Design And Application Of Portable Nuclear Detection System Based On Net Of Things

With the rapid development of economy, the automobile has become one of the indispensable transports in the modern society. To solve the heavy road traffic problems, many studies have been done on intelligent transportation systems and V2V technology. Lane-level positioning of the vehicle is one of the most popular researches. Along with the deep analysis of imaging theory and improvement of vision detection method, research on lane-level positioning by monocular vision is carried out. The main research contents and achievements are summarized as follows:(1) Make a deep research on image pre-processing method for lane detection. As Red color and Green color account for a larger proportion of road line, R and G from RGB are selected to extract the grayscale image. The first Region of Interest (ROI) is below the line where gray value steps in vertical direction. Followed by median filter, a new edge feature points detection method based on Sobel is proposed. A new iterative method is also used in image thresholding. In order to eliminate disturbance information and strengthen the lane edge feature, researches on edge skeleton extraction according to the gradient direction and rain denoising with the shift subtraction method are carried out.(2) Make a research on lane-level positioning of vehicle based on multi-lane detection. After Hough Transform is used to extract the possible lane lines on the road, the least squares fitting method is used to fit different combinations of continuous edge. Features of line parameter and location, color contrast and edge linearity are used to correct the lines. Experiment results demonstrate that this proposal reduces the false detection rate and improves the accuracy of lane detection. Through the statistics and clustering from numbers of image frames, left and right lane lines can be easily distinguished and the vehicle’s lane-level location can be got by the result.(3) Make a research on lane change tracking. Near-field road lane lines can be tracked in a dynamic ROI extracted form Kalman predictor with high robustness. Change of the line angle is used to monitor whether the vehicle is departing or changing the driving lane. Parallel Perspective Projection Model which is used to determine the new lane line parameters makes the system real-time and continuous for vehicle’s lane-level positioning.