[Retracted] Self-Correction Ship Tracking and Counting with Variable Time Window Based on YOLOv3

Table 1

The process of this tracking algorithm.

---

Tracking algorithm

---

(1) Determine the tracking target in the first frame of the image, set the target as a template, and obtain its HSV and LBP feature vectors

(2) Continue target detection and obtain a number of targets to be selected

(3) Obtain the HSV and LBP features of each target to be selected, calculate their similarity rates with the tracking target in turn, and determine a candidate target with the highest similarity rate

(4) Use a variable time window self-correction model to avoid count repetition, misjudgment of the direction, and jitter

---