We present a Multi-perspective Tracking (MPT) framework for intelligent vehicle. An iterative search procedure is proposed to associate detections and tracklets in different perspectives.
The multi-camera array has drawn attention ofresearchers in recent years, and has been configured anddeployed on intelligent vehicle to capture the panoramic views.Understanding surroundings is crucial for the ego-vehicle. Thispaper presents a Multi-perspective Tracking (MPT) frameworkfor intelligent vehicle. An iterative search procedure is proposedto associate detections and tracklets in different perspectives.This procedure iteratively assigns determined states and estimatesnon-determined states for the detections and tracklets.An inherent determined and non-determined graph is utilized toreinforce this procedure. For more reliable associations betweenperspectives, a Siamese convolutional neural network is employedto learn feature representation. The supervised classificationand verification signals are added to train the network. Thefeatures in different conventional stages are integrated togetheras the discriminative appearance model. The experiments areconducted on a MPT data set with five perspectives. The proposedframework is tested in each pair of adjacent perspectives for theability to associate target objects between perspectives.