Presentation on theme: "DDRS – Dietary Data Recording System"— Presentation transcript:
1 DDRS – Dietary Data Recording System The Dietary Data Recording System supports the dietary assessment of cancer patient participants by facilitating collection of more objective measurement of dietary intake in real time and with moderate cost.
2 Motivation Relationship between Diet and Health Increase study participants satisfaction!Inaccuracies from under reporting intakeCost Reduction(Doubly Labeled Water)Address issues with traditional dietary recording systemsInconvenience of remembering food log
3 DDRS Dataflow Solution components nutritionist, device use, data managementStudy participants log food intake on a device.Data uploaded to a server which handles image processing and volume estimate.The collected and calculated values are imported into a web interface for review by a clinical coder.The Coders prepare the data for entry into NDSR which can be used for performing nutrient analysis based on the data.
4 Data Flow Meal Entries Via Android App Segmentation Volume Estimation Data PublicationCoder InterfaceNDSR - Nutrition Analysis
11 Other Techniques of Segmentation and Volume Estimation Brightness down by -100Contrast up by 70Grayscale
12 raw image blurred image To solve this problem we make two kinds of operations ahead of Canny operation, which are Smooth and Erode. Both of them are used to blur the image to remove tiny noise dots. In the right image, we can see the plate edge is almost perfect and this bright noise is merged into its neighbors and its edge is vague.raw imageblurred image
13 Canny on the blurred image Now we can implement Canny. The result image has just two smooth contours that we care about without noise or any other contour we don’t need. The contours can be clearly seen by our eyes.blurred imageCanny on the blurred image
14 But it’s not enough just showing an image with contours But it’s not enough just showing an image with contours. The ultimate goal is to make the computer recognize the objects. So right now we need to mark the contours and store them in the memory by some data structure. In OpenCV there is a library function called cvFindContours(), which is able to detect all the closed contours in an image and store them as a linked list. Here is the result. It was the Canny operation that makes this function able to find contours very accurately without any trivial noise.cvFindContours()
15 3D Point CloudWe will use a cellphone with a laser on it. Unlike usual laser devices, this laser will emit a line onto a surface. We hold this line to scan through the food on a plate from one side to the other.
16 Front View of an example cloud 3D Point CloudThen we can get a set of data visualized like this. This is the front view of the point cloud, because each line is made from points and all the lines form the cloud shaped data. We will use this data to estimate the volume of the thing it covers.Front View of an example cloud
17 Top View of another example 3D Point CloudThis is the top view of another example cloud data. The black part is the plate holding the food, and the blue part is the food. At last we use the segmentation image to cancel the plate part in the image so that we have only the food point cloud. Then we can use some math model to calculate the food volume.Top View of another example
18 Intelligent Trash Can for Smart Home/Restaurants Data CollectionData ProcessingData AnalyticsApplicationsKinectsCheap camerasWeighing scaleBetter Waste ManagementCost ReductionGreener planetImage detectionVolumetric AnalysisCarbon footprint measurementDemographic waste information
19 Call for Papers - LinuxFest What – LinuxFest Northwest 2013When – April ’13Where – Bellingham Technical College, WAPresentation proposals due – 15 March ’13