The methodological step followed in the present research is shown in Fig. (1), and the variables considered for analysis are shown in Table 1 . The primary step followed in this study is conducting a videographic survey at locations of different land use types. The second step involves data extraction from the recorded video and removing outliers from the extracted data. The third step involved in this study is conducting statistical analysis to identify the classification of data. In the fourth step, the crossing speed of the pedestrian is converted into crossing time High/Low. This is achieved at each location by assuming a crossing speed of 1.2 m/s for each site, and with reference to this base value, crossing time High/Low is determined. The final step involved predicting the categorical choices crossing time High or Low using the binary logit model. Further, this study helps in identifying the variables which are having a significant influence on the crossing time of pedestrians.

Fig. (1). Study methodology.

The district of Kozhikode is a fast-developing city with a population of 30,86,293 in 2011, with females 52.34% and males 47.66% [11 ]. Further, the city has a total area of 2,206 sq.km [ 40 ] and constitutes 38.25% of urban areas [ 39 ], and the total number of registered transport motor vehicles as on March 2018 is 69,518, and registered non-transport vehicles is 3,42,515 [41]. Three midblock sites selected from the city with different land use types are considered for the study. The selected midblock location is of mixed land use type located at the center of Kozhikode city. A schematic diagram of a typical midblock location with basic pedestrian behaviour is shown in Fig. (2).

The variables considered for the present research are listed in Table 1, description of each variable is shown below:

• Driver Yielding – It represents the driver's charac-teristics in giving way to pedestrians at the zebra crossing location. When a pedestrian tries to cross the road, if the driver on the road reduces the speed and allows the pedestrian to cross indicates driver-yielding behaviour-yes. If the driver does not yield to the pedestrian, it shows the non-yielding behaviour of the driver.
• Stage – This behaviour represents how many inter-mediate points the pedestrians stop during the road crossing activity. The number of intermediate stops represents the number of stages [10], which is shown with an example in Fig. (2).
• Age – In the present study, the data is collected from video data, so the age group of the pedestrian was identified by visual observation, i.e., children (0 – 15 years), young (16 – 25 years), adults (26 – 60 years) and senior (more than 60). The classification of age group is represented as children, young, adults and seniors based on the observation from the video.
• Gender – This represents the classification of a male and female group of pedestrians involved in road crossing activity.
• Crossing Time – The time taken by the pedestrian to cross the road is called crossing time, and it is interpreted with the base crossing speed of 1.2m/s. If the crossing time is higher than the base reference time, it is crossing time High.
• Crossing Behaviour – The behaviour of pedestrians that represents the pattern of road crossing is said by path changing and straight crossing behaviour. Path-changing [10, 42] behaviour occurs if the pedestrian crosses the road in an irregular pattern and reaches the other end of the road. Fig. (2) shows an example path-changing behaviour of a pedestrian. Straight crossing [10] occurs if the pedestrian crosses the road perpendicular without making any deviation at the crosswalk.

The survey was conducted in the year 2019 on all three selected locations simultaneously for a period of three days on a normal working day. The data was collected using a video camera placed at a suitable vantage point in all locations. The video is recorded during morning and evening peak hours for a duration of 6 hours from the selected midblock locations with a stretch of 15m on either side of the road from the zebra marking. The peak hour in the morning represents 8:30 am to 11:30 am, and the afternoon peak hour represents 3:30 pm to 6:30 pm. The characteristics of selected sites from Kozhikode are shown in Table 2. The selected locations are characteristics of different land-use types; location 1 is near the bus terminal, which is used mostly by commuters travelling inside and outside the city. Also, location 1 provides access to shopping needs. Since location 1 is near the bus terminal, most bus users walk either to the bus stop on the other side or move inside the bus terminal from the bus stop. Location 2 is of mixed land-use type which is used for shopping and health-related activity by the locals. Location 3 is near the hospital zone, which has regular pedestrian and vehicle movement on all working and non-working days. The locations of all the sites have footpaths and zebra marking for pedestrians.

The recorded video data were analysed through a custom-made tool developed using vb.net, which extract the time of crossing pedestrian and other variables in excel format. The tool can be played forward and backward with reduced player speed to get accurate pedestrian crossing time.

The road crossing speed of pedestrians is compared with crossing patterns like straight and path-changing behaviour. The result identifies that the mean crossing speed of path-changing pedestrians is 1.26m/s and straight-moving pedestrians' mean crossing speed is 1.56m/s, where the mean speed of path-changing pedestrians is lesser when compared to straight-moving pedestrians. This shows that the crossing speed of path-changing pedestrians is reducing due to increased crossing distance. This indicates that their road contact time is higher due to their increased crossing distance when compared to the straight crossing distance, which reduces the safety of path-changing pedestrians.

The t-test is used to compare the crossing speed of the pedestrian with respect to the crossing behaviour. The Independent sample t-test result shows the p-value is <0.05, which rejects the null hypothesis and concludes that there is a significant difference between the crossing speed of Path change and Straight pedestrian at 5% of significance, t (8718) = -30.479, P=0.000.

The two-way ANOVA test is conducted to compare the mean crossing speed of pedestrians with respect to gender and crossing behaviour (path change and straight). The descriptive statistics show the mean crossing speed of female pedestrians (straight = 1.47m/s, path changing = 1.27m/s) and male pedestrians (straight = 1.59m/s, path changing = 1.26m/s). The interpretation shows crossing speed of males is higher than females, but the crossing speed of straight-moving pedestrians is higher than path-changing pedestrians. This indicates the road safety of pedestrians irrespective of age group path changing shows a higher exposure to the road than straight moving pedestrians. The ANOVA test shows a 'p-value <0.05, rejects the null hypothesis, and concludes there is a significant difference among the gender and behaviour (path change and straight).

A one-way ANOVA test was performed to compare the effect of path-changing and straight-moving pedestrians on crossing time separately for Locations 1,2, and 3. The result of the one-way ANOVA test carried out for Location 1 revealed that there was a significant difference in mean crossing time between the straight crossing and pedestrian crossing with path change (F(1, 2817) = [23.37], p = 0.00). Further, similar results were obtained from the data collected from locations 2 and 3, showing that there was a significant difference in mean crossing time between the crossing behaviour of pedestrians (F(1, 2407) = [9.859], p = 0.002) and (F(1, 3488) = [17.728], p = 0.00) respectively. Tables 3 and 4 show the results of ANOVA result and descriptive statistics based on different locations. It is observed that the mean crossing time for path-changing pedestrians is higher than for straight-moving pedestrians. Apart from the increased crossing time, the non-yielding nature of motor vehicles for pedestrian crossing with path change at locations other than marked crossings becomes more vulnerable to safety issues.

The results of the previous section show that the road crossing time is different for path-changing and straight-moving pedestrians. So instead of using the crossing speed, crossing time (HIGH/LOW) with reference to the base condition (crossing time for 1.2m/s) is used in the further analysis. The crossing time base condition is calculated with respect to the width of the road and the speed. If the road width is 14 m, then the base crossing time will be 11.6 seconds with respect to 1.2m/s, and if the observed value is above 11.6 seconds, then the crossing time is High. A chi-square test is conducted to identify the variation in pedestrians' crossing time (High/Low) with respect to behaviour. Further, the dataset is changed by replacing crossing speed with crossing time High/Low. The pedestrian behaviour (path change and straight) with respect to crossing time High/Low is compared by Chi-square test; the result shows a p-value <0.05, so there is a significant difference between crossing time (high/low) and their behaviour. When comparing gender and behaviour, there is no significant difference, but there is a significant difference among the different age groups and behaviour. The results of the stage crossing show that there is a significant difference between the behaviour (path change and straight) and different stages of crossing. Also, the result of driver-yielding behaviour shows a significant difference with respect to the behaviour of pedestrians. Further, the parametric analysis and binary logit model are developed by including crossing time High/Low as a variable instead of crossing speed.

The parametric analysis for location 1 shows that the percentage of pedestrians making path change is 57 and straight crossing is 43. Since the location is near the bus terminal, the pedestrian population in that area is regular or practised road users. Further, the zero-stage crossing behaviour is observed in a higher percentage. When comparing to all locations, the crossing time HIGH is observed to be lesser; this also may be due to the reason that the stage crossing behaviour is less observed. The classification from location 2 of mixed land-use type observes 93.5% of pedestrians make path-changing behaviour. Since location 2 is of mixed land use type, the people using the location of daily users are habituated to make path-changing behaviour. Though the path-changing behaviour is an unsafe pattern of road crossing, it is observed high at location 2, but the percentage is a little lesser at locations 1 and 3. Among all the sites, the percentage of children is very less, and also, in our case, the children are observed in associated with higher age groups. In context with the stage of crossing, nearly 80% of the pedestrian is observed to cross with zero stage crossing, i.e., do not involve in making stage crossing. But a smaller extent of 10% is observed in locations 1 and 2. This zero-stage crossing is a healthier crossing decision due to having fewer encounters with the driver's behaviour. In view of genders, the percentage of males is higher at all the locations. The detailed classification of variables based on different locations is shown in Fig. (3). Further, previous research on the pedestrian crosswalk level of service shows the behaviour of pedestrians changes with respect to different land use types [43]. Also, the study highlights the importance of land use in planning and designing a pedestrian facility. The classification identifies an unequal distribution of pedestrian population for different cases. It is due to the reasons that pedestrians have probabilistic behaviour during all their activities. But the behaviour can be predicted to a reasonable extent using the binary logit model described in the next section.

Fig. (3). Classification of Pedestrian behaviour based on different variables.

The analysis of mean crossing time at study location 1 shows that female pedestrians require 5.2% higher crossing time than the male group. Similarly, the mean crossing time of female pedestrians was observed to be 1.89% and 2.14% higher than the same for male pedestrians at locations 2 and 3, respectively. With reference to male pedestrians, every unit increase in female pedestrians will result in road crossing time (High). The study identifies that with an increase in female pedestrians, the chance of having crossing time (Low) decreases. When comparing males and females, it is observed that the female gender is exposed for a longer time at a crosswalk. Also, an earlier study shows that the crossing speed of females is lesser than males [45].

The results of the analysis of the age of pedestrian crossing at location 1 show that the mean crossing time of senior pedestrians is 12.13% higher than young, 29.84% higher than children, and 19.38% higher than adult groups. The same has been observed as 13.76%, 23.21% and 19.81% higher while comparing with young, children and adult groups at location 2 and 5.54%, 21.68% and 2.97% higher at location 3 for young, children and adult pedestrians, respectively. From the result with reference to young pedestrians, the increase in senior pedestrians results in higher road crossing time. In case of an increase in adult and child pedestrians, the crossing time is lesser when compared with a young pedestrian. When there is an increase in senior pedestrians, the chance of having crossing time (Low) decreases, which leads to greater road exposure than other age groups. The earlier study by [31] confirms the senior pedestrian crossroad at a lesser speed. Further, there is an unequal distribution of the population of different age groups due to the location of land-use type. Moreover, the percentage of children is very low, and they perform crossing activities in association with elders.

The stage crossing behaviour is observed when the road has more lanes. In our study, increasing stage number results decrease the chance of crossing time (Low). Our present study shows the increase in the stage is observed with an increase in road crossing time. This can be interpreted as road exposure increases with an increase in stage number. But in our case, the maximum road width is 14m; hence, stages number greater than two are observed in a very low proportion. Further, earlier studies on stage crossing show stage crossing is an important parameter in analysing the crossing behaviour [46].

The yielding behaviour of drivers has a greater impact on pedestrian-safe road crossing. Based on our analysis, the vehicle absence circumstance shows a road crossing time (High) for pedestrians. This high road crossing time is due to the reduced crossing speed adopted by the pedestrian in the absence of a motor vehicle. But the result of earlier works shows that an increase in pedestrian platoon size increases driver-yielding behaviour during road crossing [47].

The crossing pattern is a variable represented as straight and path-changing behaviour, whereas path change represents an irregular pattern of road-crossing behaviour. An increase in path change behaviour results in road crossing time (High) for pedestrians. It shows that crossing time is higher when the pedestrian makes a path-changing pattern of the road crossing. The earlier study [46] shows the safety margin is very lower when they make an oblique pattern of road crossing than straight road crossing. The oblique pattern of road crossing is also a type of path-changing behaviour observed at the crosswalk.

The path change is observed to be an unsafe method of road crossing, but it is mandatory to improve road user's safety by implementing suitable policies. As per the regulation given by the central motor vehicle act [ 48 ], motorists should give way to the person crossing the road when it is not regulated by a traffic light or traffic police at a zebra marking. The driver should show courtesy and consideration for the safety and convenience of users by anticipating their behaviour. The law describes that the driver should yield at the zebra marking area. But in our study, an increase in path change behaviour results in increased road crossing time for pedestrians. This is due to the irregular pattern followed by pedestrians during the road-crossing activity. Enforcing new policies could increase the safety of pedestrians crossing the road and other road users. Also, recent statistics show 6% of road accidents in urban areas occur at pedestrian crossings [2]. According to statistics, a pedestrian was killed every 81 minutes in traffic crashes in 2020 [3]. Further, the previous study observed the path change is greater at the location without zebra marking [10]. To overcome this undisciplined crossing behaviour and increase road user safety, providing guard rails at the median and footpath other than zebra marking locations can be an effective solution for the issue, even though it involves considerable expenditure. Also, the location where guard rails cannot be provided is recommended to provide easily identifiable footpath marking indicating the locations of zebra marking facilities. The footpath marking could guide the pedestrian to nearby zebra marking, which psychologically drives the pedestrian to use the zebra marking location.