基于卷积神经网络的散点透镜识别方法  

Recognition Method of Scatter Lens Based on Convolution Neural Network

作　　者：刘宇飞 黄秋实 霍彦明 王昱 

机构地区：[1]河北科技大学电气工程学院,石家庄050018

出　　处：《日用电器》2021年第5期60-64,共5页ELECTRICAL APPLIANCES

摘　　要：目前国内外的电路板焊接技术正处在飞速发展阶段。贴片机是用来实现高速、高精度贴放元器件的关键的设备,但贴片机对不同器件的识别能力亟待提高。本文针对此问题提出一种基于卷积神经网络(Convolutional Neural Networks,CNN)的识别方法,旨在提高贴片的速率、精度和稳定性。经过大量测试得出,本文提出的这种包含卷积计算且具有深度结构的前馈神经网络识别方法得到的识别精度远高于现在工业化生产常用的激光识别和相机识别,与此同时此方法的应用也大大提高了贴片速率和稳定性。At present,the circuit board welding technology at home and abroad is in a stage of rapid development.The placement machine is a key equipment to achieve high-speed,high-level placement of components.The placement machine's ability to recognize different components needs to be improved.This paper proposes a recognition method based on convolutional neural network for this problem.The purpose is to improve the speed,accuracy and stability of the patch.A fter a large number of tests,it is concluded that the recognition accuracy of the neural network recognition method with deep structure including convolution calculation proposed in this paper is much higher than that of laser recognition and camera recognition commonly used in industrial production.At the same time,the application of this method is also Greatly improve the patch rate and stability.

关 键 词：CNN 识别 贴片机 透镜 目标检测 

分 类 号：TN405[电子电信—微电子学与固体电子学] TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]