The Development of a Real-time Object Recognition System on the Mobile Devices

• Main Authors: TU,CHUN-SHENG, 凃俊升
• Other Authors: CHEN,DING-HORNG
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/ppukr4

碩士 === 南臺科技大學 === 資訊工程系 === 104 === While image recognition and image retrieval have been popular issues in image processing in recent years, most recognition and retrieval of images have been carried out on computer, failing to reach such recognition “anytime and anywhere”. The current study, therefore, aimed to solve such limitation by combining smartphones with image recognition installed in computer. Wi-Fi and the built-in camera of a smartphone were used to transmit images to a computer for image recognition. After processing, subsequent results were sent back to the smartphone with the use Wi-Fi. With regard to image processing, FAST (Faster and Better: A Machine Learning Approach to Corner Detect) was employed for feature retrieval. ORB (Oriented FAST and Rotated BRIEF) was adopted to depict features as feature descriptors, which were further grouped to create related image vocabulary with BOVW (Bag of Visual Words). Support Vector Machine (SVM) was then utilized to analyze image vocabulary for subsequent image classification. The results revealed that, with the proposed method, the average image recognition rate was 79.8%. Smartphone devices and computers could be utilized for real-time image transmission and image recognition.