基于分段切换模型的点集优化车道线识别方法

针对车辆的横向运动偏差及相关安全监控技术存在缺陷的问题,提出一种基于分段切换模型的点集优化车道线识别方法。首先建立稳定可靠的分段切换车道线模型,然后利用单目视觉和毫米波雷达的传感器组合对车辆进行识别,筛选掉干扰车道线识别的无效候选点,如果缺少有效的识别点,则利用卡尔曼滤波和车道线等宽条件进行有效点的补充。实验证明,该方法能够准确有效的识别车道线,对于减少车辆行驶过程中安全事故的发生具有重要意义。

基于锚点集优化的硅片图像缺陷检测方法

为有效解决机器视觉算法识别硅片图像缺陷时存在检测速度慢、准确率低和稳定性差等问题,提出一种锚点集的优化方法。将Faster R-CNN检测网络与ResNet相结合进行特征提取;针对硅片缺陷数据集的特点,从理论上通过期望的定位精度推导一种优化的锚点集;利用优化后的锚点集训练缺陷检测模型。实验结果表明,所提检测模型识别精度提升至94.1%,检测时间达到160 ms,满足工业检测的需求。

Immunity clone algorithm with particle swarm evolution

Combining the clonal selection mechanism of the immune system with the evolution equations of particle swarm optimization, an advanced algorithm was introduced for functions optimization. The advantages of this algorithm lies in two aspects. Via immunity operation, the diversity of the antibodies was maintained, and the speed of convergent was improved by using particle swarm evolution equations. Simulation programme and three functions were used to check the effect of the algorithm. The advanced algorithm were compared with clonal selection algorithm and particle swarm algorithm. The results show that this advanced algorithm can converge to the global optimum at a great rate in a given range, the performance of optimization is improved effectively.

Ekeland变分原理及其应用(英文)

本文将Ekeland变分原理作了推广 。