Optimization and Design of Driverless Vehicle Software System Based on Image Recognition

An intelligent vehicle is a system that integrates environment perception, path decision planning, automatic driving, and other functions. In order to improve the tracking and motion performance of intelligent vehicles, a system that includes image preprocessing, image processing, path tracking planning, and intelligent vehicle control is designed and optimized. First, the principles and implementation effects of three basic threshold algorithms and the image denoising algorithm are discussed. Second, the traditional edge extraction algorithm and track condition judgment algorithm are improved. Then, a path tracking planning method based on the midline algorithm and an edge fitting algorithm based on the least square algorithm are proposed and simplified. Finally, aiming at solving the shortcomings of the traditional PID algorithm that cannot update the values of K_p, K_i and K_d, an intelligent vehicle control system based on the PID algorithm and fuzzy control is proposed and verified by simulation and experiment. The results show that the designed filtering algorithm can effectively reduce the image noise. The improved edge extraction algorithm has an obvious filtering effect on the abnormal data in the process of intelligent vehicle operation. The difference between the straight and bent track obtained by the improved track condition judgment algorithm is 7.39, which is larger than 1.78 obtained by the traditional algorithm. The improved algorithm is sensitive to the change in the track bending degree and overcomes the problem that the performance of the traditional algorithm decreases with the bending degree. Using the simplified edge algorithm, an edge fitting algorithm based on the least square algorithm is developed. This algorithm is similar to the algorithm with the R-squared greater than 0.994, and the number of edge points used for calculation is reduced from the original 48 points to 2 or 3 points, which greatly improves the operation efficiency of the intelligent vehicle. Using the fuzzy-based PID control algorithm, after the target speed changes, the output curve reaches the target speed at 0.44 s, the maximum excess is approximately 16.4 rpm, and the algorithm becomes stable at the target speed after 7.9 s, which is less than 1.2 s, 56.7 rpm and 12.1 s of traditional PID control respectively. Thus, using the proposed fuzzy-based PID control algorithm, the control performance of the system can be significantly improved. The experimental results of real intelligent vehicle show that the proposed fuzzy PID control algorithm can significantly improve the control effect under high-speed operation.

關鍵字

Intelligent vehicle ； Image preprocessing ； Edge extraction algorithm ； Track condition judgment algorithm ； Path tracking planning ； Fuzzy-based PID control

並列摘要

智能車輛是集環境感知、路徑決策規劃、自動駕駛等功能於一體的系統。為提高智能車輛的跟蹤和運動性能，設計並優化了包括圖像預處理、圖像處理、路徑跟蹤規劃和運動控製在內的智能車輛系統。首先，討論了三種基本閾值算法和圖像去噪算法的原理和實現效果。其次，對傳統的邊緣提取算法和路況識別算法進行了改進。然後，提出並簡化了基於中線算法和基於最小二乘算法的邊緣擬合算法的路徑跟蹤規劃方法。最後，針對傳統PID算法無法更新K_p、K_i和K_d值的缺點，提出了一種基於PID算法和模糊控製的智能車輛控製系統。實驗結果表明，所提出的濾波算法能有效地降低圖像噪聲。改進的邊緣提取算法對智能車運行過程中的異常數據具有明顯的濾波效果。改進的路況識別算法得到的直、彎路況差值為7.39，大於傳統算法得到的1.78，改進算法對軌道彎曲度變化敏感，克服了傳統算法性能隨彎曲度變化而下降的問題。在簡化邊緣算法的基礎上，提出了一種基於最小二乘算法的邊緣擬合算法，該算法與R^2大於0.994的算法相似，計算所用的邊緣點數量由原來的48個點減少到2或3個點，大大提高了智能車的運行效率。使用模糊PID控製，在目標速度的變化後，輸出曲線在0.44s達到目標速度，最大超出量約16.4rpm，目標速度在7.9秒後穩定，分別小於傳統PID的1.2s, 56.7rpm, 12.1s。因此，采用所提出的基於模糊PID控製算法，可以顯著提高系統的控製性能。真實智能車的實驗結果表明，所提出的模糊PID控製算法能顯著提高高速運行時的控製效果。

並列關鍵字

無資料

主題瀏覽