BROWSE
Detail
Conference Paper
Low Environmental Impact Object Recognition: Adaptive Process and Recognition Based Segmentation
Cited - time in 
Share 
Share
- Authors
- Kyekyung Kim, Sangseung Kang, Jaehong Kim
- Issue Date
- 2016-12
- Citation
- International Conference on Big Data Applications and Services (BigDAS) 2016, pp.113-117
- Language
- English
- Type
- Conference Paper
- Abstract
- In this paper, we describe object recognition adapts to environmental changes that is invariant to illumination effects and object location, object rotation or object type changes. Object recognition highly depends on environment of consciousness, especially, the object recognition has low performance in a real production field. Two types of object recognition error are appeared as follows: Illumination effects cause image distortion, loss of image data or noise insertion that result in object segmentation error and recognition. Another error has occurred by randomly located objects and various type of object. Therefore, we present object recognition method is less sensitive and adaptive to environment changes. A complemented image processing, recognition based object segmentation, combined feature extraction, scale or rotation invariant object recognition with neural network are used to enhance segmentation and recognition performance. We have experiment with ETRI database and obtained 98.7% recognition rate using 2,160 images. Experimental image data have captured under unconstrained environment such as uncontrolled lighting condition.
- KSP Keywords
- Adaptive process, Combined features, Environmental change, Environmental impact(EI), Feature extractioN, Image Distortion, Image data, Image processing, Invariant object recognition(IOR), Lighting condition, Low environmental impact