 km/h.

#### *2.4. Data Analysis*

For the first part of the analysis, we grouped the data into three categories: (1) correctly recognized signs, (2) unrecognized signs and (3) wrongly recognized signs. For each category, a percentage distribution of the results was calculated and compared. For the second part, the results from groups (2) and (3) were joined into one group, unrecognized or wrongly recognized signs, since they all in all presented the error of TSRS. On top of that, scenarios were grouped based on the level of modification they were submitted to: control (scenario 1), minor changes (scenarios 2 + 3 + 4), medium changes (scenarios 8+9) and major changes (scenarios 5 + 6 + 7) in order to analyze the differences between the control scenario (1) and each level of graphical change. Mean values for each vehicle and group were computed and repeated measures ANOVA with Bonferroni adjustment (alpha level 0.05) were used to test the differences between the control condition and each group.

In the second part of the analysis, we tested the differences between vehicles per scenario. Since the values in the data set for this analysis were dichotomous, Cochran's Q test was used (alpha level 0.05). As one of the assumptions of Cochran's Q test is that a sample has two mutually exclusive categories, the analysis was conducted on the correctly recognized signs categories and not on unrecognized or wrongly recognized signs.
