Evaluation of Data Needs, Crash Surrogates, and Analysis Methods to Address Lane Departure Research Questions Using Naturalistic Driving Study Data (2011)

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.

A P P E N D I X C Assessment of Availability of Roadway Data Elements in the SHRP 2 Naturalistic Driving Study Data Acquisition SystemBackground A preliminary test run of the SHRP 2 data acquisition system (DAS) was evaluated to determine what roadway data elements could feasibly be extracted in the event they are not available in roadway data sets from SHRP 2 Safety Project S04A (Roadway Information Database Developer, Technical Coordination, and Quality Assurance for Mobile Data Collection) or Safety Project S04B (Mobile Data Collection). A list of roadway fac- tors that are necessary to answer lane departure research questions was developed by the CTRE team as part of their efforts in Safety Project S04A. The research team for this proj- ect reviewed the list of roadway data elements and DAS data set and commented on which roadway factors can be extracted from the DAS data set which may be useful to researchers for Safety Project S08 (Analysis of the SHRP 2 Naturalistic Driving Study Data), as well as to other researchers. Ideally, roadway information will be available from the mobile mapping data collection for Safety Project S04B. However, Safety Project S04B data collection will not cover all areas where naturalistic driving study data will be collected, and not all necessary fac- tors will be collected under Safety Project S04B. It is there- fore important to comment on whether these factors can be reduced from the naturalistic driving study data when they cannot be obtained from existing data sets or the roadway data set. Evaluation of the accuracy of the lane tracking system and of the global positioning system (GPS) is beyond the scope of this project. Description of DAS VTTI instrumented a test vehicle with what is expected to be the final version of the data acquisition system that will be used by the pilot study sites in SHRP 2 Safety Project S07 (In- Vehicle Driving Behavior Field Study). The CTRE team received two sets of test runs from VTTI. The data were col-124lected over the test route that had been surveyed for vendors participating in the mobile mapping data collection rodeo for SHRP 2 Safety Project S03 (Roadway Measurement System Evaluation) (Figure C.1). The data received included two video files with four video views (Figure C.2) and a database with raw system data that included GPS coordinates for the following vehicle data: • Vehicle kinematics (e.g., speed, forward acceleration); • Lane position; • Presence of lane lines; • Steering wheel position; • Turn signal state; • Temperature; and • Light level.Methodology to Extract Data Elements from DAS The team reviewed the GPS traces and the forward and back videos to determine roadway data items that could be extracted. The team assessed the data using three methods as described in the following sections. Comparing List of Roadway Data Elements to Forward Video First, the team used a list of roadway data elements that had been identified as part of SHRP 2 Safety Project S04A. The list identified roadway data elements that had been indicated as being important in addressing either road departure or inter- section crashes research questions. Research questions from Safety Projects S01 (Development of Analysis Methods Using Recent Data), S02 (Integration of Analysis Methods and Development of Analysis Plan), S05 (Design of the In-Vehicle Driving Behavior and Crash Risk Study), and S06 (Technical Coordination and Quality Control) were reviewed and data

125Figure C.1. Trace of DAS activity data.Figure C.2. Video views provided with DAS.elements identified. Next, the team reviewed data elements included in the Highway Safety Information System (HSIS), Model Inventory of Roadway Elements (MIRE), and the Model Minimum Uniform Crash Criteria (MMUCC). Data elements included in those inventories not already identified were added to the list. A survey was conducted under SHRP 2 Safety Project S04A to solicit additional user input. Roadway data elements not already identified were added to the list and a final list of roadway data elements that would be necessary to comprehensively answer lane departure or intersection safety questions was completed. The team reviewed the forward video, and as each item was encountered, they noted whether the data element could beseen in the forward video and whether it was likely that data reductionists could identify and extract the data element. The list of data elements and assessment of whether it could be col- lected using the forward video is provided in Tables C.1 to C.5. The team primarily tested whether a feature was present. Loca- tion of an object can be associated with a corresponding road- way feature. For instance, a vehicle trace can be located to a particular curve and forward imagery could indicate presence of several chevrons. As a result, the curve could be coded as having chevrons. However, spacing of the chevrons or exact location of the chevrons could not be determined. An approx- imate location could be identified if the forward video frame, vehicle’s spatial position, and location of a proximate feature on an aerial image could be linked. This would be significantly affected by the accuracy of the GPS.Length cannot be accurately calculated. However, length of an object, such as length of a horizontal curve, could be roughly estimated if the begin and end points are located in the forward frame of the video, which provides an estimate of time (t). Vehicle speed (v) can be extracted from vehicle data in the DAS and length (L) calculated as: As indicated, it is possible to approximate location and length of an object. However, there was no feasible way to test the accuracy of these measurements for this exercise. As a result, the effort focused on identification of features. Elements included in the list of roadway data elements that could not fea- sibly be extracted are indicated as not applicable (NA). In some cases a feature was included in the list but the data reduction- ists did not view that object in the DAS database. This may be because the feature was not present in the test run made by the L v t= 

126Table C.1. Identification of Curve Data Elements and Roadway Cross-Section Features Data Element Features DAS Horizontal curvature Vertical curvature Cross section Shoulder Presence Length Location Presence and type of spirals Tangent length between adjacent curves Radius or degree of curve Presence and amount of superelevation Direction of curve Curve deflection angle Vertical curve length Tangent length between adjacent curves Grade (percent) Grade direction Terrain type Grade length Lane width Surface width Number of lanes Lane direction (one way, two way, auxiliary, reversible) Cross slope Type (e.g., regular; two-way left-turn lane) Turn lane length Surface type Median type Median width Curb type Type and characteristics of bicycle facilities On-street parking type Right/left shoulder type Right/left shoulder paved width Right/left total shoulder width Right/left shoulder slope Right/left shoulder condition Yes Could be approximated from time between frames and vehicle speed Begin and end points can be approximately located NA Could be approximated from time between frames and vehicle speed NA NA Can be determined from driver perspective; left-hand vs. right-hand NA Difficult to establish the begin point of a vertical curve NA NA NA Can determine flat vs. hilly vs. mountainous Difficult to establish the begin point of a grade NA NA Yes Yes NA Yes NA Yes Yes; could differentiate between grass, flush, raised NA Yes Either was not present or could not be identified Yes Yes NA NA NA NA

127Table C.2. Identification of Signs by Type Data Element Features DAS Regulatory signs Warning signs Guide signs Service signs (e.g., camping, food) Other Speed limit Pass/no pass zones Other (lane end, do not enter, no parking) School area Railroad crossing Stop Yield Horizontal alignment signs and location (e.g., chevron, curve advisory speed) Roadway cross-section changes (e.g., lane ends) Vehicular warning (e.g., horse and buggy) Nonvehicular warning (e.g., deer, pedestrian, snowmobiles) Object markers Speed reduction Slippery when wet Guide destination signs (type and location) Route signs (type and location) Route sign auxiliary signs (type and location) Advance turn and directional arrow auxiliary signs (type and location) Sign type School crossing Yes; in some cases sign text was blurred Either was not present or could not be identified Yes; in some cases sign text was blurred Yes; although text was blurred in some cases, sign shape was distinct Either was not present or could not be identified Yes; shape was distinct Yes; shape was distinct Yes; chevrons can be identified; with curve advisory signs, in some cases sign text/symbol was blurred Yes; in some cases sign text was blurred Either was not present or could not be identified Yes; in some cases sign text was blurred Yes; shape was distinct Yes; in some cases sign text was blurred Either was not present or could not be identified Yes; although text was blurred in some cases, sign shape was distinct Yes; in some cases sign text was blurred Yes; in some cases sign text was blurred Yes; in some cases sign text was blurred Yes; in some cases sign text was blurred Yes; shape was distinctVTTI instrumented car (e.g., no roundabouts were present) or the feature may have been present but data reductionists did not observe an instances of the feature. It should be noted that both data sets were collected under clear conditions in the daytime. There was some glare, but there were no adverse ambient conditions. Consequently, the ability to identify features does not account for that variation in conditions that will be present in the full-scale study. Table C.1 indicates which curve and roadway cross-section features could be identified. The items apply to both tangent sections and intersections. The presence of horizontal and vertical curves could be easily determined for pronounced curves. A very flat vertical curve or horizontal curve with a large radius would be difficult to identify. Most of the features of a curve cannot be determined (e.g., radius, supereleva- tion). Width of objects cannot be determined using any of the DAS data elements.Figure C.3 illustrates some of the cross-section data elements as viewed by the data reductionist.A list of sign types that were identified as being important to lane departure or intersection research questions is listed in Table C.2, along with an indication if they can be identified in the forward imagery of the DAS. However, the text or symbols on the sign face was often difficult to read. Signs with distinct shapes (e.g., stop, yield) were the easiest to identify. The sign face for speed limit signs was usually legible. It was more diffi- cult to identify signs when there was significant glare or foliage along the roadway. It was also difficult to detect signs when there was on-street parking or the test vehicle was traveling on an inside lane away from the road edge. This was particularly problematic when other vehicles were between the test vehicle and road edge. Table C.2 indicates if the sign could be identi- fied. Figure C.4 illustrates some of the sign data elements as viewed by the data reductionists.

128Table C.3. Identification of Pavement Marking and Lighting Data Element Features DAS Pavement markings Illumination Edge line Centerline (e.g., dashed, solid) Location of pass/no pass Lane line Center island Arrows (e.g., merge, left only) Text (e.g., Slow, School Ahead) Raised pavement markings Stop and yield lines Crosswalks Parking Other (e.g., speed hump, HOV, colored pavement, curve ahead) Overhead lighting type Overhead lighting location Overhead lighting characteristics (e.g., lumens) Type of in-pavement lighting Location of in-pavement lighting Yes Yes NA Yes Yes Yes Yes Either was not present or could not be identified Yes Yes Yes Yes, found instance of colored pavement Presence and type of mast arm could be determined NA NA Did not encounter in database Did not encounter in databasePavement markings were evaluated and are listed by type in Table C.3. Pavement markings in almost all cases were eas- ily identified. Pavement markings included lane lines, painted median/gore areas, and on-pavement markings such as stop bars or turn lane designations. In most cases, a qualitative assessment of condition was possible. For instance, markings could be categorized by grouping such as “like new,” “good condition,” “faded but visible,” or “faded barely visible.” An assessment of lighting is also included in Table C.3. Examples of pavement markings and lighting from the analysis are shown in Figure C.5.Evaluation of roadway surface elements and identification of objects in the clear zone are listed in Table C.4. Surface type could be identified between asphalt, concrete, and gravel. No surface condition data elements (e.g., friction, roughness) can be determined. Only obvious roadway defects, such as patch- ing, could be identified in the forward video. Clear zone ele- ments were also included. The type of objects within the clear zone could easily be determined. This includes trees, utility poles, guardrails, and so forth. Examples of roadway sur- face and clear zone elements from the analysis are shown in Figure C.6.Table C.5 provides information about how well counter- measures and access management features could be extracted.Only centerline rumble strips were identified in the forward video. Edge line, shoulder line, and advance stop line rumble strips were either not located along any of the roadways or were present but could not be identified. One speed feedback sign and one flashing beacon were identified in the data. Although it was not possible to read the text on the feedback sign, it was possible to determine through the forward video whether the sign was activated or turned off. Presence of driveways and type of median can be identified in the DAS data set. Together they can be used to estimate level of access control. Bridge charac- teristics were also included in Table C.5. Bridge type and pres- ence of barriers and abutments could be determined from the forward video. Examples of data elements as they appeared in the data set are provided in Figure C.7.The data elements listed in Tables C.1 to C.5 apply to the entire roadway. Data elements in Table C.6 are specific to intersections. Several cross-section features were identified as important elements that are specific to intersections. Number of lanes for each approach of an intersection could usually be determined if the cross streets were visible within the forward video frame. The number of lanes could always be determined for the approach where the vehicle was located. Left- and right- turn prohibitions can be determined if the corresponding sign can be identified.

129Table C.4. Identification of Roadway Surface and Clear Zone Characteristics Data Element Features DAS Road surface Roadway defects Clear zone Surface type (e.g., gravel, asphalt, PCC) Surface friction Macro-texture Pavement roughness Pavement condition Roadway rideability Pavement edge drop-off Roughness Surface irregularities Road debris (best source would be forward video) Type of objects within clear zone (tree, utility pole, sign) Clear zone distance Slope beyond edge of shoulder Presence, type of guardrail Guardrail end Guardrail face Curb presence Curb type Right-of-way Roadside hardware types and location (e.g., barriers, culverts) On-street parking Concrete barrier Other longitudinal barriers Yes NA NA NA NA NA NA NA Some irregularities could be determined, but it was difficult to distinguish from shade Yes; observed plastic bag flying down the street Yes NA NA Yes Yes Yes Yes Yes NA Could only identify if hazard marking was present Yes Yes YesThe type of intersection control by approach could be determined in all cases for the approach where the vehicle was located. In some cases, type of control could be deter- mined for adjacent or opposing approaches. When a signal is present, it can be inferred that all approaches are signal- ized. When a stop sign is present, it can be assumed that the opposing approach is also stop controlled if it cannot be deter- mined from the forward view. When no control is noted for the approach where the test vehicle was traveling and con- trol for adjacent approaches cannot be determined, it may be difficult to determine if the intersection is uncontrolled or whether the adjacent streets have stop control. When stop control is present for the approach where the vehicle is located and the control cannot be identified for adjacent approaches, it may be difficult to determine whether the intersection is two-way or four-way stop controlled. No instances of advancestop line rumble strips and red-light-running cameras were found. It is possible that they were present but were not identified. The signal phase (red, yellow, green) could be determined for the approach where the vehicle was traveling and can be inferred for other approaches by vehicle movements. Signal phase state could be identified in almost all situations, with the exception of turn arrows. Turn arrows were very difficult to discern in many cases. Signal progression could be inferred based on the number of times the vehicle received the green phase through a series of intersections. Several intersection features are shown in Figure C.8 as they were identified in the forward video.Red-light running is a significant cause of many intersec- tion crashes. The team therefore reviewed the forward video view to determine whether a data reductionist could identify

130Table C.5. Identification of Other Countermeasures and Access Data Element Features DAS Other counter- measures Access Bridge structures Type of edge line or shoulder rumble strips Location of edge line or shoulder RS Type of centerline rumble strips Location of centerline rumble strips Advance stop line rumble strips Type of speed feedback signs Channelizers, delineators Presence of safety edge Automated speed enforcement Cable barrier Crash attenuators/cushion Vertical deflection (e.g., speed tables, raised intersection) Application of high friction surfaces Driveway density Roadway facility type (e.g., collector, arterial) Access control Type (overpass, underpass, water crossing) Bridge deck width Barriers (e.g., railing) Abutments Either was not present or could not be identified Either was not present or could not be identified Either was not present or could not be identified Either was not present or could not be identified Either was not present or could not be identified Encountered one speed feedback sign; speed display was blurred, but it was possible to tell if activated or not Yes NA Could only be detected if signing was present; either was not present or could not be located Yes Yes Was not present in database NA Driveways could be identified and counted NA Could determine presence and type of median as well as driveways; access could be inferred Yes NA Yes Yeswhether the instrumented vehicle ran the red light, although red-light running is technically not a roadway data element. The signal state could be observed in most cases, so the signal change from yellow to red and the position of the instrumented vehicle relative to the stop bar could usually be observed. Fig- ure C.9 shows the signal turn from green to yellow to red. Since signal state can be determined, a situation where the front of the vehicle has crossed the stop bar after the signal turns red could be identified as running the red light. Figure C.10 shows the instrumented test vehicle crossing the stop bar while the signal is yellow.Evaluation of Percentage of Time That Feature Characteristics Can Be Identified The second evaluation method assessed the number of times characteristics of a particular type of feature could be identified.The team selected several critical items that would be necessary in evaluating lane departure or intersection crashes. At least 10 of the data items were identified, if present, and a determination made as to whether features could be extracted. For instance, a number of regulatory and advisory signs were identified. Then the number of times the sign message could be detected was recorded. For instance, an advisory sign may be detected by color and shape, but the text or symbol could not be read. Results are shown in Table C.7 for the majority of features. As the table indicates, chevrons were located three times in the DAS databases. The chevron symbol and number of chevrons could be identified in all three instances. Work zone signs were encountered five times and were assessed to deter- mine whether the sign text or symbol could be interpreted. In all five cases, no text or symbol could be identified. Ten route signs were extracted and 70% of the time the route or guiding (text continues on page 137)

131Figure C.3. Identification of cross-section elements: (a) illustrates change from asphalt to concrete (surface type); (b) paved/earth shoulder; (c) sidewalk along roadway; and (d) raised median. (b)(a) (d)(c)

Figure C.4. Identification of sign types: (a) difficulty identifying signs with glare or vegetation present; (b) guide sign; (c) warning sign face visible; (d) warning sign face not visible; (e) speed limit sign; and (f) turn prohibition. (b)(a) (d) (c) (e) (f )

Figure C.5. Identification of pavement markings and lighting: (a) crosswalk; (b) turn arrows; (c) other on- pavement markings; (d) fading school sign; (e) overhead lighting (cobra head); and (f) decorative street light. (b)(a) (d) (c) (f)(e)

134Figure C.6. Identification of roadway surface and clear zone elements: (a) guardrail within clear zone and (b) pavement damage. (b) (a)Figure C.7. Identification of countermeasures: (a) colored pavement marking and (b) speed feedback sign (activated). (b) (a)

135Table C.6. Identification of Intersection Features Data Element Features DAS Cross section Intersection control Other counter- measures Other Intersection/interchange type (includes railroad) Number of lanes by approach Number of approaches Left- and right-turn prohibitions Intersection skew angle Channelization (islands) Intersection offset (whether crossroad approach centerlines are directly opposed or offset by some distance) Intersection offset distance Control by approach (includes railroad crossing signals) Signal characteristic (e.g., signal head configuration, lens size) Type of signalization (e.g., fixed, actuated) Presence of left-turn arrow (indication of left-turn phasing) Detector type Overhead beacons Pedestrian signal Pedestrian signal features (e.g., push button) Advance stop line rumble strips Red-light running countermeasures Sight distance Signal progression Traffic signal state Red-light running Yes Yes; for the approach where the vehicle is located Yes If the corresponding sign can be identified NA Yes NA NA Yes; for the approach where the vehicle is located Lens head configuration could usually be determined NA Was very difficult to see turn arrows, but signal head configuration could be determined, so presence of left arrow could be inferred NA Either was not present or could not be identified Could detect presence if in line of sight when passing through intersection NA Either was not present or could not identify NA An estimate can be made based on driver’s line of sight from forward view Signal state could be determined in most cases, so progression could be inferred Yes, except for left/right arrow Yes, if stop line can be identified

136(b)(a) (d) (c) (f) (e) Figure C.8. Identification of intersection features: (a) channelization; (b) turn prohibition; (c) signal head configuration; (d) signal phase easily identified (note green left-turn arrow); (e) signal phase difficult to identify for left turn; and (f) signal phase difficult to identify.

137(b) (a) (c) Figure C.9. Features to identify red-light running: (a) signal identified as green; (b) signal identified as yellow; and (c) signal identified as red.information could be read. Ten on-pavement markings were also selected and, in all cases, the type of marking could be identified. This included left turn and “SCHOOL.” Sixteen speed limit signs were extracted and the numeric speed limit could be identified for all of the signs. In some cases, the text “Speed Limit” may not have been legible, but the sign could be identified since it was distinct. Fifteen reg- ulatory signs other than speed limit, stop, or yield, were identified by shape and color. The sign message could be determined for 73% of the signs. Signals were identified for the approach where the instrumented vehicle was trav- eling for 26 intersections. The overhead signal phase could (continued from page 130) be determined 100% for through movements. In some cases it was difficult to determine the state for left-turn arrows. Thirty-five warning signs were identified and an attempt was made to identify the sign text or symbol. The signs only included those with the traditional diamond shape. Warning signs such as channelizers, which have a distinct shape, were not included in this assessment. When the sign message could be determined, it is indicated in Table C.8. When the message was not legible, it was listed by whether it was text or symbol, since this could be determined even if the exact message could not. The sign message could be determined for around 46% of the signs.

138Figure C.10. Vehicle crossing stop bar during yellow.Table C.7. Assessment of Ability to Identify Features Data Element Feature Assessed Times Encountered Times Recognized Chevrons Chevron symbol and number of chevrons 3 3 100.0% Work zone sign Message 5 0.0% Guide/route sign Message 10 7 70.0% On-pavement marking Message 10 10 100.0% Speed limit Speed limit 16 16 100.0% Regulatory 15 11 73.3% Signal Phase 26 26 100.0% Stop sign Message 3 3 100.0%Table C.8. Assessment of Warning Signs Times Times Warning Sign Type Encountered Recognized Total warning signs 35 16 End of divided roadway 3 3 Lane ends 1 1 Left-hand curve 2 2 Merge 2 2 Reduced speed ahead 3 3 Right-hand curve 1 1 Signal ahead 2 2 Start divided highway 1 1 Watch for deer 1 1 Unknown symbol 9 0 Unknown text 10 0Comparison of Data Elements in DAS with Safety Project S03 Rodeo Elements Several of the data elements that were collected as part of SHRP 2 Safety Project S03 data collection rodeo with mobile mapping vans were used to determine whether the same data item could be identified in the DAS data set. Data elements from the rodeo data set were overlaid with the GPS vehicle traces from the DAS data. The GPS data points nearest the feature in question were selected and corresponding time stamps noted. The time stamps were located in the forward video and the forward view searched for several frames before and after to locate the object. The process is depicted in Figure C.11. If the object was located, it was noted and compared against the description for the data element in the rodeo data set.Eighteen signs were extracted from the rodeo data. Two signs could not be located in the DAS in any reasonable prox- imity. Many of the signs were identified by shape or color, butin many cases the text or symbol was not legible. Results are shown in Table C.9.Table C.10 shows identification of pavement markings. As noted, seven pavement markings that were selected in the rodeo data could be identified in the forward video.Miscellaneous other objects were also compared between the two data sets as shown in Table C.11. Two streetlights were located in the mobile mapping data. One could be identified in the forward video in the approximate location. The second streetlight was located in an area of heavy vegetation and sig- nificant glare was present in the forward image. As a result, the streetlight could not be distinguished from other background features. One segment was indicated as having centerline rum- ble strips in the mobile mapping data. The approximate begin

139Figure C.11. Method to compare objects in mobile mapping database with features identifiable in the DAS forward view.Table C.9. Comparison of Signs Between Rodeo Data and DAS Sign Type How Sign Was Identified Speed limit Could not locate Exit sign Could not locate Merge Merge symbol Food next right Identified blue sign with text but could not read text Gas next right Identified blue sign with business symbols Overhead guide Overhead guide sign Adopt a highway Identified small blue sign but could not read text School speed limit Flashing beacon when flashing Object marker Object marker School sign School sign State route Route sign Object marker Object marker School sign School sign Right lane must turn Identified white sign Right lane must turn Identified sign over right-turn lane Right lane must turn Identified sign over right-turn lane Stop here on red By textTable C.10. Comparison of Pavement Markings Between Rodeo Data and DAS Pavement Markings How Markings Were Identified Stop bar By shape School Text “SCHOOL” Stop bar By shape Right turn only By shape Right-turn arrow By shape Stop bar By shape Stop bar By shapeTable C.11. Comparison of Other Objects Between Rodeo Data and DAS Object How Object Was Identified Street light By shape Street light Could not identify, heavy tree cover and bad sun angle Centerline rumble strips Identified approximate beginning of rumble strips Guardrail Identified end of guardrail

140point of the rumble strip section was located. The presence of guardrail in the forward video was easily identified. Summary The team reviewed sample data sets that were collected using what is expected to be the final version of the instrumented vehicle data acquisition system. The data were collected by VTTI using a test vehicle that traversed the same route where data were collected for a demonstration of mobile mapping vendor capabilities. Three methods were used to evaluate how well roadway data elements could be identified using data from the DAS and how feasible data extraction using this method would be. In the first method, data reductionists reviewed the forward video from the DAS to assess which roadway features could be identified using manual data reduction. The team had devel- oped a list of relevant roadway data elements as part of their work for Safety Project S04A. The team determined which of the data elements could be collected, if applicable. Mostfeatures could be recognized to some degree in the DAS for- ward video. The presence of a sign could be determined in most cases, but text and symbols were frequently illegible. In the second method, the team extracted a sample of some data elements and examined them to determine what percent- age of the time certain features about the element could be identified. For instance the team identified 35 instances of warning signs. The text or symbol could only be recognized slightly less than half of the time. In the third method, location of several data elements col- lected in Safety Project S03 for the mobile mapping rodeo were compared with the GPS position and the time stamped from the DAS. The location of the object in the forward video was then identified, if possible. In summary, a large number of roadway features, including traffic signal state, could be identified in the DAS. The general location of roadway features relative to the roadway could be determined, but actual location could not be established. For instance, it was possible to determine that a school crossing sign was located just before a crosswalk.

Integration of Analysis Methods and Development of Analysis Plan (S02) Roadway Measurement System Evaluation (S03) Roadway Information Database Developer, Technical Coordination, and Quality Assurance for Mobile Data Collection (S04A) Mobile Data Collection (S04B) Design of the In-Vehicle Driving Behavior and Crash Risk Study (S05) Technical Coordination and Quality Control (S06) In-Vehicle Driving Behavior Field Study (S07) Analysis of the SHRP 2 Naturalistic Driving Study Data (S08)