As shown on the Multiple Object Tracking(MOT) Benchmark (above), Dahua Technology TRK Method set a world record and took the 1st place in  a multiple pedestrian tracking challenge on August 2nd, 2018 - with a multi-object tracking accuracy(MOTA) of 54.1±13.0%.

Object tracking technology is a core technology of video structured analysis. The principle is to use the initial target positioning information to complete the association of the same target in the time series, and complete the behaviour discrimination and analysis according to the associated information. Multiple object tracking technology handles trajectory association of all target objects in a video sequence.

To prepare for this competition, Dahua Technology has upgraded the deep learning multiple object tracking algorithm frame based upon the advantages of target detection, face recognition, search by image and other algorithms. Using current methods such as feature reuse, together with multi-model fusion technology, the tracking performance has been greatly improved. Better solving problems such as obscuration, deformation, light changes, random motion, excessive movement, Dahua Technology’s object tracking technology has been widely applied to newly launched intelligent products, especially cameras, NVRs and servers that are based upon video data structured algorithms.