Grouping Intervals Sample Clauses

Grouping Intervals. The design of our speed estimation algorithm is based on a revised “region growing” concept. Region growing is a technique traditionally used in image segmentation applications that allows computer vision systems to separate areas of an image into regions depending on criteria of interest, such as color or texture. Xxxxxxx and Xxxxxxxx (2001) stated that “a region grower begins at a position in an image and attempts to grow each region until the pixels being compared are too dissimilar to the region to add them.” The position at which region growing begins is known as the seed for the algorithm. This same idea can be used to discern between intervals containing only SVs and those containing LVs. One characteristic of region growing is that the statistics used for determining membership in a region are updated each time a new member is added to the region. Applying this concept to speed estimation, we first group m 20-second intervals into a period of length m/3 minutes (for example, if m=15 intervals, then the length of a period is 5 minutes). All the m interval data are then processed simultaneously to identify intervals with only SVs. Intervals with no vehicles present are eliminated from the analysis. The occupancy per vehicle (O/V) is calculated for each remaining interval, and the periods are sorted in order of ascending occupancy per vehicle to prepare for region growing. Once the periods are sorted, it is assumed that the smallest non-zero O/V value consists of only SVs, which will serve as the seed for the region growing algorithm. Xxxx and Xxxxx (2003) found that an assumption that the smallest two non-zero O/V intervals contain only SVs was violated less than 3 percent of the time when 5-minute periods with the typical traffic composition (about 10 percent LVs) on I-5 were used. Therefore, the assumption that only the first interval contains only SVs would be violated even less frequently. The group occupancy per vehicle (GOV) is calculated by using the occupancy and volume measurements for all intervals already identified as being in the group (i.e., let interval x be the last identified interval in the group): (i) x (i) i=1 Then the occupancy per vehicle ratio (ΔO/V) of interval x+1 is calculated: ΔO / V = O(x + 1) /V (x + 1)