Search

Small object detection is one of the main obstacles hindering the development of object detection technology. In small object detection tasks, the performance of universal object detectors often drops sharply. We find through experiments on common models that for small objects, the box prior based on anchor detectors and the point prior without anchor detectors are suboptimal. The current anchor based or anchor-free label allocation methods will generate many false positive small objects samples, resulting in a decrease in detector attention to small objects. Therefore, we propose a new detector CLAHR for small object detection. In response to the issue of high sensitivity to small objects using Intersection over Union (IoU), where small deviations can lead to poor quality of assigned positive and negative samples, CLAHR utilizes the insensitivity of Gaussian distribution to small perturbations of small objects to transform the predicted box and manually annotated true box into Gaussian distribution. Then, CLAHR uses Wasserstein Distance (WD) to measure the distance between the predicted actual receptive field (ARF) and the manually labeled true receptive field (TRF) of the real box, rather than using IoU or central sampling strategies to allocate samples. Considering that IoU threshold based and central sampling strategies tend to favor large objects, we further design a cascaded label assignment (CLA) module to achieve balanced learning for small objects samples. Extensive experiments on the COCO, AI-TOD, and VisDrone datasets demonstrate the effective-eness of the proposed approach.

Person re-identification (Re-ID) is an identification method based on the overall characteristics of human body, which is usually used to judge whether there is a specific person in the image or video sequence. The clothing characteristics of persons have very important impact on Re-ID methods. But in some special scenarios, for example, a criminal maybe change his clothes in order to escape tracking, then his clothes will be different in the query image from those in the gallery image, which will cause the existing Re-ID algorithms difficult to match the target person in the gallery image. In order to solve this problem, a Re-ID model based on clothing information transfer is proposed in this paper. Given a query image of a target person, all the persons are re-dressed in the gallery image with the clothing feature in the query image, therefore the difference of clothing feature in person matching process can be eliminated. Finally, the model is verified on the dataset PRDDC (person Re-ID with different clothes), and experimental results show the significant improvement of the proposed algorithm compared with the algorithms without clothing feature transfer module.

It studied the explosion propagation of the hybrid mixture of LDPE dust and a small amount of ethylene (below its LEL) in a closed duct, to examine the impacts of dust concentration, dust particle size, and ethylene concentration on flame propagation behavior and overpressure characteristics. The dust concentration affected the explosion behavior of the LDPE dust/ethylene hybrid mixture in two ways. The flame gradually became discrete and unstable as the dust concentration increased, while the maximum and average flame propagation velocity, Pmax, and (dp/dt)max increased first and then decreased. The dust particle size and ethylene concentration had a significant effect on the hybrid mixture explosion. As the particle size decreased or the ethylene concentration increased, the discrete flames quickly merged into a bright continuous flame. Further, the maximum and average flame propagation velocity, Pmax, and (dp/dt)max increased, whereas the fluctuation degree of the flame propagation velocity with time showed a decreasing trend. There was no obvious agglomeration phenomenon among small LDPE particles. The explosion overpressure and flame propagation showed a strong synergistic effect. Compared with the semi-open duct, the closed duct with a more uniform dust/gas hybrid mixture showed a more continuous flame structure, a shorter flame propagation distance, and a lower flame propagation velocity. These conclusions can provide a scientific basis for explosion risk analyses and process safety designs involving the production of LDPE powder. [ABSTRACT FROM AUTHOR]