Active Visual Servo Based On Bionic Multi-feature Fusion Recognition and Distance Estimation

In the complex natural background, the image features of spatial dynamic objects usually change severely, so target recognition method based on single feature could not adapt to the recognition requirements. Due to unpredictable changes of target distance, finding real-time solution of image Jacobi matrix is more difficult, so as achieving active visual servo of spatial dynamic target. Inspired by the visual characteristics of frog eye and bionic recognition mechanism, a bionic spatial-temporal fusion recognition algorithm of dynamic target was provided, and a distance estimation method based on monocular vision information was designed, finally an active visual servo system adapted to three-dimensional tracking applications was established. The physical experiment results show that the bionic recognition method inhibited the background information effectively and enhanced moving target information in multidimension, which is better than the method based on single feature. By using the target distance estimation method, the system solved calculation problem of active vision image Jacobi matrix and fulfilled the requirements of active three-dimensional visual servo.