In order to solve the problem that traditional wafer positioning is easily affected by lightness,noise and other fac-tors and consume a lot of resources,the wafer detection algorithm YOLOv5s-wafer was constructed.The wafer detection dataset was constructed firstly,the lightweight network GhostNetv2 was used as the backbone feature extraction network to reduce the amount of model parameters.Then the CA attention mechanism was introduced into the feature fusion network to enhance feature extraction capabilities.Finally,EIOU was used as positioning loss function to improve the wafer detection accuracy.The experi-mental results show that the detection algorithm average accuracy is 99.3%,the parameter number is 4.637×106,which achieves an ideal balance between the detection performance and the lightweight of the algorithm.
针对变电设备智能物联监测中的分布式智能感知节点微型化设计需求,设计了一种新型的超声局部放电测量转换电路.使用高集成度仪表放大电路对高阻抗的超声局部放电传感器输出信号进行了低阻抗变换和固定增益放大,设计通带为 20～300 kHz的宽带滤波电路以适应变电设备上带宽不同的传感器,最后通过增益可调仪表放大电路适应不同灵敏度的传感器输出信号.电路能同时连接2 个超声局部放电传感器并对其测量信号进行转换,电路尺寸为83 mm×50 mm×15.5 mm.设计了相应的电路性能测试实验并给出了其在智能物联系统中的应用方法.实验结果表明:电路各级增益误差在整个通带频段内低于1%,1000 倍增益条件下的输出端噪声RMS值不超过10 mV.
针对MEMS压阻式压力传感器微型化发展过程中精度难以保证的问题,通过有限元仿真分析方法,研究固支结构、封装约束方法等变量对压力传感器输出的影响.结果表明,固支结构的引入会导致压敏电阻所处路径出现24.93%的应力偏差,不同封装方法会使路径上产生0.54%～2.42%的应力偏差.当固支结构宽度与压力敏感薄膜膜厚的比例大于 3 ∶ 1 时,压敏电阻所处位置应力积分均值稳定.结果表明,固支结构的约束方法、结构宽度均会对压力传感器的输出产生重要影响,合理设计固定约束结构可为压力传感器的微型化高性能设计提供保障.
In view of the influence of camera imaging deviation caused by plane refraction on visual measurement of the glass protective cover of explosion proof camera,a camera model under non-parallel plane refraction was derived,and on this basis,a correction method of refraction distortion based on dynamic virtual focal length was proposed,which effectively solved the problem that the refraction distortion was difficult to correct when the camera optical axis and the normal vector of refraction surface were not parallel.The experimental results show that the root mean square error of the corrected image deviation decreases by 68.69%and 72.82%,respectively,and the relative error of the corrected measurement is less than 0.5%.The positioning accuracy and measurement accuracy of the camera are obviously improved after correction.
In order to solve the automatic feeding problem of a series of non-standard screws,a screw flexible grasping system based on machine vision was proposed.Firstly,the Eye-to-hand visual positioning and grasping system was designed,and the ac-quired images were identified by image preprocessing such as binarization and particle analysis.Secondly,for the screw target,a series of circular characteristic parameters of the screw center direction were obtained by fitting the circle operation,and the cen-ter line was obtained by fitting the least square method,furthermore,pose parameters of grasping point and angle of the screw were solved.Finally,the ROI region conforming to the grasping characteristics of the claw was designed with the grasping point as the center to judge the grasping.The upper computer sent the pose information to the controller,guided the robot to move to the corre-sponding position,and completed the grasping of the screw workpiece through the end claw.Through the experimental verifica-tion,the positioning accuracy of the system is 99.5%.
