Alexander L. Honda

A neuromorphic system for object detection and classification

Unattended object detection, recognition and tracking on unmanned reconnaissance platforms in battlefields and urban spaces are topics of emerging importance. In this paper, we present an unattended object recognition system that automatically detects objects of interest in videos and classifies them into various categories (e.g., person, car, truck, etc.). Our system is inspired by recent findings in visual neuroscience on feed-forward object detection and recognition pipeline and mirrors that via two main neuromorphic modules (1) A front-end detection module that combines form and motion based visual attention to search for and detect “integrated” object percepts as is hypothesized to occur in the human visual pathways; (2) A back-end recognition module that processes only the detected object percepts through a neuromorphic object classification algorithm based on multi-scale convolutional neural networks, which can be efficiently implemented in COTS hardware. Our neuromorphic system was evaluated using a variety of urban area video data collected from both stationary and moving platforms. The data are quite challenging as it includes targets at long ranges, occurring under variable conditions of illuminations and occlusion with high clutter. The experimental results of our system showed excellent detection and classification performance. In addition, the proposed bio-inspired approach is good for hardware implementation due to its low complexity and mapping to off-the-shelf conventional hardware.

Robust static and moving object detection via multi-scale attentional mechanisms

Real-time detection of objects in video sequences captured from an aerial platforms is a key task for surveillance applications. It is common to perform expensive frame to frame registration as preprocessing to moving object detection in this type of application, and there is no principled approach to the detection of stationary targets.We explore the Spectral Residual algorithm,6 a fast linearithmic run time saliency model which requires no training and has no temporal dependencies, and is capable of detecting proto-objects in a single image. In this paper we describe methods for enhancing the Spectral Residual saliency algorithm to generate candidate object detections from video sequences captured from moving platforms. These object candidates can then be passed to a classification module for further processing. We describe a method that makes the Spectral Residual algorithm more robust to natural variances in color images, and a pyramidal approach to make the processes more robust to objects of varying size. Furthermore we describe a technique for processing the resulting saliency map into a set of tight bounding boxes suitable for extracting image regions for classification. These methods result in a system that is fast, robust, and efficient with reliable performance for low SWaP surveillance platforms.