一种阈值动态调整的仿生同步自主定位方法
Independent Location and Mapping Method with Bionic Synchronization Based on Dynamic Threshold

针对仿生导航中鼠类同步定位与构图(RatSLAM)算法对连续经历场景出现漏匹配和错误匹配的问题,提出一种阈值动态调整的仿生同步自主定位与构图(DT？？RatSLAM)方法。该方法在局部视图细胞生成阶段,引入阈值增强项和阈值衰减项,将前几帧图像的识别结果作为激励,实时动态地调整当前帧图像的模板识别阈值,并根据阈值判断是否生成新的局部视图细胞,然后综合连续场景的模板识别结果,对经验图进行闭环修正。仿真实验结果表明,与固定阈值条件下的RatSLAM算法相比,基于阈值动态调整的DT？？RatSLAM方法在模板识别方面不仅能够保证不出现错误的匹配,而且正确识别率达到了99.6％,促使运行体能够对生成的经验图及时进行修正,提高了经验图的准确性,在算法运行时间方面比最优固定阈值的RatSLAM算法平均减少了13.5％。
An independent location and mapping method with bionic synchronization based on the dynamic threshold (DT？？RatSLAM) is proposed to solve the problem that there exist missing matching and false matching in the model of the rat simultaneous location and mapping (RatSLAM) for the continuous experienced scenes in the bio？？inspired navigation. An enhanced threshold factor and an attenuated threshold factor are introduced in the stage of forming local view cells. The recognition results for the first few frames are used as the excitation to dynamically adjust the template recognition threshold of the current frame image in real time, and to judge whether new local view cells need to be generated according to the threshold. Then the experience map is corrected through closed loop detection by integrating the recognition results for the continuous experienced scenes. Experimental results and a comparison with the method under the fixed threshold show that the proposed method not only avoids false matching, but also makes the correct recognition rate reach 99.6％ in the aspcet of template recognition, which urges the agent to modify the experience map in a timely manner and to improve the accuracy of the experience map. Moreover, the running time of the proposed method is also reduced by 13.5％