Summary: | 碩士 === 國立中正大學 === 電機工程研究所 === 104 === With the popularity of health awareness, there are more and more software and
hardware that can help people record diet and measure calorie. Traditional written
records, electronic barcode and measureing calorie with look-up table became
incovenient. With the popularity of smartphone, the way using image recognition to
record personal diet and measure calorie become a more convenient and high
accuracy choice.
Vegetable and fruit recognition is a image recognition method using vegetable
and fruit images. Because we use hand to measure food weight, we do not have to
prepare the other measuring tools. In the thesis, we take photos of vegetable and fruit
with smartphone, then we extract color features, shape features and texture features
with technology of image processing. The matching rate of our method is 98.5% and
the error rate of calorie measurement is 4.5%. Compare to other food regocnition
method[23],our method use less kind of features ,but we have higher matching rate
(98%>95%) based on same SVM kernel(RBF). It is obvious that our method of
vegetable and fruit recognition is higher accuracy and convenient method.
|