Refine
Document type
Is part of the Bibliography
- Yes (8)
Keywords
- Computer vision (3)
- Lane detection (2)
- Problem-centered interviews (2)
- Traffic scenarios (2)
- ADAS (1)
- Advanced driver assistance systems (1)
- Cameras (1)
- Crosswalk detection (1)
- Decision support systems (1)
- Driver assistance (1)
- Informatics (1)
- Machine intelligence (1)
- Navigation (1)
- Road segmentation (1)
- Software engineering (1)
- System integration (1)
- Vision use cases (1)
- Visual impairment (1)
- Visualization (1)
- Visually impaired (1)
Herein, we research the possibilities of assisting visually impaired pedestrians moving in traffic situations by using camera-based detection of relevant objects in their immediate surroundings. Therefore, we use and adapt algorithms from the field of driver assistance. We present a road background segmentation based on watersheds, whose results are used as input for the presented crosswalk and lane detection algorithms. The crosswalk detection is based on the application of two 1D mean filters and the lane detection on local computations of the EDF (Edge Distribution Function). In our evaluation, the described algorithms achieved good hit rates of 99.87 % (road segmentation), 98.64 % (crosswalk detection), and 97.89 % (lane detection).
We introduce an algorithm that performs road background segmentation on video material from pedestrian perspective using machine learning methods. As there are no annotated data sets providing training data for machine learning, we develop a method that automatically extracts road respectively background blocks from the first frames of a sequence by analyzing weights based on mean gray value, mean saturation, and y coordinate of the block’s middle pixel. For each block labeled either road or background, several feature vectors are computed by considering smaller overlapping blocks within each block. Together with the x coordinate of a block’s middle pixel, mean gray value, mean saturation, and y coordinate form a block’s feature vector. All feature vectors and their labels are passed to a machine learning method. The resulting model is then applied to the remaining frames of the video sequence in order to separate road and background. In tests, the accuracy of the training data passed to the machine learning methods was 99.84 %. For the complete algorithm, we reached hit rates of 99.41 % when using a support vector machine and 99.87 % when using a neural network.