Research On Vision-based Surveillance System Of Vehicles For Violation Of Traffic Line And Illegally Parked

Detection and Capture of vehicles for violation is a an important part of intelligent transportationsystems, it played a major role in constraining traveler’s behavior, reducing vehicle’s violation andtraffic accidents, improving traffic efficiency, etc. We design a system to monitoring and record thevehicles which violation of traffic line or illegally parked.First, according to the request of the regulatory system, the hardware architectures and softwareprocess was presented. And describes the operating principle. The system in this paper has largemonitoring scope and high robustness.Second, in view of the issue that most high speed dome camera have no angle feedback, we addthe angle encoder into the pan-tilt holder’s shaft to build the closed-loop system and investigate thedata collection and transmission base on ARM chip. Base on the feature of the camera, we build asnapshot measurement and control model to establish the correspondences between the pan-tilt holderand the target’s location information.Then, in order to extract vehicles from road scene image, we studied the non-parametric andparametric background model for foreground object detection. Through experiment analyzed thequality of these algorithm. We combined the history pixel sequence and ViBe to build the backgroundmodel. Comparison test confirmed that our algorithm had advantages in detection rate and falsepositive rate.In addition, the current image measurement algorithm fussy operation, we design the vertical(direction of the lane) and (vertical direction of the lane) horizontal ranging algorithm, base on thecamera imaging principle and width line calibration. According to the experiment, the accuracy ofmeasurement meets the requirement of snapshot.Last, in the large monitoring range traffic system the camera need the help of pan-tilt holder toaim the target, and capture the HD graphics. So we design the formula to calculate the parameterssuch as pan, tilt and focus with the snapshot measurement and control model. It can predictive thevehicles’position and control the pan-tilt to snapshot the picture.In order to detect the vehicles crossed line we design a robust algorithm, then capture it base onlocation predictive. In order to detect illegally parking, we design the snapshot scheme to capture it.The system in this paper was test in the real traffic environment, and obtained satisfactory results.