Energy Acquisition And Analysis System For Diabetic Pregnant Women Based On Image Processing

Gestational diabetes is a high-risk pregnancy that affects both the mother and the fetus.At present,10 out of every 100 pregnant women in China suffer from the disease,which is more than three times higher than the proportion of 3% who suffered 10 years ago.The diet,exercise,and blood glucose of gestational diabetes patients are important indicators that doctors pay more attention to.The above information needs to occupy a large amount of time.Therefore,if a diabetes maternal health management system is established,it can help the pregnant woman to collect information related to diabetes,such as diet,exercise,etc.,anytime,anywhere,and automatically analyze the relationship between them,which can help the doctor to give a more advice with the patient's own situation.This article mainly discusses the food data collection and analysis methods and the specific implementation of the system in this system.In the data collection phase,the paper first uses the MySQL database to build a food nutrition information database,and then uses the MVC framework and JavaEE technology to implement a WeChat service number that allows diabetic pregnant women to enter daily diet,exercise,and blood glucose information anytime and anywhere.After that,doctors can provide users with personalized health guidance through the above information input by users.Among them,in the diet data acquisition module,the most commonly used methods of manual input of food name and gram number are changed,but the food images are automatically identified and analyzed by the user through the mobile devices,and the variety and energy of the food are automatically identified and analyzed by the system,so as to reduce the complexity of the users' input information.In the data analysis phase,we identify and analyze the way users absorb nutrients and energy through uploaded food images.This thesis innovatively adopts Faster RCNN target detection method to locate and recognize food.When users take pictures,the distance and angle between mobile devices and food are not fixed.Therefore,this study uses the coins that people often take around with a fixed size as object of reference,that is,each time a coin is taken with the food to be photographed and uploaded.In this study,we first built a food image library containing coins,which contains 11000 image data of 20 dishes.Then we mark the location of coins and food in the image.Then,the 5500 images in the image library are trained to train the RPN and ROI layers in Faster RCNN,and the remaining 5500 images are tested to identify the food types and position coordinates,and the coordinates are converted to area.Finally,by calculating the ratio of the area of each dish to the coin,the weight of food is calculated and converted to the corresponding caloric,which is the basis for analyzing the user's diet data.