Automated Recognition and Sorting of Remote Control Devices


Automated sorting system (courtesy of BEKO International, Inc.). From the right part, remote controllers are fed automatically one-by-one from the assembly line. A camera acquires their image and visual processing determines their type. Accordingly, the remote controllers are directed to one of the control stations. Those whose types are unidentified are directed to a basket as seen in the front.


We developed an automated remote controller sorting system for a TV manufacturing plant BEKO International, Inc. - currently operational as seen in the figures. There are five different types of remote controllers with ten different colors. Each type of remote controller can be distinguished based solely on the outer shape. In this application, all the different types of remote controllers are being manufatured on the same assembly line. After being manufactured, they are subjected to functional testing - which varies according to their type. Hence, an automated visual sorter station - placed on the assembly line - ensures that an incoming remote controller is sent to the appropriate control station.

The visual processing of the inspection system was based on the shape assessment algorithm in my M.S. thesis at Bogazici University. To manage time constraints, we used selective perception and potential functions in our algorithm and implemented it on Matrox Imaging card. The shape assessment algorithm is based on selective perception. The novelty of this approach is that instead of processing the whole image, only the areas that are deemed "interesting" and hence calling for attention are analyzed. The attentional sequences thus generated can then be used for a variety of tasks including shape determination. Since only a small portion of the whole image is processed, visual processing can be real-time and flexible without requiring special hardware. Finally it is simple enough so that new items may be easily defined.


1. Some of remote controllers with different shape and color.

2. Once the remote control device arrives the imaging capsule, its picture is taken.

3. then the image is processed in the matrox imaging card.

First, the system is presented with a sample of each remote controller type. It selectively attends to the image - thus coming up with a sequence of fixation points. It then finds the contour segments going through these fixation points and then merges them to find the closed outer shape as explained in publications listed below. It is then set to learning mode where the parametric representation of the outer shape is retained for future use. This is repeated for each type. After being presented with samples of all types, the system is ready to operate in sorting mode. When a "to-be-sorted" remote controller comes to the sorting station, it is selectively attended and the attentional sequence thus generated is used to eventually generate the invariant parameters of its outer shape. Since the outer shape is sufficient for determining its type, this parameter set is then compared with those of the models. According to the comparison results, the remote controller is directed to one of the possible control stations through an electro-mechanical setup as seen in figures. If the system cannot assign the incoming remote controller to any of the five classes, then the situation is announced as a miss and the remote controller is sent to a basket.

This project on the automated sorting of remote controllers has been made possible with cooperation of BEKO International, Inc. We gratefully acknowledge the contributions of BEKO team to the mechanical design and implementation efforts.

This system was among the five finalist in the national best industrial project competition in 1999 in Turkey.


Collaborators: Prof. Isil Bozma & Burak Gokturk.

4. Once the type of the RC device is determined, it is directed to corresponding line.

5. Remote control devices move along assembly line

6. and accumulated in baskets to be packaged.




[Research]
        [Rapid Gain Change in Thermal Imagery]
        [Tracking Vehicles In Airborne Video Imagery]
        [Dense Motion and Appearance Estimation]
        [Implicitization by Matrix Annihilation]
        [Modeling and Measurement Using IPs]
        [Automated Sorting of Remote Controllers ]
        [Shape Assessment by Selective Fixations ]



Related Publications
  • H. I. Bozma and Hulya Yalcin,
    "Visual Processing and Classification of Items on a Moving Conveyor: A Selective Perception Approach,"
    Robotics and Computer Integrated Manufacturing, 18(2): 125-133, April 2002.
    [Abstract] [pdf]
  • Hulya Yalcin,
    "Theoretical and Practical Aspects of 2D Shape Assessment using Selective Fixations and Artificial Potential Functions"
    M.S. Thesis, Systems and Control Engineering, Bogazici University, February 1999.
    [Abstract] (Download [pdf])
  • Resume | Research | Main Page
    Carnegie Mellon University, Robotics Institute
    5000 Forbes Av., Pittsburgh, PA, 15213

    hulyayalcin@gmail.com