This paper presents a new computer interface system based on laser spot detection and moving pattern analysis of the detected laser spots in real-time processing. We propose a systematic method that uses either the frame difference of successive input images or an autoassociative multilayer perceptron (AAMLP) to detect laser spots. The AAMLP is applied only to areas of the input images where the frame difference of the successive images is not effective for detecting laser spots. In order to enhance the detection performance, the AAMLP is trained by a new training algorithm that increases the sensitivity of the input-to-output mapping of the AAMLP allowing a small variation in the input feature of the laser spot image to be successfully indicated. The proposed interface system is also able to keep track of the laser spot and recognize gesture commands. The moving pattern of the laser spot is recognized by using a multilayer perception. It is experimentally shown that the proposed computer interface system is fast enough for real-time operation with reliable accuracy.
In this paper, we present a multi-vehicle tracking method that uses integrated position and motion tracking methods to minimize missing and false detection. No existing state-of-the-art vehicle detection method can detect all the vehicles on the road and remove all false positive alarms. Therefore, a robust tracking-by-detection algorithm is necessary to minimize the number of false positive and false negative alarms. In multi-vehicle tracking, there are three types of errors such as false negative alarms, false positive alarms, and track identity switches. False negative and false positive alarms are caused by an imperfect detection algorithm, while track identity switches are caused by measurement-to-track pair confusion. Our tracking-by-detection method minimizes these errors while processing in real-time for online application. Sparse false positive alarms are reduced by a track initialization procedure. Motion tracking with selected features can minimize false negative alarms. A data association algorithm with complementary global and local distance prevents tracks from connecting measurements incorrectly. The proposed method was evaluated and verified in challenging, real road environments. The experimental results demonstrate that our multi-vehicle tracking method remarkably reduces false positive and false negative alarms and performs better than previous methods.
Knowledge-based clustering and autonomous mental development remains a high priority research topic, among which
the learning techniques of neural networks are used to achieve optimal performance. In this paper, we present a new
framework that can automatically generate a relevance map from sensory data that can represent knowledge regarding
objects and infer new knowledge about novel objects. The proposed model is based on understating of the visual what
pathway in our brain. A stereo saliency map model can selectively decide salient object areas by additionally considering
local symmetry feature. The incremental object perception model makes clusters for the construction of an ontology map
in the color and form domains in order to perceive an arbitrary object, which is implemented by the growing fuzzy
topology adaptive resonant theory (GFTART) network. Log-polar transformed color and form features for a selected
object are used as inputs of the GFTART. The clustered information is relevant to describe specific objects, and the
proposed model can automatically infer an unknown object by using the learned information. Experimental results with
real data have demonstrated the validity of this approach.
The noise problem, such as the fixed pattern noise (FPN) due to the process variation, should be considered when designing a vision chip. In this paper, we proposed an edge detection circuit based on biological retina using an offset-free column readout circuit (OFCRC) to reduce the FPN occurring in the photo-detector. The OFCRC consists of one source follower, one capacitor and five transmission gates. Thus, it is simpler than a conventional correlated double sampling (CDS) circuit. A vision chip for edge detection has been designed and fabricated using a 0.35μm 2-poly 4-metal CMOS process and its output characteristics have been investigated.
Numerical increment of analog circuits causes power consumption to increase and requires a larger chip area. In designing an analog complementary-metal-oxide-semiconductor (CMOS) vision chip for edge detection, power consumption should be considered. It restricts the number of the edge detection circuit which is based on the edge detection mechanism of vertebrate retina. In this paper, we applied electronic switches to an analog CMOS vision chip for edge detection to reduce the power consumption. Also, we propose a method to implement vision chip with higher resolution, which is to separate pixels for edge detection into a 128×128 photodetector array and a 1×128 edge detection driving circuit array. The capability to minimize power consumption was investigated by SPICE. Estimated power consumption with 128×128 pixels was below 20mW.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.