Automatic Interpretation of Outdoor Scenes.

Abstract

This paper describes recent work on a neural network approach to outdoor scene interpretation. The results of evaluating a range of automatic region-based segmentation techniques based on a new segmentation quality metric are presented. The optimal technique is used to segment images of natural outdoor scenes. A powerful set of features designed for outdoor scene analysis is extracted from regions in the segmented images and used to train a neural network to recognise eleven different classes of objects, including sky, road, building and vegetation. The system is tested on a large number of images which have been hand-labelled to provide ground-truth segmentations and interpretations. This fully automated system achieves an impressive mean classification accuracy of 81.4% per image on unseen data, and runs in approximately 1 CPU minute on a Sun SPARCstation 20.

Automatic Segmentation and Classification of Outdoor Images Using Neural Networks

“…This technique is the most successful closed-region segmentation technique which w e h a v e found, outperforming the K-means (58.9%), Marr{ Hildreth (21.2%) and region-growing techniques (41.1%) described in Ref. 6.…”

“…In previous work 6 we h a v e shown a method by which, given a ground-truth segmentation for an image such a s i s a v ailable as part of the Bristol Image Database, a quantitative measure of its similarity t o a m a c hine-generated segmentation may b e calculated. This similarity measure computes the fraction of pixels which are not inside both the region to be measured and its`ideal' counterpart.…”

A Scene Images Classification Method Based on Local Binary Patterns and Nearest-Neighbor Classifier

“…Texture could also play an important role in this kind of application, because it is much more robust than color with respect to changes in illumination [3]. Setchell and Campbell [4] used Gabor texture features for classifying pre-labeled regions from images in the Bristol Image Database [5]. Castano et al assessed the performance of two Gabor filtering based texture classification methods on a number of real-world images relevant to autonomous navigation on crosscountry terrain [3].…”

Automatically Locating an Area of Interest and Maintaining a Reference Image to Aid the Real-Time Tracking of Objects

“…Campbell, et. al., [2] use neural networks to automatically recognise features in a scene, such as road, grass, etc. This approach could be used to identify the area of interest in a scene, such as the road in a traffic tracking application, though this was not the purpose of their system.…”