Intelligent Vehicle Road Recognition And Early Warning System Based On Binocular Vision

With the increasing development of global science technology and economyscale, the automotive industry is also growing, along with the continuous increasing inthe number of per capita holdings of automobile. It has caused great problems, themost serious of which is frequently happened traffic accidents. Automotive safetytechnology based research also emerges as the times require, driving safety assistancesystem of which has become a hot research topic.As one of the assistant safe driving system, road warning system based binocularvision has become the hot spot. By processing the information gathered from thecamera on the car, whether there is a vehicle in front of the current vehicle can bejudged. Combining with car’s speed and road conditions, corresponding measurescould be taken to guarantee the safety of the driver during the driving the process,which could effectively reduce the car collisions, rear-end and other accidents. Thispaper has carried on the following work:(1) Hardware acquisition platform. Binocular vision acquisition platform consistsof a tripod, two CCD and two Easycape image acquisition cards. Through thisplatform, a lot of campus road pictures can be collected and transmitted to the hostcomputer for processing.(2) Image processing and research on the detection and algorithm of lane andobstacle. First of all road image should be preprocessed, including gray processing,filter, image binarization and edge detection, and then improved Hough transform canbe used to detect the carriageway and identify two carriageways. So an area of interestformed by both of carriageways could be found. The shaded portion of the targetvehicle can be identified by histogram thresholding method, which can be marked asthe location of the target vehicle.(3) Research on fault location algorithm based on binocular. The distancebetween target vehicle and platform can be calculated by binocular parallax transverse mode. Through the integration of monocular calculated relative depth difference, theoriginal algorithm of binocular distance measurement can be improved. Theexperimental data show the improved location algorithm fully can reduce the errorrate ranging and improve the execution efficiency of the program effectively.(4) Construction and research early warning system. First, a safe distance modeland the permanent magnet synchronous motor model is built. Combining with thecorresponding vehicle control strategies, early warning system is built inMATLAB/Simulink simulation. Experiment results show that this system hasachieved the desired effect.