摘要 |
PROBLEM TO BE SOLVED: To provide a driving scene label generation device which estimates a scene label fitted to a divided driving scene.SOLUTION: A driving scene discretizing unit 3 divides a time sequence of driving data supplied from a driving behavior data collection unit 2, into partial sequences each representing some driving scene by a double articulation analyzer (DAA). A feature quantity distribution generation unit 4 generates, for each partial sequence, an analysis object distribution representing features of driving behavior data and image data, and a topic proportion calculation unit 6 calculates topic proportions representing the proportions of driving topics (base feature distributions) in the analysis object distribution. An estimation distribution setting unit 7 sets a distribution for label estimation on the basis of the topic proportions and base tag distributions prepared per base feature distribution, and a scene label estimation unit 8 estimates a scene label fitted to the driving scene by using the set distribution for label estimation. |