A semi supervised learning based deep learning image dehazing algorithm is proposed to address the issues of high cost of real data annotation and domain differences between synthesized data and real scenes in foggy image dehazing tasks.Combining supervised and unsupervised dual branch training frameworks to achieve mixed feature fusion and adversarial training,that is,using convolution operations and attention mechanisms to extract mixed features,and using supervised methods to convert foggy images into clear images.Afterwards,unsupervised processing was achieved through generator conversion and loss function optimization to obtain real dehazing images.The results showed that the proposed algorithm achieved a structural similarity value of over 0.99 on the fog image test set,and its information entropy,visual information fidelity,and standard deviation values of 12.49,0.569 2,and 57.23 on highway image dehazing were significantly better than other compared algorithms.This method can adapt well to distant foggy scenes and effectively provide a solution for dehazing real scenes at low annotation costs.
上下料搬运机械手作为自动化生产线向智能化转型升级的重要组成部分,对生产线智能化和提升生产线效率起着至关重要作用.以搬运机械手为研究对象,基于形态学矩阵法对机械手结构设计方案进行优化.通过疲劳仿真与有限元仿真方法对搬运机械手关键结构部件进行分析,验证理论推导准确性.结果表明,采用计算机控制的平板式机械手为最佳设计方案,且齿轮传动方式疲劳值更小,夹爪机构应力值更小,应力分布更均匀,最大应力47.5N/mm2.所提出机械手可有效延长其工作寿命,为机械手结构设计与力学性能优化奠定理论基础.
针对水下环境噪声干扰以及水下时延误差等问题,提出基于自适应滤波长基线组合AUV定位算法.该算法采用长基线组合导航定位系统,以捷联惯性导航系统为主,长基线定位系统为辅,消除时延误差实现误差补偿;再引入自适应滤波算法,使用卡尔曼滤波算法对系统进行信息融合来矫正累积误差实现自适应滤波定位.仿真实验表明,该算法能有效消除水下复杂环境中的噪声干扰对AUV定位精度的影响,提高算法对未知噪声的适应性和定位精度.
利用铁酸钴纳米颗粒(CoFe2O4NPs)催化鲁米诺-H2O2的化学发光,结合单宁酸会抑制CoFe2O4NPs的催化活性这一性质,首次建立一种简单、快速且超灵敏的单宁酸检测方法.在最佳实验条件下,CoFe2O4NPs-鲁米诺-H2O2体系对单宁酸的检测线性范围为5.0×10-9～5.0×10-7 mol/L,检测限为2.5 ×10-10mol/L.对比其他检测方法,该检测限极低,而且该检测体系选择性良好,有望将该方法运用于食品及水样中单宁酸的检测.
为满足专注力训练领域的需求,克服传统正念训练方法中存在的局限,以Arduino为核心,以TGAM的专注度和放松度为数据采集观测值,设计脑电波专注力训练系统.系统分专注力训练反馈和辅助正念两种模式.在专注力训练模式中,通过舒尔特方格游戏,LCD实时显示专注力数值,并且达到设定的专注力值之后,LED灯点亮.在正念模式中,通过电脑播放正念音乐,利用Arduino模块控制三色LED灯的状态,不同的专注状态时LED灯显示不同的颜色.使用VS2017开发了上位机,借助上位机完成舒尔特方格专注力训练和正念辅助音乐播放功能.测试结果表明,该系统性能稳定,使用方便,在专注力训练领域有较好应用前景.
