Guidebook on Pedestrian and Bicycle Volume Data Collection (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

121 A p p e n d i x C Introduction This booklet will serve as a guide for Research Assistants working on NCHRP 07-19: Methods and Technologies for Pedestrian and Bicycle Volume Data Collection. According to the NCHRP website, “the objective of this research is to assess existing, new, and innovative technologies and methods and provide guidance for transportation practitioners on how to best collect pedes- trian and bicycle volume data. The assessment should consider, among other factors, feasibility, availability, quality, reliability, cost, and compatibility. The guidance should include methods to (a) efficiently mine and manage existing data sources; (b) acquire and use data from new and innovative technologies; and (c) summarize and disseminate pedestrian and bicycle volume data for site-specific, local, and system wide needs assessments, project development, and safety management.” The bulk of the research entails installing automated bicycle and pedestrian counting devices in the field and evaluating the accuracy of these technologies under a variety of con- ditions. We are installing counters at 13 sites spread across Davis (CA), San Francisco (CA), Portland (OR), Minneapolis (MN), and Washington (DC)/Arlington (VA). Additionally, we will use some pre-existing counters located in Montreal (QC, CAN). The accuracy evaluation involves recording video footage of the installation sites on a short-term basis and selecting 15-minute intervals of the video to conduct manual counts of bicyclists and pedestrians. These manual counts will then be compared to the automated counts gathered by our equip- ment in the field. This is where you come in! Performing manual counts of bicyclists and pedestrians is a vital component of this research, and will enable us to determine how well the equipment is working. If you have any questions on any items in this process, please do not hesitate to ask. It will be far better to work out any issues as they arise than to try and troubleshoot in hindsight. Count Protocol Used for NCHRP Project 07-19 About This Count Protocol These instructions reproduce the instructions given to the project staff who developed the project’s ground-truth counts by performing manual counts from video. They are intended to provide a template for others who may wish to per- form similar types of counts. In these instructions, “I” refers to the data collection supervisor and “you” refers to the staff performing the counts.

122 Guidebook on pedestrian and Bicycle Volume data Collection Key Points 1. Thank you for your help! Your work on this project will help to advance the state of the art of bicycle and pedestrian data collection, which in turn has great implications for roadway safety analysis, project prioritization, and bicycle/pedestrian planning in general. 2. Accuracy is of the utmost importance. Please take your time and try not to stress. It’s com- pletely normal to make occasional mistakes, get stressed out, get sleepy, etc. If you feel any of these things happening, take a short break, go drink some coffee, do some push-ups, whatever you need to stay alert. 3. Please do not work on this for more than 3 hours at a time. If you would like to work for more than 3 hours in a day, schedule it around your other obligations or take a break in the middle for at least 30 minutes. Do some homework, go get something to eat, do whatever you would like to do other than count bicyclists and pedestrians. 4. You will undoubtedly find situations where you have to make a judgment call as to how to count a particular event. I’ve done my best to write everything out very clearly in this guide- book, but you never know exactly what you’re going to find. Whenever you run into a gray area, make your best guess, and jot down a note of what the situation was, when it occurred, and in which video file. Either send me an email or hold onto it for our weekly meeting so we can decide on an “official ruling.” However, it’s important to not get bogged down with these details. 5. Before performing any official counts, please read this guidebook at least twice. We will also do an in-person training. This will give you an opportunity to perform practice counts and will allow us to perform a test for inter-rater reliability (i.e., how accurate are your counts relative to the counts recorded by another student?). 6. Please follow all directions in the guide closely—I know there are a lot, but the more consistent that we can be, the stronger our results will be. 7. I will print out a copy of this guide for each RA, as well as a couple to keep in the office. Keep a copy handy when you’re working in case you need to look anything up. How to Count Bicycles and Pedestrians What Counts in Each Category The first thing to consider is how to classify any events that occur. Generally, you will record a count every single time a pedestrian or bicyclist passes through the detection zone (explained below) of each piece of equipment. There is a certain degree of nuance associated with what qualifies as a “pedestrian” and what qualifies as a “bicyclist.” Below are lists of par- ticular cases and where they fit within the schema of pedestrians/bicyclists for the purposes of this project. Pedestrians • Pedestrians • Pedestrians walking dogs (count as one pedestrian, but write this down in the notes field any time you see it) • People walking with walkers • Pedestrian carrying a child in arms or wheeling a baby in a stroller (count as a single pedes- trian but write this down in the notes field any time you see it) • People rollerblading or skateboarding • People in wheelchairs • People riding on scooters (e.g. razor scooters—also, make a note of this in the notes field)

Count protocol Used for nCHRp project 07-19 123 Bicyclists • People riding bicycles • Pedestrian walking a bicycle (according to the law, a person walking a bicycle is typically char- acterized as a pedestrian, but we are interested in how well the counting devices detect bicycles, so this should be counted as one bicycle) • Children riding bicycles or tricycles • People on Segways (write this down in the notes any time you see it) Special Case • People riding a tandem bicycle (record this as one bicycle count for an inductive loop, piezo- electric sensor, or pneumatic tube; record this as two people for an infrared or radio beam counter; write this down in the notes field any time you see it) • Bicycles with trailers (record this as one bicycle for an inductive loop, two bicycles for piezo- electric sensor or pneumatic tube, and as however many people go by for an infrared or radio beam counter; write this down in the notes field any time you see it) • Anybody loitering in the detection zone (Make a note of this—this type of behavior can give false counts. Record as a single count, and note roughly how long people loitered in total for the 15 minutes) Detection Zones (by location) In this project, we are testing the following detector/counter types: passive infrared, radio beam, pneumatic tubes, inductive loops, piezoelectric sensors, as well as combinations of these devices. For the purposes of performing manual counts, however, these can be thought of as either being screenline detection zones (i.e., when a pedestrian or bicyclist crosses an imaginary line) or areal detection zones (i.e., when a pedestrian or bicyclist travels through any part of an area). Passive infrared is an example of a screenline—these devices work by detecting body heat in front of them, within a fairly constrained band of space. A count is recorded any time a person crosses this “screenline.” Inductive loops, on the other hand, have an areal detection zone. These devices are installed in or on top of the pavement, and detect the metal in bicycles passing above them. Hence, a count is recorded any time a bicycle passes within the area defined by the bounds of the loop (or maybe slightly more or less). To help with understanding what should be considered a count, a summary of the equipment types being tested and how they operate is provided in Table C-1. Each count site’s video frame has been divided into numbered zones. The coded numbers for each zone correspond with specific data fields where you should enter the individual counts. This is done to determine whether people counted in the video are within the specified detection zones for each piece of counting equipment. Images depicting these numbered zones have been prepared and are provided later in this booklet. Keep these on hand for reference while perform- ing counts in case of uncertainty with whether or not a count should be recorded. Technology Bikes/Peds How it works Detec on Zone Inducve Loops Bikes Detects metal in the detecon zone Area above loop Passive Infrared Both Detects body heat Screenline Radio Beam Both Radio beam between transponder and receiver records a count when broken Screenline Pneumac Tubes Bikes Rubber tubes record counts when compressed Screenline Piezoelectric Sensor Bikes Strips on ground record counts when deformed by bikes passing over Screenline Table C-1. Equipment types tested in NCHRP 07-19.

124 Guidebook on pedestrian and Bicycle Volume data Collection At the site in Montreal shown in Figure C-1, we have three numbered detection zones in a cycletrack. Any time a bicyclist passes through any part of detection zone 1 (the red zone), they should be counted under “Initial Event 1.” The “Initial Event” field numbers are indicated in the box in the upper left-hand corner. For example, if the same bicyclist rode through Box 1 and Box 3, then that bicyclist would be counted under Event 1 and Event 3. We will now get into the more technical aspects of the counting process. Key Software and Files The following are the key pieces of software/files that you will be using for this project, and where you can find them on the SafeTREC Network. • KeyCounter (W:\Grants\G2012_NCHRP_DataCollection\Task5_DataCollection\KeyCounter)—This software counts the number of times you press whatever set of keys you tell it to watch. Copy this over to your local hard drive. KeyCounter can be downloaded from http://skwire. dcmembers.com/wb/pages/software/keycounter.php • VLC Media Player (W:\Grants\G2012_NCHRP_DataCollection\Task5_DataCollection\VideoLAN\VLC)— This is a media player that includes speedup/slowdown options, as well as it works well with the video data formats that we’re receiving. Copy this to your local hard drive as well. Alter- natively, VLC Media Player can be downloaded from http://www.videolan.org/vlc/index.html • Count Database (W:\Grants\G2012_NCHRP_DataCollection\Task5_DataCollection\NCHRP0719Count Database.accdb)—This database is where all data pertaining to the study will be stored. This includes manual counts (your input), automated counts, weather data, and site information. • External Hard Drives—These contain all of the video files. They are kept in the SafeTREC office—ask if you don’t know where to find them. 1 2 3 Event 1: Bicycle (Inducve Loop) Event 2: Bicycle (Inducve Loop) Event 3: Bicycle (Inducve Loop) Event 4: All Bicyclists (count every bicycle once, regardless of where they are) Figure C-1. Montreal University video zone.

Count protocol Used for nCHRp project 07-19 125 Selecting Video All of the video data will arrive to SafeTREC on DVDs shipped from project team members or contractors. Video that needs to be reduced will be stored on the external hard drives. Each external drive has a folder called “NCHRP0719Videos” that contains all of the video files. At the top of this folder is also an Excel file detailing what video clips are on the drive. Video is stored in .vob format, with videos sorted by site/time period. They are grouped into 6 hour blocks. A list of video segments to be reduced will be kept in the file W:\Grants\G2012_NCHRP_DataCollection\ Task5_DataCollection\ManualCounts\VideoList.xlsx. Should you need them for any purpose, the original DVDs are stored on the bookcase in the SafeTREC Office Room #107. Performing Counts And now, for the fun part: how to conduct counts. • First, plug in the external hard drive (to the wall and a USB port on your computer). Open up the external hard drive in Windows Explorer and navigate to the folder of videos. • Navigate to the folder corresponding to the time period that you are planning to conduct counts on. Within each folder, the video files are broken up into smaller chunks (1-2 hours), labeled along the lines of “VIDEO_TS-1-01.vob.” Take a guess as to which one your time period should be in (based on loca- tion within the 6 hour period) and open up the file. If you just double click on it, Windows might have trouble figuring out what to use—direct it to the VLC media player software. If you got the video wrong, select one before or after to find the correct time period. Navigate using the timestamp at the bottom of the video frame. • Open your printed count protocol manual (this book!) to the Appendix Page corresponding to the site that you are analyzing. • Open up KeyCounter as well. Under “Preset keystroke groups,” select “0-9” and press “Add keys to list.” If you click “Start monitoring,” the program will record how many times you press the numeric buttons (even if KeyCounter isn’t your active window). If you want to temporarily stop it from count- ing (like if you go to a web browser or a word document), hit “Stop monitoring.” This will keep your counts active without recording more (until you press “start monitoring” again). “Reset counts” will delete all counts. • Play the video. Whenever a pedestrian or bicyclist crosses a screenline or passes through a detection area, press the corresponding number button. Be very careful to only press the numbered keys when you are recording a count. • VLC media player has hotkey options to control the video—it is highly recommended to use these for pausing, speeding up, or slowing down videos. The hotkey options can be custom- ized in the VLC media player by clicking Tools ≥ Preferences. Navigate to the “Hotkeys” tab at the bottom of the left-hand panel. You can use whatever keys you would like, although don’t

126 Guidebook on pedestrian and Bicycle Volume data Collection use any number keys as you should only be pressing these to record counts. The default set- tings for keys you may want include: Play/Pause Space Faster (fine) ] Slower (fine) [ Normal rate = Very short backwards jump Shift+Left Short backwards jump Alt+Left • As you perform the count, note the following information. These will be recorded as binary (yes/no) characteristics, so give a general assessment across the count period: – Cloudy: It is completely overcast (not partly cloudy) – Foggy: There is visible fog between the camera and background objects – Raining: Enough rain drops can be seen falling to make the pavement wet – Wet Pavement: The pavement is wet – Snowing: Frozen precipitation can be seen falling from the sky – Snow on the Pavement: Frozen precipitation has collected on the pavement – Shade: The detection zone(s) have a shadow falling across them – Daylight: It is fully daylight – Darkness: It is fully dark • Count for the specified time period, being careful to count for exactly 15 minutes. Pay attention to the timestamp at the top of the video to get the number of seconds correct (and the timestamp at the bottom to ensure that you have the right hour and minute!). • After you have finished the 15 minute period, switch to the KeyCounter window and note the number of key strokes for each numbered key. Press “Stop Monitoring” so you don’t acciden- tally alter your counts when entering data into the database. Move to the “Data Entry” step. What to Do if You’re Having Trouble Counting There might be situations where you have trouble counting everything going on. This is likely to occur when there are many detection zones (3+), mixing of modes, and/or high volumes. Do not despair, there are some good tricks to try and smooth out the process. When you get to a period where this seems to be the case, try the following: • Focus on 1-3 detection zones at a time, which can be chosen however you would like to make things easier to count. Note the timestamp when you start using this technique (to avoid miscounting). Run through the difficult period counting on one set of detection zones (again note the ending timestamp), then move the video back to the point where you started and run through again, this time counting the other detection zones. The PATH software will save all of your counts with their timestamps, so the counts will get coded approximately the same as if they had been conducted simultaneously. • Slow down the video (I find × 1⁄8 time is best). • Scroll through the video with the slider bar under the playback pane—this is particularly helpful if you’re having trouble seeing the path taken by a bicyclist. What to Do if a Count Period Runs Across Two or More Video Files If a specified count period runs across the end of a single video file, this will be noted in the Video List. This is not likely to occur, because videos will generally be merged to encompass entire count intervals. However, if a count period spans multiple video clips, follow this procedure: • Ensure that the timestamps in the video pane match up for the two video files • Perform the count across the first video file

Count protocol Used for nCHRp project 07-19 127 • Process the data output and make a note of the count results • Perform counts on the second video clip up to the end of the count interval • Add the count results for the two time periods together Data Entry Once you have calculated values for your 15-minute block of video, open up the count data- base. Open the form “ManualCountsDataEntry”—this will be your primary interaction with the count database. Figure C-2 is a screenshot of this form. For every 15-minute interval that you reduce, fill out this form. This will create a record in the database for each interval. A description of each field follows: • Date: Date the video was taken. • Time: Beginning time of the 15 minutes being reduced. • Video ID: The ID number given to the video, as found in the Video List document. • Student Name: Your name. This is a dropdown menu with all of the Student RAs’ names in it—just find yours. • Count Site: This is another dropdown menu, with an option for each of the count sites. These are all in coded forms, which are summarized later in this document. Figure C-2. Manual count data entry form.

128 Guidebook on pedestrian and Bicycle Volume data Collection Each of the following are checkboxes—check them if the given criterion is satisfied for the count interval, based on the instructions outlined previously in this document. – Cloudy – Foggy – Raining – Wet Pavement – Snowing – Snow on the Pavement – Shade – Daylight – Darkness • Count for Event #: These should be reported exactly as they are in the KeyCounter pane. • Notes: Any notes about the data reduction period. For example, anything that you weren’t sure of, odd occurrences, or reasons we might want to look more closely at this period. After you have entered all of this data, save the form (Ctrl-s). If you have more count intervals to enter, use the record navigator at the bottom of the pane. The rightmost button (“New (blank) record”) is the one to click. Otherwise, you can close the database. Finally, don’t forget to check off the video as “complete” in the list of videos, to avoid duplicating efforts. Now you can start all over with a new video, or else go do something else. Count Site Overviews & Event Definitions 1 2 Event 1 : Pedestrians Event 2 : Bicyclists Berkeley

Count protocol Used for nCHRp project 07-19 129 Event 1: Pedestrians (passive infrared) Event 2: Bicyclists (passive infrared) Event 3: Bicyclists (passive infrared and inducve loops) 3 2 1 Davis (Loyola) 1 2 Event 1: Pedestrians (passive infrared) Event 2: Bicyclists (passive infrared and inducve loops) Davis (Sycamore)

130 Guidebook on pedestrian and Bicycle Volume data Collection 1 2 Event 1: Bicyclists (Pneumac tubes) Event 2: Bicyclists (Inducve loops) Event 3: All Bicyclists riding on street Arlington (Clarendon) Event 1: Pedestrians Event 2: Bicyclists NOTE: If anybody lingers or backtracks between these lines, make a note of it. This shouldn’t be very common. There are mulple technologies here, but they all funcon as screenline sensors. 2 1 Arlington (Key Bridge)

Count protocol Used for nCHRp project 07-19 131 2 1 3 Event 1: Pedestrians (passive infrared) Event 2: Bicyclists (passive infrared, piezoelectric, radio beam) Event 3: Bicyclists (piezoelectric) Arlington (Four Mile Run) 1 2 Event 1: Pedestrians (passive infrared) Event 2: Bicyclists (pneumac tube, inducve loop) Event 3: All bicyclists on street Washington, DC (L Street)

132 Guidebook on pedestrian and Bicycle Volume data Collection 1 2 3 Event 1: Bicycle (Inducve Loop) Event 2: Bicycle (Inducve Loop) Event 3: Bicycle (Inducve Loop) Event 4: All Bicyclists (count every bicycle once, regardless of where they are) Montreal (University, for video from 07/2012) Event 1: Bicyclists (Induc ve Loop) Event 2: Bicyclists (pneuma c tubes) Event 3: Bicyclists (Induc ve Loop) 1 2 3 Montreal (University, for video from 07/2013)

Count protocol Used for nCHRp project 07-19 133 1 Event 1: Bicyclists (Pneumac Tube) Event 2: All Bicyclists (count every bicyclist once, regardless of where they are) Montreal (Milton) 1 2 Event 1: Cyclists (Inducve Loop) Event 2 : Cyclists (Inducve Loop) Event 3: All Bicyclists (count every bicycle once, regardless of where they are) Montreal (Maison)

134 Guidebook on pedestrian and Bicycle Volume data Collection 1 2 Event 1: Cyclists (Inducve Loop) Event 2: Cyclists (Inducve Loop) Event 3: All Bicyclists (count every bicycle once, regardless of where they are) Montreal (Parc) 1 2 3 Event 1: Bicyclists (pneumac tubes in bike lane) Event 2: Bicyclists (Pneumac tubes in bike lane) Event 3: All users (passive infrared, on sidewalk) Event 4: All Bicyclists (“into” video) Event 5: All Bicyclists (“out of” video) Minneapolis (15th Street)

Count protocol Used for nCHRp project 07-19 135 15 4 3 2 Event 1: Pedestrians (infrared, radio beam) Event 2: Bikes (pneumac tubes) Event 3: Bikes (inducve loops) Event 4: Bikes (inducve loops) Event 5: Bikes (pneumac tubes) Minneapolis (Midtown, video from 11/2013) 1 2 3 4 Event 1: Pedestrians (passive infrared, radio beam) Event 2: Bicyclists (pneumac tubes) Event 3: (inducve loops) Event 4: (inducve loops) Minneapolis (Midtown, video from 06/2013)

136 Guidebook on pedestrian and Bicycle Volume data Collection 1 Event 1: Pedestrians Event 2: Pedestrians passing on the le side of the bike rack Portland (5th Avenue) 1 2 Event 1: Bicyclists Event 2: Pedestrians Portland (Eastbank Esplanade)

Count protocol Used for nCHRp project 07-19 137 1 1 3 12 Event 1: Bicyclists (inducve loops) Event 2: Pedestrians across enre sidewalk (passive infrared) Event 3: Bicyclists (pneumac tubes) Event 4: All bicyclists riding on facility San Francisco (Fell Street) Codes for Database SiteID Locaon Name City Berk Bridge near VLSB Berkeley Dav_Loyola Loyola Path (West of Pole Line Road) Davis Dav_Villanova Villanova Path (West of Sycamore Street) Davis DCA_L L Street (exact locaon TBD) Washington DCA_Clarendon Clarendon Boulevard (E of Danville Street) Arlington Co DCA_FourMile Four Mile Run Trail (E of I-395 Underpass) Arlington Co DCA_Key South End of Key Bridge (Lee Highway and Lynn Street) Arlington Co Min_15th 15th Avenue SE (N of University Avenue SE) Minneapolis Min_Midtown Midtown Greenway (E of Humboldt) Minneapolis Mon_Parc Avenue du Parc (N of Avenue des Pins) Montreal Mon_University Rue University (E of Milton) Montreal Mon_Maison Boulevard de Maisonneuve Montreal Mon_Rachel Rue Rachel Montreal Mon_Milton Milton Street Montreal PDX_5th 5th/6th Avenue Transit Mall Portland PDX_Eastbank Eastbank Esplanade Portland SF_Fell Fell Street (W of Sco Street) San Francisco