摘要
为了提高机器人在环境识别中的实时性和准确性,在分析了SIFT描述算子性能及原理的基础上,对SIFT算法在计算描述算子阶段提出自适应修改步长的改进,以及在原先的匹配方法的基础上又加入二次反向匹配,既在一定程度上提高了实时性又对关键点的消除错配起到较好的效果。仿真结果也证明了其有效性。
In order to improve the real-time and accuracy ot environment recognition for robot, improving on sampling step in the phase of calculating the descriptors and adding Second reverse match to matching method on the basis of anglicizing performance and principle of SIFT descriptors, which satisfy requirement of enhancing speed and accuracy. The simulations also prove its sufficiency.
