2D Shape Assessment using Selective Fixations and Artificial Potential Functions



(a) Fixation points generated on the pincers that is lying on a textured surface.
(b) Ordered sequence of boundary points generated using previously found fixation points.


Studies in vision science have revealed that biological systems work by allocating limited computational resources to only the interesting parts of an incoming image. This is done by saccades, rapid eye movements that direct the optical axis to a target fixation.

Traditionally, research in machine vision has concentrated on thorough analysis of acquired images. This contrasts with human perception. For human visual behaviour, selectively gathering information about the environment is characterized by the ability to fixate on a point of interest and the ability to select new fixation locations. Humans can shift the attention by concentrating on a part of the field of view. Motivated by human visual system, recent studies in machine vision have tried to mimic this behaviour. These studies depend on detection of visual targets and interrogation around those targets instead of processing whole image. The process of identifying and selecting new visual targets is referred to as selective fixation control. Remarkable efficiency on computational grounds has been observed in machine vision systems motivated by fixation control. In this manner, visual resources are allocated to process only a small part of the whole scene.



Paths leading to next fixation point and the surfaces represented by artificial potential functions for first, second and third fixation points.


In preliminary research, we have devised a visual inspection system - based on a simple heuristic algorithm - which selectively fixates on the interesting parts of an incoming image and uses the attentional sequence thus gathered in a task-dependent manner - specifically the task of tracing an object's outline. In this thesis, we developed a unified mathematical framework based on artificial potential functions - which formalizes this algorithm. It turns out that this formalism tantamounts to a feedback based approach - that naturally leads to the automatic generation of camera actuator commands that cause the camera to find a sequence of fixation points and use the attentional sequence thus generated a in top-down driven task - such as the tracing of salient parts of the objects. In contrast to open loop, process-all systems, the provable correctness of such a closed loop system can be investigated.

In this approach, the visual processing consists of a continuum of pre-attentive and attentive stages. An attentional sequence thus generated represents the visual data spatio-temporally. This processing occasionally followed by cognition, where attentional sequences are processed according to the demands of the task at hand.




(a) Fixation points generated on the industrial object occluding the elliptically-shaped paper. The intensity values along region of occlusion are close. (b) Sequential boundary points found on object in (a) where paper and industrial object is perceived as a composite object. (c) Fixations points generated on the industrial object occluding the elliptically-shaped paper. Intensity values along region of occlusion are not close. (d) Sequential boundary points found on object in (c) where paper and the industrial object is perceived as two distinct objects.










We employed this approach for the automated sorting of remote controllers in a TV manufacturing plant. Here, with an extensive development and experimental evaluation, nearly 100% reliability is achieved. Furthermore, selective visual processing allows inspection times of about 100ms.


Automated remote controller sorting system.




[Research]
        [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])
  • H. Yalcin, H. Isil Bozma,
    "An Automated Inspection System With Biologically Inspired Vision,"
    Proceedings of IROS'98 (IEEE/RSJ International Conference on Intelligent Robotic Systems),
    Victoria, Canada, October 1998.
    [Abstract] [ps] [pdf]
  • H. Yalcin, M. Tezol, H. Isil Bozma,
    "BUVIS - A Novel Approach to Real-Time Automated Visual Inspection,"
    submitted to Texas Instruments DSP Challenge '97 and received honor award as "runner up", 1997.
    [Abstract] [ps], [pdf]

    • Resume | Research | Main Page
    Carnegie Mellon University, Robotics Institute
    5000 Forbes Av., Pittsburgh, PA, 15213
    hulyayalcin@gmail.com