Vehicle-type identification through automated virtual loop assignment and block-based direction biased motion estimation

摘要: This paper presents the concept of automated virtual loop assignment and loop-based motion estimation in vehicle-type identification. A major departure our method from previous approaches is that loops are automatically assigned to each lane; size much smaller for accuracy; number per lane large. Comparing this with traditional ILD, there a advantages. First, may be varied fine-tune detection accuracy fully utilize computing resources. Second, no failure rate associated installation maintenance cost can kept minimum. Third, re-allocated anywhere on frame, giving flexibility detecting different parameters.