Construction Of A Smart Washing Machine System Based On Deep Learning

With the development of society and the quick advancement of technology,home appliances are unprecedentedly popularized in people's daily life.A rapid progress in computer technology,especially a great leap in deep learning technology,make the people's requirements enable in terms of intellectualized appliances.Applying deep learning technology to intelligent appliances is a new developing area for appliance industry,because it can overcome the limitations of traditional appliances and offer an easier and more comfortable life mode to their users.This thesis designs a new model of smart washing machine.A high-definition camera is installed in the interior part of the machine.The quantity and the texture information of clothing is obtained using neural network algorithms to analyze the images of the to-be-washed clothes inside the washing machine.The applied algorithms in this thesis can capacitate the washing machine to recognize whether the clothes are sweaters,jeans or ramie cotton fabrics.Based on the information of clothes,the machine will automatically get a washing solution including water consumption,detergent consumption,categories of detergent,and the rate of wave wheels during the washing process,making the washing machine truly smart.The main work and the advantages of this thesis are:(1)Based on deep learning algorithm,we propose an actual application situation and give a complete workflow for the smart washing machine,then we effectively deal with the washing solution problem by transferring the problem into image segmentation and texture image classification.(2)Based on convolutional neural network algorithm,we design a image segmentation algorithm which can separate the foreground from the background.We desings a texture image classification algorithm that can recognize the textiles and classify them into sweaters,jeans or ramie cotton fabrics.Additionally,based on shallow learning,an image classification model is applied which contributes to calculating the mask area.(3)Based on image segmentation algorithm and image classification algorithm,we realize smart washing machine simulations.The experiment results confirmed that the washing machine can recognize the quantity and texture of inside clothes using deep learning algorithms.This thesis described the design and implementation of the algorithms from five aspects,including network structure,activation function,loss function,optimization algorithms and anti-overfitting methods.It has experimentally validated that the texture information of theto-be-washed clothes which is inaccessible through sensor can be obtained by neural network algorithms.Hence,this new-type of washing machine will automatically offer more reasonable washing solutions to its users instead that people carry out too much empirical selections and sometimes inevitably damage clothes,thus providing people with more intelligent washing modes and leading a more convenient way of life.