Research On Food Volume Estimation Cloud Computing Technology In Diet Health

Obesity has become an epidemic all over the world. The cause of obesity is diet in additionto hereditary factors. Reasonable energy intake and eating habits can improve the rapidly spreadof obesity. A method has been built to find a solution with information technology aim to thediet health problem. First, a wearable device has been developed to take diet pictures, and thenthe ellipse automatic detection algorithm is used to detect the plate in the diet pictures to get theelliptic equations of the plate. The plate can be used to calibrate the camera to get the relativeposition and rotation angle between the food in the picture and the camera. After the cameracalibration, the virtual reality technology is used to build virtual scenes according to the dietcameras environment. Then build3D object which is similar to the food in the picture. User canmanipulate the object by rotating, zooming or moving to fit the real food in the picture. Then thevolume of the real food can be estimated according to the volume of the virtual object.This method has basically been finished, but there are still some drawbacks: First, thepopularity of the wearable device is too low. The wearable device is still a laboratory productand not mature enough to enter commercial market. Besides, this product is only for specialgroups such as the elderly or patients with obesity which means it cannot be used by more users;Second, virtual reality and other functions are running on the desktop. It will take a very longtime to process the pictures due to the large amount of calculation. Finally, the traditional userinput methods like mouse and keyboard are not suitable to manipulate the3D objects in thevirtual reality environment.To solve these problems, this food volume estimation method need to be improved. First,the smart devices like smartphones can be used to take diet pictures instead of wearable devices.With the popularity of smartphones and other smart devices, the functions of picture taking anddata processing have been developed rapidly. These smart devices can take pictures which areclear enough for research. Second, the processing capabilities of the smart mobile devices andthe multi-touch technology on them, makes them very suitable to finish the relative functions inthis method. Third, a cloud computing platform has been built to store the large number dietpictures taken by users. These pictures will provide basic and accurate material for diet healthresearch. Fourth, a stable and efficient cloud processing platform has been built to process thepictures stored on cloud platform. The functions of ellipse automatically detection and cameracalibration is deployed on the cloud computing platform.The structure of Cloud plus Client is used to implement this system, in which the Cloud and the Client will cooperate with each other. The functions of Client mainly developed on themobile devices, including drawing or adjusting the ellipse when the plate cannot be detectedautomatically or the detection is not very accurate. At the same time, users can create objectsand fit them to the food to calculate the volume of the food on the Client to get more accuratefood volume.The Cloud is used to store pictures taken by users with a distributed architecture, andprocess the mass pictures stored on the Cloud by detecting the plate in the pictures andcalibrating the camera. The calculation result can be returned to the Client to allow the Client tocreate3D scenes according to the result.The innovations include:1) This system uses virtual reality technology to calculate food volume. This workproposed a Cloud and Client cooperate method, researched and implementedautonomous and intelligent3D multi-touch interactive technology to achieve a virtualreality model to fit real food.2) To fix the HDFS file system storing small files problem, this work proposed a new filesystem with two level indexing mechanism. By using Virtual Big File method, solvedthe massive food small file picture management issues.3) To fix HDFS copy strategy problem, this work proposed a new copy strategy basedon the performance of DataNode. This strategy is based on the performance ofDataNode and the comprehensive performance of files. This strategy is verified byexperiments.4) This work proposed a new cloud computing platform scheduling algorithm, whichtakes full advantage of the computing resources and task parameters. This algorithmadded task execution feedback, and obtained reasonable scheduling result.