Object Detection and Recognition in Cluttered Scenes Using Fully Scale and In-Plane Invariant Synthetic Discriminant Function Filters

A demanding pattern recognition problem is to recognize objects despite distortions in position, orientation and scale in cluttered backgrounds. A system capable of detecting target objects despite any kind of geometrical distortion has many potential applications since it will be able to detect target objects when the orientation and position of the target or camera is unknown. In this work we report the use of fully invariant correlation filters for object detection in cluttered images despite any kind of geometrical distortion of the target object. A mapping technique is combined with the correlation filters, capable of creating invariance to various types of distortion of the target object. Synthetic Discrimination Function (SDF) based techniques provide a solution to the problem of invariant correlation filter design, expected distortions being included in the filter design. A log r-theta mapping is applied to the input image to give invariance to in-plane rotation and scale by transforming rotation and scale variations of the target object into vertical and horizontal shifts. The SDF filter is then trained using the logmapped image. A Maximum Average Correlation Height (MACH)based SDF filter is then employed to create invariance to orientation and gives good tolerance to background clutter and noise. The MACH filter is trained on the log r-theta map of the target for a range of orientations and applied sequentially over regions of interest in the input image. Areas producing a strong correlation response are then used to determine the position, in-plane rotation and scale of the target objects in the scene.