Design And Implementation Of Android Application For Water Meter Detection Based On Deep Learning

With the popularity of smart phones and various applications,mobile internet is increasingly close with our life.Besides,the rapid development of artificial intelligence in recent years has also accelerated the arrival of the new technological revolution.More and more traditional enterprises began to focus on artificial intelligence and mobile internet.Thus,the combination of artificial intelligence and mobile Internet becomes an important issue for us.The traditional billing model of the Water Service Company results in a big waste of time and labor on calculating the amount of water use because they have to check the water meter for every house monthly.This paper introduces the deep learning technology to the traditional Water Service Company business by designing a water meter detection system based on Android platform.This paper has the following contributions:1.We apply three methods,including Hough transform method,least square method and deep learning method to detect the circle area of the water meter.The experimental results show that deep learning method performs the best.In this paper,the algorithm is designed to extract the feature of the water table disc by using the convolution neural network.After the feature fusion,the candidate window is extracted into the classification layer and the regression layer.In this paper,we annotate the object as circles,which is different from the rectangle labels in traditional object annotations.So the Monte Carlo sampling method is used to calculate the IoU to distinguish the positive and negative samples.This paper designs and implements a deep learning based circle area of the water meter detection algorithm to extend the application of deep learning,which indicates the successful attempt to apply deep learning methods to solve the traditional enterprises' problems.2.We develop an Android application for water meter detection with the function of registration,login,identification,management and etc.The unit testing and black box testing shows perfect client function and user experience.The stress test proves the robustness and good usability of this application.3.We design the network request framework and the picture cache framework to optimize the network request and picture cache process,which helps to improve the performance,to save system resources and to improve the user experience greatly.The network request framework mainly utilizes multithreading technology,caching technology and network transmission related technology,unified management of all network requests.Picture cache framework designed a multi-level image cache unified management of the picture,and to achieve a variety of cache elimination algorithm.The network request framework and the picture cache framework support dynamic configuration with strong scalability.