Overview
The beneficiary of our services is a company part of the top 100 Forbes, with more than 100 billions EUR in revenues. They are world leaders in Consumer Electronics, Home Appliances, Home Entertainment, Mobile Communications, Solar and Display Technologies.
The challenge was to develop an object detection system for automotive use. This application should identify obstacles and evaluate distances to obstacles, using only one video source from a moving camera.
The targeted platform is a mobile and embedded hardware with a monocular vision that should be power efficient in real time.
Our solution
We developed a solution that uses sensor information for ground plane estimation:
- We aggregated obstacle information from multiple methods, using different algorithms
- All results were incorporated in one final output
- We solved a broad range of issues: input stabilization (vibration), image registration, dynamic object identification
- Our solution involved a lot of parallelization and optimization work, for GPUs, SIMD/MIMD processors
- It is able to work in different environments (day/night, sunny/cloudy etc)
Gallery
