研究车祸中严重碰撞车辆图像边界准确分割问题.车祸中,如果发生碰撞较为严重,两车图像碰撞部位交汇的像素分布较为密集,像素会产生变异.传统的边沿检测算法多是基于像素差异进行边界分割,当车祸中像素密度分布密集导致像素变异的情况,会造成像素聚类效果不好,分割不完整,分割的准确性不高.提出了一种基于密度分布函数的车祸图像边界检测算法.算法通过计算数据车祸像素邻域半径内每个像素点对它的高斯影响函数之和,将其作为该像素对象的密度,再通过变异系数刻画像素对象密度分布特征从而提取车祸图像边界点.实验结果表明,算法提高了车祸图像边界分割的准确度.%Research accurate segmentation of vehicle image boundary with serious collision. This paper put forward an image edge segmentation algorithm based on density distribution function. It calculates the sum of Gaussian influence function of each pixel point within the neighborhood radiu, and takes it as the object pixel density to extract the boundary point of the accident image. The experiment results show that the algorithm can improve the accuracy of segmentation.
展开▼
