Abstract: In order to solve the problems of insufficient perception image processing accuracy and poor anti-interference ability of path recognition in complex road environment, this paper designs a multi-stage optimization method that integrates the Otsu's method, the eight-neighborhood boundary tracking algorithm and the median filter, and verifies the effectiveness of the scheme by MATLAB simulation. The proposed method achieves systematic improvement through a three-level processing mechanism: the original Otsu's method with dynamic threshold adjustment is used to improve the image segmentation accuracy under complex lighting, the path contour features are extracted by the eight-domain boundary tracking algorithm, and the path topology optimization and noise suppression are realized by combining median filtering, so as to improve the anti-interference ability of the system as a whole. Experiments show that the proposed method reduces the standard deviation of path coordinates by about 50%, which not only improves the image processing accuracy, but also significantly strengthens the robustness of path recognition, thereby improving the tracking reliability of the intelligent car.

Key words: image processings; path recognition; MATLAB; Otsu's method; eight-neighborhood algorithm

摘要： 针对智能小车在复杂道路环境中存在的感知图像处理精度不足与路径识别抗干扰能力 较差的问题，文章设计了一种融合大津法、八邻域边界追踪算法及中值滤波的多级优化方法， 并通过 MATLAB 仿真验证方案有效性。所提方法通过三级处理机制实现系统性改进：采用动 态阈值调整的原始大津法提升复杂光照下的图像分割精度，利用八邻域边界追踪算法对路径 轮廓特征进行提取，结合中值滤波实现路径拓扑优化和噪声抑制，提升系统整体的抗干扰能 力。实验表明，该方法使路径坐标标准差降低约 50%，在提升图像处理精度的同时也显著强 化了路径识别的鲁棒性，进而提高了智能小车的循迹可靠性。

关键词: 图像处理；路径识别；MATLAB；大津法；八邻域算法