针对在动态射频识别(Radio Frequency Identification,RFID)室内定位环境中,传统的室内定位模型会随着定位目标数量的增加而导致定位误差增大、计算复杂度上升的问题,文中提出了一种基于近端策略优化(Proximal Policy Optimization,PPO)的RFID室内定位算法。该算法将室内定位过程看作马尔可夫决策过程,首先将动作评价与随机动作相结合,然后进一步最大化动作回报值,最后选择最优坐标值。其同时引入剪切概率比,首先将动作限制在一定范围内,交替使用采样后与采样前的新旧动作,然后使用随机梯度对多个时期的动作策略进行小批量更新,并使用评价网络对动作进行评估,最后通过训练得到PPO定位模型。该算法在有效减少定位误差、提高定位效率的同时,具备更快的收敛速度,特别是在处理大量定位目标时,可大大降低计算复杂度。实验结果表明,本文提出的算法与其他的RFID室内定位算法(如 Twin Delayed Deep Deterministic Policy Gradient(TD3),Deep Deterministic Policy Gradient(DDPG),Actor Critic using Kronecker-Factored Trust Region(ACKTR))相比,定位平均误差分别下降了36.361%,30.696%,28.167%,定位稳定性分别提高了46.691%,34.926%,16.911%,计算复杂度分别降低了84.782%7,70.213%,63.158%。

In the Radio Frequency Identification(RFID) dynamic indoor positioning environment,the positioning error and the computing complexity of traditional indoor positioning model will increase with the increase of the number of positioning targets.This paper proposes an RFID positioning algorithm based on Proximal Policy Optimization(PPO),which regards the positioning as Markov decision-making process.Firstly,the action evalution is combined with random action and the return of the action is then maximized to select the best coordinate value.Meanwhile,under the premise of limiting the action to a certain range,the algorithm introduces clipped probability ratios,using post-sample and pre-sample action alternatesly,then,with stochastic gradient ascent updates multiple epochs policy of minibatch and with the critic network evaluate the action.Finally,the PPO positioning model is obtained by training.This method can effectively reduce the positioning error and improve the positioning efficiency.At the same time,it has a faster convergence speed,especially when dealing with a large number of positioning targets,it can greatly reduce the computational complexity.Experiment results show that,compared with other RFID indoor positioning algorithms,such as Twin Delayed Deep Deterministic policy gradient(TD3),Deep Deterministic Policy Gradient(DDPG) and actor-critic using Kronecker-Factored Trust Region(ACKTR),the average positioning error of the proposed method decreases respectively by 36.361%,30.696% and 28.167%,the positioning stability improves by 46.691%,34.926% and 16.911%,and the computing complexity decreases respectively by 84.782%,70.213% and 63.158%.

Multispectral imaging (MSI) has become a new fast and non-destructive detection method in seed identification. Previous research has usually focused on single models in MSI data analysis, which always employed all features and increased the risk to efficiency and that of system cost. In this study, we developed a stacking ensemble learning (SEL) model for successfully identifying a single seed of sickle alfalfa (Medicago falcata), hybrid alfalfa (M. varia), and alfalfa (M. sativa). SEL adopted a three-layer structure, i.e., level 0 with principal component analysis (PCA), linear discriminant analysis (LDA), and quadratic discriminant analysis (QDA) as models of dimensionality reduction and feature extraction (DRFE); level 1 with support vector machine (SVM), multiple logistic regression (MLR), generalized linear models with elastic net regularization (GLMNET), and eXtreme Gradient Boosting (XGBoost) as basic learners; and level 3 with XGBoost as meta-learner. We confirmed that the values of overall accuracy, kappa, precision, sensitivity, specificity, and sensitivity in the SEL model were all significantly higher than those in basic models alone, based on both spectral features and a combination of morphological and spectral features. Furthermore, we also developed a feature filtering process and successfully selected 5 optimal features out of 33 ones, which corresponded to the contents of chlorophyll, anthocyanin, fat, and moisture in seeds. Our SEL model in MSI data analysis provided a new way for seed identification, and the feature filter process potentially could be used widely for development of a low-cost and narrow-channel sensor.