Multisized Object Detection Using Spaceborne Optical Imagery
Movable object detection in aerial or satellite imagery is of great practical interest owing to its variety of applications in numerous fields including traffic monitoring, airport surveillance, parking lot analysis, search and rescue, determining transportation infrastructure etc. However, the problem is highly challenging since such remote sensing images are acquired from high altitudes causing atmospheric distortions, illumination and viewpoint variations, partial occlusions, and clutter (especially in urban environments). Moreover, objects when viewed from an elevated platform (satellite, drone etc.) present a difficult to understand prospect and arbitrary orientation, which subsequently leads to sub-optimal performance of machine (deep) learning models trained over datasets containing terrestrial object images.
We develop the methodology to addresses the highly challenging problem of vehicle detection from high resolution remote sensing imagery by introducing a novel medium size annotated dataset named Satellite Imagery Multi-vehicles Dataset (SIMD) along with an adapted single pass deep multi-scale object detection framework with the aim to detect multi sized/type objects for catering above-ground perspective of vehicles. The dataset images are acquired from multiple locations in the EU/US regions available in the public Google Earth satellite imagery. Specifically, it comprises 5,000 images of resolution 1024 x 768 and collectively contains 45,096 objects in 15 different classes of vehicles including cars, trucks, buses, long vehicles, various types of aircrafts and boats. In the proposed architecture, we demonstrate the relevant modifications needed to translate the state-of-the-art object detection frameworks to solve the object detection problem from remote sensing imagery. The proposed architecture has been evaluated on SIMD and a public dataset VEDAI. The comparative analysis has been performed with existing off-the-shelf single-shot object detection models including YOLO and YOLT yielding superior performance measured with standard evaluation strategies.
Figure : Proposed CNN architecture with details of layers arrangement, up sampling, concatenation, and detection levels. It is shown that our model detects multiscale objects from same input image in single pass using various multilevels.
- Dr Muhammad Shahzad
- Dr Muhammad Moazam Fraz
- Muhammad Haroon
- M. Haroon, M. Shahzad, M.M. Fraz , “Multi-sized Object Detection Using Spaceborne Optical Imagery”, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (JSTARS), Vol. 13 , Jun 2020. pages 3032 – 3046. IF: 3.392