Thermal infrared pedestrian tracking via fusion of features in driving assistance system of intelligent vehicles

Abstract

Driving in nighttime or poor illumination is much more dangerous than in daytime or rich illumination conditions so it is important to develop driver assistance systems to support safe driving in nighttime. This paper presents a pedestrian tracking algorithm using a thermal infrared camera. The proposed algorithm first computes response maps with normalized grayscale and Felzenszwalb's histogram of oriented gradient in the framework of correlation filtering. Subsequently, according to response maps, this paper proposes an approach to estimate weight ratios in fusion. The fused response map is ultimately used to complete target detection in the framework of correlation filtering. In order to test the performance of the proposed tracker, extensive comparison and tests have been performed using open-source data sets. By testing through 10 different challenges in image processing and tracking, it is evident that the performance of proposed algorithm is superior to seven representative tracking algorithms with a quite promising real-time performance.

Thermal Infrared Single-Pedestrian Tracking for Advanced Driver Assistance System

“…Li et al presented a tracker which online selects the target-aware features [24]. Ding et al investigated different features including normalized grayscale and Histogram of Oriented Gradient (HOG) within the DCF framework [25]. The combination of the different layer features of VGGNet was used in the Multi-layer Convolutional Features for Thermal infrared tracking System (MCFTS) to represent the TIR target [26].…”

“…We set the original BACF tracker only using HOG feature as the baseline method. To further validate the superiority of the proposed fusion scheme, we also implement the tracker using the concatenation scheme that directly concatenate different feature maps along the direction of the channel, referred as concatenation [25].…”

Thirtieth anniversary edition and special issue on detect and avoid

“…14 The paper provided by M. Ding et al provides, finally, an insight on automotive DAA methodologies, by proposing an algorithm for pedestrian detection and tracking using a thermal infrared camera, stimulating cross-fertilization among aerospace and automotive domains in the field of obstacles detection and tracking. 15…”

A Novel Domain Transfer-Based Approach for Unsupervised Thermal Image Super-Resolution

“…In recent years, long-wavelength infrared (LWIR) images, a.k.a. thermal images, have shown to be useful to efficiently solve problems from different domains (e.g., security monitor [ 10 ], medical imaging [ 11 ], car assistance [ 12 ], visual inspection [ 13 ], and human detection [ 14 ], among others) because thermal images have the information of the radiation emitted by the surface of an object (temperature above zero [ 15 ]) captured by thermal cameras. As mentioned above, thermal cameras play an important role in different areas.…”