GridTrack: Detection and Tracking of Multiple Objects in Dynamic Occupancy Grids

Abstract

Multiple Object Tracking is an important task for autonomous vehicles. However, it gets difficult to track objects when it is hard to detect them due to occlusion or distance to the sensors. We propose a method, "GridTrack", to overcome this difficulty. We fuse a dynamic occupancy grid map (DOGMa) with an object detector. DOGMa is obtained by applying a Bayesian filter on raw sensor data. This improves the tracking of the partially observed / unobserved objects with the help of the Bayesian filter on raw data, which has a powerful prediction capability. We develop a network to track the objects on the grid and fuse information from previous detections in this network. The experiments show that the multi-object tracking accuracy is high with the usage of the proposed method.

Estimation of Appearance and Occupancy Information in Birds Eye View from Surround Monocular Images

“…To obtain the occupancy tensor O with dimension |S|×grid X ×grid Y (where |.| indicates cardinality), we pass the above tensor of dimensions C × grid X × grid Y through the first three layers of the ResNet-18 [10] backbone, followed by two up-sampling layers. We follow the same paradigm for the appearance reasoning tensor A -passing the projected BEV tensor through the first three layers of ResNet-18 [14], followed by two up-sampling layers to get n c × grid X × grid Y tensor (where n c = #color channels).…”

“…We use them to label/tag the actors and track their temporal behavior. With steeringless and pedalless cars around the corner, [11], [12], [13], it is only natural that Multi-Object Tracking (MOT) [14], [15], [16] and language-based navigation [17], [18], [19], [20], [21] will be inevitable features for any future SDV. Complementary appearance cues present in the BEV occupancy space and the RGB image space provide strong priors for the above-mentioned tasks.…”

“…In MOT, existing work focuses mainly on costs derived from occupancy information such as position, shape, and orientation [14], [15], [16]. In [16], to exploit the appearance cues, the authors back-project different object instances in BEV into image space.…”