Features Extractors Evaluation Based V-SLAM Applications

This paper presents a study related to feature extractors (detectors and descriptors) for V-SLAM applications. A review is performed with regards to well-known and optimized features descriptors like ORB, BRIEF, SURF, ZEON and bio-inspired descriptors like FREAK or HOOFR. Combinations with different widely used corner detectors like FAST, AGAST, STAR and SURF is also evaluated in terms of execution time resulting for both detection and description tasks. Results are analyzed to consider only the best combination for an automotive application. Evaluation results highlight the bio-inspired approaches considering the compromise between image processing accuracy and execution times.