Application Study of Blind Spot Monitoring System Realized by Monocular Camera with CNN Depth Cues Extraction Approach 12-02-04-0016

The image from monocular camera is processed to detect depth information of the obstacles viewed by the rearview cameras of vehicle door side. The depth information recognized from a single, two-dimensional image data can be used for the purpose of blind spot area detection. Blind spot detection is contributing to enhance the vehicle safety in scenarios such as lane-change and overtaking driving. In this article the depth cue information is inferred from the feature comparison between two image blocks selected within a single image. Convolutional neural network model trained by deep learning process with good enough accuracy is applied to distinguish if an obstacle is far or near for a specified threshold in the vehicle blind spot area. The application study results are demonstrated by the offline calculations with real traffic image data.