At present, most underwater positioning algorithms improve the positioning accuracy by increasing the number of anchor nodes which resulting in the increasing energy consumption. To solve this problem, the paper propo...At present, most underwater positioning algorithms improve the positioning accuracy by increasing the number of anchor nodes which resulting in the increasing energy consumption. To solve this problem, the paper proposes a localization algorithm assisted by mobile anchor node and based on region determination(LMRD), which not only improves the positioning accuracy of nodes positioning but also reduces the energy consumption. This algorithm is divided into two stages: region determination stage and location positioning stage. In the region determination stage, the target region is divided into several sub-regions by the region division strategy with the smallest overlap rate which can reduce the number of virtual anchor nodes and lock the target node to a sub-region, and then through the planning of mobile nodes to optimize the travel path, reduce the moving distance, and reduce system energy consumption. In the location positioning stage, the target node location can be calculated using the HILBERT path planning and trilateration. The simulation results show that the proposed algorithm can improve the positioning accuracy when the energy consumption is reduced.展开更多
孪生区域提议网络跟踪算法是一种高效的目标跟踪算法,通过锚框规避了图像金字塔对跟踪性能带来的影响,但这种跟踪方法受制于区域提议网络本身的局限性,在目标旋转时,跟踪精度将受到较大损失。而其他对旋转鲁棒性较高的方法则因为使用了...孪生区域提议网络跟踪算法是一种高效的目标跟踪算法,通过锚框规避了图像金字塔对跟踪性能带来的影响,但这种跟踪方法受制于区域提议网络本身的局限性,在目标旋转时,跟踪精度将受到较大损失。而其他对旋转鲁棒性较高的方法则因为使用了复杂的旋转结构,导致算法的跟踪速度大幅下降。为了解决旋转目标对区域提议网络跟踪精度的影响,提出了旋转区域提议网络的孪生神经网络跟踪算法,通过AO-RPN(arbitrary-oriented region proposal network)结构将旋转与区域提议网络相统一,引入角度预测分支,在目标跟踪的过程中,直接对旋转的目标进行搜索,并得到最小外接矩形。该方法在保持较高跟踪速度的同时,精度超过了对目标进行旋转采样或使用局部特征进行跟踪的算法。通过在数据集OTB2015、VOT2016和VOT2018上进行的大量实验。结果表明,该算法在遮挡、形变、光照等多种复杂情况下表现出了较强的鲁棒性和适应性。展开更多
基金supported by National Natural Science Foundation of China (Nos. U1806201, 61671261)Key Research and Development Program of Shandong Province (No. 2016GGX101007)+1 种基金China Postdoctoral Science Foundation (No. 2017T100490)University Science and Technology Planning Project of Shandong Province (Nos. J17KA058, J17KB154)
文摘At present, most underwater positioning algorithms improve the positioning accuracy by increasing the number of anchor nodes which resulting in the increasing energy consumption. To solve this problem, the paper proposes a localization algorithm assisted by mobile anchor node and based on region determination(LMRD), which not only improves the positioning accuracy of nodes positioning but also reduces the energy consumption. This algorithm is divided into two stages: region determination stage and location positioning stage. In the region determination stage, the target region is divided into several sub-regions by the region division strategy with the smallest overlap rate which can reduce the number of virtual anchor nodes and lock the target node to a sub-region, and then through the planning of mobile nodes to optimize the travel path, reduce the moving distance, and reduce system energy consumption. In the location positioning stage, the target node location can be calculated using the HILBERT path planning and trilateration. The simulation results show that the proposed algorithm can improve the positioning accuracy when the energy consumption is reduced.
文摘孪生区域提议网络跟踪算法是一种高效的目标跟踪算法,通过锚框规避了图像金字塔对跟踪性能带来的影响,但这种跟踪方法受制于区域提议网络本身的局限性,在目标旋转时,跟踪精度将受到较大损失。而其他对旋转鲁棒性较高的方法则因为使用了复杂的旋转结构,导致算法的跟踪速度大幅下降。为了解决旋转目标对区域提议网络跟踪精度的影响,提出了旋转区域提议网络的孪生神经网络跟踪算法,通过AO-RPN(arbitrary-oriented region proposal network)结构将旋转与区域提议网络相统一,引入角度预测分支,在目标跟踪的过程中,直接对旋转的目标进行搜索,并得到最小外接矩形。该方法在保持较高跟踪速度的同时,精度超过了对目标进行旋转采样或使用局部特征进行跟踪的算法。通过在数据集OTB2015、VOT2016和VOT2018上进行的大量实验。结果表明,该算法在遮挡、形变、光照等多种复杂情况下表现出了较强的鲁棒性和适应性。
