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.
NCHRP 07â19(02) Final Report 68 ChapterÂ 4:Â ConclusionsÂ andÂ SuggestedÂ ResearchÂ CONCLUSIONSÂ Research findings from Project 07â19 Phases 1 and 2 indicate the automated count technologies that were tested performed with the accuracy summarized below (reported values are Weighted Average Percent Deviation and Pearsonâs r between automated and manual count). ï· Passive infrared. Three products were tested, with average undercount rates ranging between 1.6% and 27.0%, and linear correlation rates between 0.949 and 0.988. ï· Active infrared. One product was tested, with an undercount rate of 7.6% and a correlation value of 0.998. ï· Thermal imaging camera. One product was tested, with an overcount rate of 2.7% and correlation value of 0.912. ï· Pneumatic tubes. Three products were tested, with undercount rates ranging between 9.4% and 59.6%, and correlation values between 0.841 and 0.993. ï· Radio beam. One product was tested with an undercount rate of 12.1% and correlation value of 0.980. ï· Radar. One product was tested with an overcount rate of 14.2% and correlation value of 0.918. ï· Inductive loops. Two products were tested, with average miscount rates ranging between â 3.1% and 4.8%, and correlation values of 0.959 and 0.997. ï· Piezoelectric strips. Two products were tested, with undercount rates of 3.4% and 5.8% and correlation values of 0.994 and 0.997. The following subsections discuss the factors found to influence the accuracy and recommendations for practitioners interested in using automated count technologies. FactorsÂ InfluencingÂ AccuracyÂ The research team found the accuracy measurements to vary notably depending on siteâspecific characteristics. Significant siteâspecific factors influencing the accuracy of the counts included proper calibration and installation of the technologies. For example, passive infrared sensors are susceptible to false positives when windows, mirrors, or other reflective surfaces are positioned behind the pathway being counted. This is because the surfaces collect heat on sunny days and can mimic the heat signature of humans and trigger false positives for the counter. Similarly, some technologies have a limited detection zone (e.g., a width no greater than 10 feet for some radio beam products, field of detection defined by inductive loop placement) and the installation design becomes particularly important.
NCHRP 07â19(02) Final Report 69 For screenline sensorâbased technologies (e.g., radio beam, passive infrared), occlusion is one of the most significant factors in undercounting. The degree to which occlusion may contribute to undercounting is a factor of pedestrian and bicycle platoons or groups of users traveling sideâbyâ side. For a specific site and time of day, it may be feasible to develop factors that are able to consistently adjust for such an effect. In this study, it was found that nonlinear correction functions improved the fit when adjusting automated counts to reflect ground truth volumes for passive infrared, active infrared, and radio beam sensors, confirming the occlusion effects. In some cases, different products implementing the same sensor technology had significantly different accuracies. This result suggests that a specific vendorâs implementation of a technology (e.g., the algorithm used to decide whether a detection should be registered) can be as important as the technology itself in determining accuracy. This result also indicates that âoneâsizeâfitsâallâ correction factors for particular sensor technologies may not be particularly useful, and that product and siteâspecific factors should be used instead. Given that siteâspecific conditions can also influence accuracy, it is recommended that users develop their own local correction factors for their devices whenever possible. FactorsÂ NotÂ FoundÂ toÂ InfluenceÂ AccuracyÂ Several factors anticipated to affect the accuracy of counting technologies were not evident in the Phase 1 testing. For example, concern has been expressed that the age of inductive loops influences their accuracy. However, the inductive loops tested by this project included loops that were 2 and 3Â½ years old, and the analysis did not indicate diminished quality in those loopsâ counting accuracy. Similar concerns have been expressed related to the age of pneumatic tubes. However, over the sixâ month duration of the testing, count quality was not observed to decline over time. The research team did not find a clear impact or effect of temperature on the accuracy of any of the technologies. The temperatures captured within the duration of this research did not reach the extremes of colds and heat included in other studies; however, for the temperature ranges captured in the research, no impact on the accuracy of the tested devices was observed. Similarly, there was no indicative or quantitative effect found on count accuracy due to snow or rain events. There were limited snow and rain events within the data set, but those that did occur did not appear to influence the quality of the data. Anecdotally, the research team is aware of situations that have occurred with active infrared technologies having a higher rate of false positives during heavy rain events; however, this phenomenon was not observed in this projectâs testing. RecommendationsÂ forÂ PractitionersÂ The research team recommends that practitioners calibrate and conduct their own ground truth count tests for the automated technologies they deploy at a given site or set of sites. This projectâs research results are intended to provide information to practitioners on the types of technologies that may be most promising for a specific circumstance, use, or location where automated count technology is being considered. The projectâs accuracy findings should not be blindly applied to other sites than those at which these technologies were tested at, and it should not be assumed that the same degree of accuracy will occur at other site locations or with other products. Practitioners
NCHRP 07â19(02) Final Report 70 can use the research approach described in this report and accompanying guidebook to, on a smaller scale, test and evaluate the performance of their automated count technologies at their installation sites. It should be noted that when automated counting technologies are used, finding an ideal counting site is just the first step in implementing a count program. As experienced by the research team, getting the roadway or path owner to approve the site can be a significant endeavor. In some cases, the approval of private adjacent developments may also be required. The research team recommends that practitioners consider the time required for obtaining necessary approvals when developing a count program. SUGGESTEDÂ RESEARCHÂ The research team identified several areas for suggested additional research. These areas include additional testing for count technologies not included or underrepresented in the NCHRP 07â19 testing, techniques to develop correction factors for bypass errors, a more robust method for extrapolating shortâduration counts to longer time periods, and adjustment factors to account for changes in pedestrian and bicycle demand (or the potential for demand) based on environmental contexts (e.g., weather, land use characteristics, urban design characteristics). Each of the following is discussed in more detail below. AdditionalÂ TestingÂ ofÂ AutomatedÂ CountingÂ TechnologiesÂ While Phase 2 of this study broadened the range of counting technologies tested, the market for nonmotorized traffic monitoring technologies is constantly evolving and therefore continued testing of new technologies is warranted. For example, fiberâoptic pressure sensors have been incorporated into commercial counting products available in Europe, but were not available in the U.S. market during the Phase 1 and 2 research. Systems that automatically produce nonmotorized counts from recorded video are becoming more prevalent but have not yet been widely tested. In addition to the traditional screenline counting applications tested in this research, this technology offers the potential to produce both directional (e.g., crosswalk) and turning movement/originâdestination counts of pedestrians and bicyclists at intersections. Automated methods also offer the potential for deriving nonmotorized counts from video recorded for other purposes (e.g., from security cameras). One particular challenge with this application is that cameras used for other monitoring purposes have changeable fields of view (i.e., the camera can be rotated, tilted, and zoomed in and out) and therefore the size and visibility of the desired count detection zone also changes. CorrectionÂ FactorsÂ forÂ BypassÂ ErrorsÂ Counters subject to bypass errors (e.g., inductive loops in bicycle lanes) can likely have their raw data adjusted based on site factors to estimate facilityâlevel volumes. NCHRP Project 07â19 simply recommends that practitioners develop these factors on a siteâspecific basis, as insufficient data were available to arrive at general conclusions. However, future projects could likely arrive at
NCHRP 07â19(02) Final Report 71 general findings on the factors that influence bicyclists to bypass counters. Potential topics to look at include facility design, microscopic traffic patterns (e.g., how many bicyclists approach from each direction), and volumes. ExtrapolatingÂ ShortâDurationÂ CountsÂ toÂ LongerâDurationÂ CountsÂ It is recommended that a robust method be investigated for identifying and extrapolating longerâ duration counts from shorterâduration counts at sites where continuous data have not been collected. A potential approach would be to create groups of sites that are considered similar in their pedestrian or bicycle volume peaking characteristics (i.e., factor groups). Current standard guidance is to match shortâterm count sites with continuous count sites in the same factor group to determine the appropriate expansion factors. However, this approach is mostly performed on an ad hoc basis (e.g., central business district vs. multiuse path in suburban context, count program managerâs local knowledge). Potential directions for this research include land useâbased assignments, data fusion with GPS tracking data, or point sampling of very short (e.g. 5â10 minute) counts. AdjustmentÂ FactorsÂ forÂ EnvironmentalÂ FactorsÂ Developing adjustment factors for different environmental factors would enable practitioners to better estimate and predict potential demand on facilities based on changes in land use, design characteristics of the roadway, and how the roadway interfaces with the surrounding land uses (e.g., density of destinations, building setâbacks, parking availability and design). This information would also be useful as input into a methodology for extrapolating shortâduration counts to longerâ duration time periods. Advances in travel demand modeling and discrete choice models to activityâ based modeling techniques are beginning to make progress in the understanding of the potential demand, movement, and mode of person trips. In addition to the continued development of activityâ based model methods, a sketchâplanning tool is also needed for practitioners to use to understand reasonable expected changes in bicycle and pedestrian traffic volumes due to changes in road, multiuse path, and land use characteristics.