Water Meter Number Detection Based On Deep Learning And Its Implementation For IOS Application

In recent years,with the continuous development of mobile communication network and Internet,mobile phones,tablet computers and other smart portable devices are gradually popularized,and people have a growing demand for relevant intelligent applications,especially in the field of computer vision and text recognition.How to combine artificial intelligence and mobile Internet to deal water detection is the main concern of this thsis.Water resource is an important resource related to the national economy and people’s livelihood.Water conservation is not only a basic requirement for individuals,but also for enterprises and society.However,the manual meter reading method of traditional water companies is a huge waste of manpower and material resources,and the intelligent meter reading system has become the most important development point of water companies in the future.this paper applies deep learning to the meter reading business of traditional water companies,and designs a set of automatic meter reading application based on iOS system.The main contributions of this thesis are listed belows:1.This paper collects and constructs a water meter data set for practical application scenarios,which contains 28993 water meter images,covering the technical challenges of different lighting,crack interference,blur,tilt,water splash,fog,shadow and other application scenarios.All images are annotated with quadrilateral to mark the position of the water meter reading in the picture.2.Most of the traditional general template detection methods cannot be directly used to detect water meter readings with rotation angles.For the real-time detection of multi-directional water meter readings issue,a new rectangular bounding box decomposition network(Rect BDN)and a new multi-directional rectangular YOLO(ROYOLO)are proposed.These two new water meter detection models can better solve the multi-directional detection problem of water meter.3.This paper develops a smart meter reading application software based on the iOS platform.The application is based on Apple Core ML.It implements the mobile terminal implementation and application deployment of the ROYOLO water meter detection model proposed in this thesis.It can complete the mobile terminal water meter reading detection and complete the cloud water meter recognition and data interaction through the cloud server engine.