Autonomous Car Steering Control and Sign Detection Utilizing Haar Cascade and PID

This study applies methods of Image processing, Haar Cascade, and PID in building a system on an autonomous car. The working system on an autonomous car has currently gained interest for the exploration. The Image processing method is utilized to detect road markings applying algorithms (BGR to RGB, Perspective transformation, threshold, canny edge, and histogram) in navigating pixel parameter values with actual distances. The parameter values in the line detection results are applied for input to the PID system, in controlling the handling of autonomous car of five seconds settling time and 1.5 cm overshoot. On the other hand, the Haar Cascade image training requires the positive (1000) and negative (1400) sample images to obtain a statistical model. Regarding the image processing test results, the correlation between the distance of the road markings and the camera pixels is 1 cm/10 pixels. This correlation hence demonstrates an accuracy of 0.1cm/pixel. The statistic test results for detecting traffic signs are obtained at a distance of 5 cm to 120 cm, with the optimal detection results at a distance of 5 cm to 90 cm. Regarding the traffic signs’ point of view at an angle of ��� to ����� optimal detection results are obtained at an angle of ��� to ����. Meanwhile, when testing dynamically, the detection performance depends on the angle and distance of the reading as on the results of the static test and the appropriate lighting factor on the results of training sample data on traffic signs. In addition, the results of the average accuracy of traffic sign detection are 92.3%