基于深度学习的智能洗衣机系统构建

发布时间：2018-03-20 02:12

本文选题：智能洗衣机　切入点：深度学习　出处：《深圳大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着社会的进步和科技的发展,家电产品得到了广泛的普及。而计算机技术,包括深度学习技术的迅猛发展,使得人们对家电的智能化需求成为可能。将深度学习技术应用在智能家电领域将是智能家电领域新的发展方向。深度学习算法可以克服传统智能家电的局限,使智能家电为人们带来更加便捷和舒适的生活模式。本文主要提出了一种新型智能洗衣机的设计模式,在洗衣机内部放置一个高清摄像头,通过该摄像头采集到待洗衣物的图像,然后利用神经网络算法对图像进行分析,得出洗衣机内部衣物量,衣物的材质信息,例如毛衣类衣物,牛仔面料衣物以及普通棉麻类衣物。根据这些信息自动提供一套合适当前环境下的洗衣方案,如洗衣过程需要的水量,需要的洗涤剂量,洗涤剂的种类以及洗衣过程波轮旋转的速率等,从而智能控制洗衣机。本文的创新点及主要工作在于:1)提出一个结合深度学习和智能洗衣机的具体应用场景,并描述了其整套工作流程,同时将问题转化成图像分割和纹理图像分类领域的问题;2)设计了一个基于卷积神经网络的图像分割算法,该算法可以实现对灰度图像的前景背景分离;设计了一个基于卷积神经网络的纹理图像分类算法,该算法能够识别毛衣面料纹理、普通棉麻面料纹理以及牛仔面料纹理;以及一个基于浅层学习的图像分类模型,该模型能够识别掩码的面积大小;3)实现了一个基于上述算法应用于智能洗衣机系统的仿真,并证实了可以通过深度学习算法实现智能洗衣机对内部衣物的衣量和衣物材质识别。本文分别从网络结构、激活函数、损失函数、优化算法以及过拟合防范方法5个方面描述了算法的设计和实现,并通过实验验证了通过神经网络算法能够获取传感器所无法获取的衣物材质信息。因此这种新型智能洗衣机能够制定更加合理的洗衣方案,避免用户在使用洗衣机的过程中做太多基于经验的选择,为用户提供更加智能的洗衣模式、带来更加便捷生活方式。
[Abstract]:With the progress of society and the development of science and technology, household appliances have been widely popularized. And the rapid development of computer technology, including in-depth learning technology, It will be a new development direction to apply the deep learning technology to the intelligent appliance field. The depth learning algorithm can overcome the limitation of the traditional intelligent home appliance. This paper puts forward a new design mode of intelligent washing machine, in which a high-definition camera is placed inside the washing machine. Through the camera to collect the image of laundry, and then use the neural network algorithm to analyze the image, get the washing machine inside the amount of clothing, clothing material information, such as sweater clothing, Denim fabrics and general cotton and linen clothing. Based on this information, an automated laundry program is provided for the current environment, such as the amount of water needed for the laundry process, the amount of washing needed, The type of detergent and the speed of washing wheel rotation in the laundry process, so as to control the washing machine intelligently. The innovation and main work of this paper is to put forward a concrete application scene combining deep learning with intelligent washing machine. At the same time, the problem is transformed into the problem of image segmentation and texture image classification. (2) an image segmentation algorithm based on convolution neural network is designed. A texture classification algorithm based on convolution neural network is designed, which can recognize the texture of sweater fabric, common cotton fabric and denim fabric. And an image classification model based on shallow learning. The model can recognize the area of mask and realize a simulation of intelligent washing machine system based on the above algorithm. It is proved that the intelligent washing machine can realize the recognition of the clothing quantity and clothing material through the depth learning algorithm. In this paper, the network structure, the activation function, the loss function, the structure, the activation function and the loss function, respectively, can be realized. The design and implementation of the algorithm are described in five aspects: optimization algorithm and over-fitting prevention method. The experiment proves that the new intelligent washing machine can make a more reasonable laundry scheme, which can not be obtained by the sensor through the neural network algorithm. Avoid users making too many experience-based choices in the process of using washing machines, provide users with a more intelligent laundry mode, and bring a more convenient way of life.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TM925.33

【参考文献】