基于深度学习的翻越行为检测

翻越行为检测对疫情管控、社会治安等有着重要意义,一定程度上可以减少因为违规翻越行为而造成的意外事故.针对目前翻越行为检测任务实时性差、需要先验知识的问题,本文将Faster RCNN+SlowFast时空行为检测算法应用在翻越行为检测任务中,将翻越行为进行拆分检测.为提高时空行为检测算法中目标的检测精度和速率将目标检测模块Faster RCNN改为实时性高且轻量化的YOLOv5;其次针对同一行为不同视角下广泛的类内多样性的问题,将Fast支路和Slow支路的residual block分别改为AC residual block和SE residual block来加强网络对关键特征与细粒度特征的学习能力,最后设计翻越行为检测算法进行攀爬与下降两种状态的连续性检测,实验结果显示该网络平均准确率达93.5%,在翻越行为检测中表现出良好的性能.     

特征挖掘与区域增强的弱监督时序动作定位

弱监督时序动作定位旨在定位视频中行为实例的起止边界及识别相应的行为。现有方法尽管取得很大进展,但依然存在动作定位不完整及短动作的漏检问题。为此,提出了特征挖掘与区域增强（FMRE）的定位方法。首先,通过基础分支计算视频片段之间的相似分数,并以此分数聚合上下文信息,得到更具有区别性的段分类分数,实现动作的完整定位;然后,添加增强分支,对基础分支定位中持续时间较短的动作提案沿时间维度进行动态上采样,进而采用多头自注意机制对动作提案间的时间结构显式建模,促进具有时间依赖关系的动作定位且防止短动作的漏检;最后,在两个分支之间构建伪标签互监督,逐步改进在训练过程中生成动作提案的质量。该算法在THUMOS14和ActivityNet1.3数据集上分别取得了70.3%和40.7%的检测性能,证明了所提算法的有效性。     

Weakly supervised temporal action localization with proxy metric modeling

Temporal localization is crucial for action video recognition. Since the manual annotations are expensive and time-consuming in videos, temporal localization with weak video-level labels is challenging but indispensable. In this paper, we propose a weakly-supervised temporal action localization approach in untrimmed videos. To settle this issue, we train the model based on the proxies of each action class. The proxies are used to measure the distances between action segments and different original action features. We use a proxy-based metric to cluster the same actions together and separate actions from backgrounds. Compared with state-of-the-art methods, our method achieved competitive results on the THUMOS14 and ActivityNet1.2 datasets.     

基于轻量级图卷积的人体骨架动作识别方法

视频中的人体动作识别在计算机视觉领域得到广泛关注,基于人体骨架的动作识别方法可以明确地表现人体动作,因此已逐渐成为该领域的重要研究方向之一。针对多数主流人体动作识别方法网络参数量大、计算复杂度高等问题,设计一种融合多流数据的轻量级图卷积网络,并将其应用于人体骨架动作识别任务。在数据预处理阶段,利用多流数据融合方法对4种特征数据流进行融合,通过一次训练就可得到最优结果,从而降低网络参数量。设计基于图卷积网络的非局部网络模块,以捕获图像的全局信息从而提高动作识别准确率。在此基础上,设计空间Ghost图卷积模块和时间Ghost图卷积模块,从网络结构上进一步降低网络参数量。在动作识别数据集NTU60 RGB+D和NTU120 RGB+D上进行实验,结果表明,与近年主流动作识别方法ST-GCN、2s AS-GCN、2s AGCN等相比,基于该轻量级图卷积网络的人体骨架动作识别方法在保持较低网络参数量的情况下能够取得较高的识别准确率。     

雷击对碳纤维增强型航空复合材料损伤的影响

依据有限元仿真软件ABAQUS,建立碳纤维增强型复合材料(CFRP)层合板雷击损伤热-电耦合有限元仿真模型。利用叠加温度场的方法来近似表示内部受损状态,通过对比实验验证热-电耦合仿真方法的正确性与有效性。利用回归统计分析技术,定量分析雷电参数与CFRP雷击损伤的相关性,并绘制相关曲线。结果表明:雷电流比能是决定CFRP雷击损伤的关键因素,纤维破坏面积、分层面积与比能具有线性相关性,树脂破坏面积、分层厚度与比能具有对数相关性。     

不同炮孔堵塞材料对矿山露天爆破效果的影响

通过矿山目前采用的单一材料堵塞的冲孔分析,利用控制变量法和对比分析法,定性地研究炮孔中应力和时间的变化规律,结合现场爆破试验,将炮孔堵塞分6种情况进行试验研究,最终得出:混合材料的炮孔堵塞相比单一材料的炮孔堵塞,具有抗剪强度高、炸药能量在炮孔中作用时间长和炸药做功能力增大等特点。现场混合材料堵塞能有效地降低冲孔现象的发生,提高炸药利用率和减少炮孔上部大块的产生。     

一种基于分层的人类动作识别和定位算法研究

人类动作识别在视频自动分析、视频检索等领域获得广泛应用,是目前的研究热点。然而现有的动作识别方法重点关注视频的非静态部分而忽略大部分静态部分,从而影响了动作识别和定位的效果。本文提出一种新的分层空间-时间分段表示法,以分层方式实现部位和整个身体的多分辨率表示,可用于运动识别和定位。该算法分为3个步骤。第一步,首先对每个视频帧进行分层分段,以得到一组分段树,每颗树是身体分段树的候选。第二步,利用视频的轮廓、接合对象结构、全局前景色等信息对候选分段树进行修剪。第三步,在时域上对剩余分段层的每个分段进行前向和后向跟踪。我们以难度较大的UCF-Sports和HighFive数据集为实验对象,对本文方法进行性能评估,实验结果表明,本文方法的性能要优于当前最新运动检测算法性能,运动定位性能与当前最新算法相当。     

Feature Fusion-Based Deep Learning Network to Recognize Table Tennis Actions

A system for classifying four basic table tennis strokes using wearable devices and deep learning networks is proposed in this study. The wearable device consisted of a six-axis sensor, Raspberry Pi 3, and a power bank. Multiple kernel sizes were used in convolutional neural network (CNN) to evaluate their performance for extracting features. Moreover, a multiscale CNN with two kernel sizes was used to perform feature fusion at different scales in a concatenated manner. The CNN achieved recognition of the four table tennis strokes. Experimental data were obtained from 20 research participants who wore sensors on the back of their hands while performing the four table tennis strokes in a laboratory environment. The data were collected to verify the performance of the proposed models for wearable devices. Finally, the sensor and multi-scale CNN designed in this study achieved accuracy and F1 scores of 99.58% and 99.16%, respectively, for the four strokes. The accuracy for five-fold cross validation was 99.87%. This result also shows that the multi-scale convolutional neural network has better robustness after five-fold cross validation.     

Reasoning about action and change

Reasoning about change is a central issue in research on human and robot planning. We study an approach to reasoning about action and change in a dynamic logic setting and provide a solution to problems which are related to the Frame problem. Unlike most work on the frame problem the logic described in this paper is monotonic. It (implicitly) allows for the occurrence of actions of multiple agents by introducing non-stationary notions of waiting and test. The need to state a large number of frame axioms is alleviated by introducing a concept of chronological preservation to dynamic logic. As a side effect, this concept permits the encoding of temporal properties in a natural way. We compare the relative merits of our approach and non-monotonic approaches as regards different aspects of the frame problem. Technically, we show that the resulting extended systems of propositional dynamic logic preserve (weak) completeness, finite model property and decidability.