In recent years, the excellent image-based object detection algorithms are transferred to the video object detection directly. These frame-by-frame processing methods are suboptimal owing to the degenerate object appearance such as motion blur, defocus and rare poses. The existing works for video object detection mostly focus on the feature aggregation at pixel level and instance level, but the blur impact in the aggregation process has not been exploited well so far. In this paper, we propose an end-to-end blur-aid feature aggregation network (BFAN) for video object detection. The proposed BFAN focuses on the aggregation process influenced by the blur including motion blur and defocus with high accuracy and little increased computation. In BFAN, we evaluate the object blur degree of each frame as the weight for aggregation. Noteworthy, the background is usually flat which has a negative impact on the object blur degree evaluation. Therefore, we introduce a light saliency detection network to alleviate the background interference. The experiments conducted on the ImageNet VID dataset show that BFAN achieves the state-of-the-art detection performance, exactly 79.1% mAP, with 3 points improvement compared to the video object detection baseline.
17 August 2023 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
As a significant research direction in remote sensing fields, unmanned aerial vehicles (UAVs) tracking has achieved rapid development in recent years. However, due to limited power and computation resources on aerial platforms, the tracking methods deployed on UAVs usually require high computational efficiency and performance. In addition, various challenges (i.e., similar object, background clutter and occlusion) have inevitably occurred during the UAV tracking phase. Therefore, considering the above issues comprehensively, this paper proposes a dynamic aberrance-repressed temporal regularized correlation filter to achieve stable tracking in UAV remote sensing videos. First, we have introduced the aberrance-repressed temporal regularizations into the discriminative correlation filter (DCF) framework. Second, a novel objective loss function is constructed to adjust the strength of each regularization for training the filter. Then, a new judgment mechanism based on the response variation is exploited to reflect the response fluctuation and applied to tune parameters of both regularizations. Finally, comprehensive experiments are done on three different UAV benchmarks, i.e., UAV123@10fps, UAVDT and VisDrone2018, to verify the performance of our tracker and have demonstrated that our tracker achieves superior performance against other total 25 state-of-the-art trackers while reaching $\sim$ 35 FPS on a single CPU.
In recent years, image deblurring has been widely investigated. In order to solve this ill-posed problem, a variety of prior models have been proposed successively. The two-tone prior has been successfully applied to text images deblurring and achieved significant results. However, the natural image with rich color clusters does not meet the two-tone prior, which requires only two color clusters in the image. In this paper, a local two-tone prior is proposed for images with complex color clusters, which decomposes the image with complex color clusters into patches with simple color clusters. We also find that the process of image blurring is a weighted average of pixel values, which will lead to an increase of intermediate pixel values. Therefore, a new measure of the dynamic range of the image is proposed, which indicates the difference of the color cluster using the average value of the color cluster, and it is rigorously proved in mathematics. Besides, we analyze the effectiveness of the proposed prior in image deblurring in detail. Experimental results on the widely used datasets show that the proposed method performs favorably against the state-of-the-art algorithms, both qualitatively and quantitatively.
Motion blur is an inevitable problem when shooting a video in motion, and it will cause serious degradation of images. In this letter, we propose a new robust incremental weighted Fourier aggregation algorithm for dynamic scene video deblurring. This is motivated by the fact that existing multiframe aggregation methods usually require a tedious iterative process to obtain sharp information in distant frames. We propose to perform alignment and aggregation once for each frame using the current frame and the previous frame result, which includes the sharp information of all the previous frames. Moreover, a criterion for evaluating the alignment of blurred images is proposed. Based on this, a robust aggregation approach is also proposed to help reduce artifacts caused by image patch alignment failure. Specifically, we decompose the current frame into a series of image patches, find consistent image patches in the deblurring results of the previous frame by a coarse-to-fine alignment algorithm, and aggregate them in the Fourier domain. Experiments show that our method can obtain improved deblurring results and reduce the computational complexity compared to the traditional aggregation method.
Abandoned object detection is a crucial problem in many computer vision tasks. Traditional method based on foreground/background extraction techniques leads to a high false alarm rate. In this paper, we propose a novel detection algorithm based on change detection and blob separation. Our proposed approach suits more practical scenarios in which objects located near each other. Experiments are conducted on several well-known benchmarks to validate the efficacy of the algorithm proposed in this paper.
Although learning-based multi-view stereo algorithms have produced exciting results in recent years, few researchers have explored the specific role of deep sampling in the network. We posit that depth sampling accuracy more directly impacts the quality of scene reconstruction. To address this issue, we proposed NTPP-MVSNet, which utilizes normal vector and depth information from neighboring pixels to propagate tangent planes. Based on this, we obtained a more accurate depth estimate through homography transformation. We used deformable convolution to acquire continuous pixel positions on the surface and 3D-UNet to account for the regression of depth and normal vector maps without consuming additional GPU memory. Finally, we applied homography transformation to complete the mapping of the imaging plane and the neighborhood surface tangent plane to generate a depth hypothesis. Experimental trials on the DTU and Tanks and Temples datasets demonstrate the feasibility of NTPP-MVSNet, and ablation experiments confirm the superior performance of our deep sampling methodology.