Front-vehicle detection and tracking method based on monocular foresight camera

Abstract

The invention discloses a front-vehicle detection and tracking method based on a monocular foresight camera. The method comprises following steps of S1: according to prior knowledge, determining a vehicle detection ROI region and a sliding window searching scale range in the vehicle detection ROI region, removing regions without remarkable texture characteristics in the vehicle detection ROI region and by use of a cascading classifier, searching the position of a vehicle in a specific scale range; S2: extracting angle point characteristics of the vehicle right in front of the own vehicle in a vehicle image candidate region, using an optical flow algorithm to track the series of angle points so as to tracking the position of the front vehicle, and tracking the position of the vehicle which is not right in front of the own vehicle by use of Kalman filter based on a kinetic model; and S3: according to a vehicle position detection result and a vehicle position tracking result, calculating the confidence of the position of the front vehicle at this moment. According to the invention, searching range can be reduced to the greatest degree; searching efficiency is improved; and rapid and effective tracking of vehicles in the full distance range is effectively achieved.

Description

technical field [0001] The invention relates to the technical field of vehicle detection, in particular to a method for detecting and tracking a preceding vehicle based on a single forward-looking camera. Background technique [0002] Advanced Driver Assistance System (ADAS) is one of the important technologies for the development of vehicle active safety, and has gradually become an important reference for NCAP in various countries to evaluate vehicle safety levels. ADAS functions such as Forward Collision Warning (FCW) and Lane Departure Warning (LDW) have become the standard configuration of existing mid-to-high-end models. For ADAS systems targeting vehicles in front, such as FCW and automatic emergency braking system (AEB), not only the perception system is required to detect the vehicle in front in a timely and accurate manner, but also the system is required to accurately measure and predict the movement state of the vehicle in front relative to its own vehicle. , to...

AI Technical Summary

Problems solved by technology

This type of method takes a long time to search, and cannot detect and track vehicles with incomplete contours at close range

Images