Intelligent Real-Time Deep System for Robust Objects Tracking in Low-Light Driving Scenario

The detection of moving objects, animals, or pedestrians, as well as static objects such as road signs, is one of the fundamental tasks for assisted or self-driving vehicles. This accomplishment becomes even more difficult in low light conditions such as driving at night or inside road tunnels. Since the objects found in the driving scene represent a significant collision risk, the aim of this scientific contribution is to propose an innovative pipeline that allows real time low-light driving salient objects tracking. Using a combination of the time-transient non-linear cellular networks and deep architectures with self-attention, the proposed solution will be able to perform a real-time enhancement of the low-light driving scenario frames. The downstream deep network will learn from the frames thus improved in terms of brightness in order to identify and segment salient objects by bounding-box based approach. The proposed algorithm is ongoing to be ported over a hybrid architecture consisting of a an embedded system with SPC5x Chorus MCU integrated with an automotive-grade system based on STA1295 MCU core. The performances (accuracy of about 90% and correlation coefficient of about 0.49) obtained in the experimental validation phase confirmed the effectiveness of the proposed method.

Auto-attentional mechanism in multi-domain convolutional neural networks for improving object tracking

PurposeMulti-domain convolutional neural network (MDCNN) model has been widely used in object recognition and tracking in the field of computer vision. However, if the objects to be tracked move rapid or the appearances of moving objects vary dramatically, the conventional MDCNN model will suffer from the model drift problem. To solve such problem in tracking rapid objects under limiting environment for MDCNN model, this paper proposed an auto-attentional mechanism-based MDCNN (AA-MDCNN) model for the rapid moving and changing objects tracking under limiting environment.Design/methodology/approachFirst, to distinguish the foreground object between background and other similar objects, the auto-attentional mechanism is used to selectively aggregate the weighted summation of all feature maps to make the similar features related to each other. Then, the bidirectional gated recurrent unit (Bi-GRU) architecture is used to integrate all the feature maps to selectively emphasize the importance of the correlated feature maps. Finally, the final feature map is obtained by fusion the above two feature maps for object tracking. In addition, a composite loss function is constructed to solve the similar but different attribute sequences tracking using conventional MDCNN model.FindingsIn order to validate the effectiveness and feasibility of the proposed AA-MDCNN model, this paper used ImageNet-Vid dataset to train the object tracking model, and the OTB-50 dataset is used to validate the AA-MDCNN tracking model. Experimental results have shown that the augmentation of auto-attentional mechanism will improve the accuracy rate 2.75% and success rate 2.41%, respectively. In addition, the authors also selected six complex tracking scenarios in OTB-50 dataset; over eleven attributes have been validated that the proposed AA-MDCNN model outperformed than the comparative models over nine attributes. In addition, except for the scenario of multi-objects moving with each other, the proposed AA-MDCNN model solved the majority rapid moving objects tracking scenarios and outperformed than the comparative models on such complex scenarios.Originality/valueThis paper introduced the auto-attentional mechanism into MDCNN model and adopted Bi-GRU architecture to extract key features. By using the proposed AA-MDCNN model, rapid object tracking under complex background, motion blur and occlusion objects has better effect, and such model is expected to be further applied to the rapid object tracking in the real world.

Re-Identification with Pre-Filtering by Image Quality for Multiple Objects Tracking

The paper considers multiple object tracking. Existing methods tend to be either resource-intensive or prone to high object densities errors failing to provide competitive performance at high frame rates without significant tracking disruptions and error accumulation. We formulate the multiple object tracking problem under the assumption of linearity and independence of the movement of objects. The factorization of the posterior distribution of objects' parameters provides proof of the equivalence of the initial problem and the tracking procedure containing two subtasks: track prediction and assignment of measurements and objects. A modification of the assignment cost is introduced to achieve the stability of assignments in challenging scenarios of tracking, such as multiple objects occlusions and missing detections. We consider adding a term that states to re-identification of the candidate by comparing its descriptor with descriptors from the track history. Given that track measurements are not equal in terms of usefulness for re-identification, we introduce the technique of track descriptor pre-filtering based on quality assessment in order to select the most relevant descriptors for re-identification and reduce method algorithmic complexity. Both known quality assessment methods and an alternative detector-based approach are taken into account. Computational experiments were conducted on MOT20-01, MOT20-02 datasets containing CCTVcameras data in order to compare the proposed method with other approaches. The results showed the computational efficiency of the proposed methods and the increased stability of tracking in complex scenarios.