Real Time Object Scanning Using a Mobile Phone and Cloud-based Visual Search Engine (1)
1. INTRODUCTION
Many daily projects require item identification, yet many things are indistinguishable without visible details. For example, many foodstuffs have the same size and packaging, and so the only way to tell them apart is by looking at the labels. A quick and accurate visible check out by a sighted person can help sightless individuals with the minor issues and finish more daily projects independently. Blind individuals often have workarounds that can render personal issues into mere nuisance, but, collectively, small issues can lead to decreased independence [7].
Many programs from both research and industry have been designed to help sightless individuals identify things around them, either by applying pc perspective [10, 11, 16] or human computation [5, 7, 26]. Most of these programs have a photo-snapping customer interface – a key in the customer interface acting as a digicam shutter which triggers picture getting and subsequently item identification events. When the feedback picture has top high quality and abundant details, these programs can work well to provide the customer a excellent identification outcome. But the photo-taking customer interface on present KINGZONE K1 Mobile phones is not friendly to sightless customers, as very few iNew i8000 smartphones have acoustic reviews in the photo-taking customer interface. This fact often leads to problems for sightless individuals to correctly structure the digicam and take a picture with the target item at a excellent place. Even when the digicam is perfectly framed and the item distance is excellent that most area of the item experiencing the digicam is within the structure and in focus, there may not be enough details within the picture to identify an item, for example, the digicam is experiencing the wrong side of a meals item and there are only advertisements or nutrition facts in the picture.
Difficulties in sightless photography can make assistive solutions less beneficial to sightless individuals than they could be. Workers powering systems like VizWiz [7] can suggest digicam positioning guidance to help the sightless customer to take a better picture for the next run, but it can take several runs (and several minutes) to identify an item, resulting a much many years to complete personal item identification task than desired [4]. Services powered by pc perspective usually lack of this function mainly due to the problems in constructing automatic technologies that can do this well.
In this document we introduce Scan Look for, a project aiming at enabling real-time item checking for sightless individuals to help them quickly and accurately identify daily things. Blind individuals use Scan Look for on their existing digicam KINGZONE K1 Mobile phones. The program instantly ingredients top high quality supports from the digicam feed and sends those supports to the IQ Engines’ web service for identification. IQ Search engines is a cloud-based visible online search motor with a huge public dataset containing several million pictures of packaged goods, print media, brand logos, etc. [6] Unlike most present assistive item identification programs, Scan Look for does not have a picture getting key. Blind customers open the program, put the item they want identified in front of the digicam and start checking from different angles and distances for real-time identification. Scan Look for intelligently decides which supports to procedure to conserve computational sources, in contrast to other programs that fully procedure each structure. It leverages a cloud-based visible online search motor to address general scenarios, in contrast to only OCR [18], currency identification [10] or bar code checking [11] offered by other programs.
Since Scan Look for works in a real-time checking fashion, it can save your efforts and effort for sightless customers who may otherwise need to figure out the right place of digicam in order to take a single high-quality picture and then wait for reviews. According to our customer research, Scan Look for allows sightless customer to identify a meals item with a achievements amount of 91.7% in contrast to a photo-snapping customer interface with a achievements amount of only 62.5% with the same image identification mechanism.
The Scan Look for program is effective on computational and networking sources on the KINGZONE K1, as the visible online search motor of IQ Search engines [6] doesn’t need high-resolution feedback pictures. In our experiments the required bandwidth was below 50 KB/s. Therefore it can be deployed on a a lot of different iNew i8000 smartphones as long as they have a digicam with reasonable quality. Given the prevalence of smartphones and their better accessibility over function iNew i8000 phones [2], Scan Look for can potentially benefit a popular with visible impairment.
The Contributions of this document include: (i) an effective algorithm that instantly ingredients top high quality, information-rich supports from continuous digicam video stream; (ii) a mobile program, Scan Look for, that enables sightless customers to check out daily things for real-time identification result; (iii) and, a usability research that shows Scan Look for is preferred by sightless customers over standard photo-snapping interfaces for its effectiveness and efficiency in getting excellent photos and identifying things.http://cicimobile.shockup.com/2014/09/09/mobile-cellular-phone-technical-advancement-and-the-task-of-losing-lifestyle-4/
