The appearance information of the target has been used as the only tracking cue for most trackers to locate the target in the video. However, when the surrounding environment changes drastically or there are similar interference targets, it usually causes target drift. We propose a tracker named MRTrack, a transformer-based spatiotemporal information de coupling network architecture to enhance the target tracking capability. We designed two training schemes to explore the effective integration of motion cues derived from optical flow with representation information. The first approach integrates the target's motion and representation information for training. The second scheme is a step-by-step training, where the target motion information is first learned, and the learned model is used for representation learning. We compare the two training methods on five generic tracking datasets. The experiment results indicate that the first training approach can better integrate motion and representation information, leading to more precise tracking results for MRTrack compared to solely relying on the appearance model. In addition, optical flow cues are used only in the training phase to guide the tracker in understanding motion information, and no additional cost is incurred during tracking inference.

The TBD (Tracking By Detection) framework has made significant contributions to the field of Multi-Object Tracking (MOT) through its flexible association strategy and modular design. In this paper, the FSC-MOT architecture is presented as a highly robust model for MOT building based on TBD. First, a novel association metric is introduced based on the results of the posterior bounding box. By step-by-step optimization and multistage integration, an efficient cost matrix is constructed without adding significant computational overhead. In addition, object data enhancement techniques are employed to address camera shake issues. The experimental results in the MOT17 and MOT20 datasets demonstrate that this method improves performance by 50% compared to BoT-SORT’s MOTA. In particular, the model achieve an impressive MOTA score of 80.7, indicating better tracking accuracy and robustness.

In modern security management, object detection in X-ray images plays a critical role in identifying hazardous items. However, this technology still faces significant challenges when dealing with complex scenarios, such as overlapping and intertwined objects. To address these issues, this paper proposes a novel network architecture called the Spatial and Channel Enhancement Detection Transformer (SCE-DETR). The model incorporates the Spatial and Channel Enhancement Module (SCEM), the Spatial and Channel Pruning Module (SCPM), and the Bilinear Downsampling Feature Fusion (BDFF) method. These innovations significantly improve the model’s ability to understand X-ray images, resulting in enhanced detection accuracy. Extensive experiments on the SIXray and CLCXray datasets demonstrate that SCE-DETR outperforms previous SOTA methods. Specifically, the model achieves an accuracy of 96.7% (AP50) on the SIXray dataset and 89.5% (AP50) on the CLCXray dataset. Compared to the baseline, SCE-DETR improves detection performance by 2.9% on SIXray and 3.4% on CLCXray.

Mud volcanoes have sparked significant current academic interest because their creation process is closely tied to groundwater. Mud volcano detection is crucial in Mars exploration, but current automatic detection technology is inadequate for this task. In this research, we address a gap in the field by building a novel and successful deep learning network that enables the first automated detection of Mars mud volcanoes. The technique can dramatically improve feature extraction and robustness. It is intended to take advantage of the integration of local binary features into a multi-branch dilated convolution framework. To facilitate the identification method and ensuing investigation, we present a novel scientific dataset of mud volcano images utilizing a single-group framework. In the experiment, we designed it to be compared to other state-of-the-art algorithms such as Atss, CascadeRCNN, Deformable Detr, Fcos, and so on, which are well-known for their ability to recognize tiny objects. The algorithm demonstrated superior performance over other algorithms, with an average accuracy increase of 10.4% and a 17.8% boost in recall rate.

Traditional Chinese medicines (TCMs) play an important role in the treatment of many diseases. For industrial production, classical TCMs identification methods suffer from high labor cost and low efficiency. Moreover the complex multi-object combinations of TCMs lead to serious feature confusion problem. In this article, we propose a novel detection network for TCMs called TCMnet. It focuses on the performance degradation caused by the images in different datasets containing different number of objects. First, an innovative multilevel feature fusion framework is proposed, which improves the generalization of the model. Then, a receptive field controlling architecture is established to limit the receptive field for reducing the confusion among multiple objects. Finally, a trainable feature resolution enhancement algorithm is proposed to increase the precision of classifier by enhancing local detail information. In the experiments, we choose 18 classes with 1800 images from our TCMs dataset. The experimental results show that TCMnet proposed in this article is able to mitigate the feature confusion problem in single-multiple object detection. In addition, TCMnet achieves a good accuracy compared with other detectors on single-object and multi-object detection tasks.

Siamese trackers usually use the target in the first frame as a fixed template, but the static template cannot adapt to target changes. The existing updater is challenging to deal with target deformation and update noise, and there is an excellent risk of updating with an inaccurate updater. In our research, a dynamic template updating strategy based on spatial-temporal information is proposed to improve the tracking accuracy of the Siamese tracker. Furthermore, Tracking Confidence Network (TCNet) is proposed to judge whether to update, which ensures that high-quality target features are used to update and reduce the noise caused by adding unreliable targets. In experiments, the proposed method is embedded into two baseline trackers: SiamRPN and SiamFC++, and tested on five popular benchmarks. The experimental results show that the proposed method can improve the performance of the Siamese trackers while maintaining real-time speed.

In the information age, image processing technologies play a vital role in the field of industrial engineering. In this article, a novel spline scheme called the hierarchical polishing splines algorithm is proposed, and it is applied to the field of computed tomography (CT) image reconstruction of coronavirus disease 2019 (COVID-19). The proposed algorithm defines a set of control lattices, wherein the density of lattice points in these control lattice ranges from coarse to fine in order, and the final applied function is produced by adding the functions derived from every control lattices. In order to demonstrate the performance of the proposed algorithm, some CT images from COVID-19 patients are selected. The experimental results show that the reconstructed COVID-19 CT images by using the proposed algorithm have good quality when compared with some widely used approaches. In addition, this article also applies the proposed algorithm to reconstruct the infected regions derived from COVID-19 CT images, and the results also show that the proposed algorithm is more efficient than others. Besides, the applicability of the proposed algorithm is discussed, wherein the COVID-19 severity is estimated based on the reconstructed COVID-19 CT images, and the applicability analysis shows that the use of the proposed algorithm to reconstruct COVID-19 CT images can help to achieve more accurate severity assessment of COVID-19 in a certain extent.

As the coronavirus disease 2019 (COVID-19) spreads around the world, industrial automated medical diagnosis systems have been developed, which complete a large amount of medical diagnosis work through computed tomography (CT) images. In these systems, how to quickly store and transmit such a large amount of CT image information has important research significance. In this article, a more targeted COVID-19 chest CT image codec is proposed to make image data not only occupy less space but also have higher image quality. First, the bilateral lung contours are extracted to calculate the position information of the region of interest (ROI). Then, a CT image is classified into four types of nonuniform image blocks according to the characteristics of COVID-19 chest CT images and ROI position information. Next, a series of new transformations are proposed for more efficient transform coding. Finally, a flexible quantization strategy is proposed for the adaptive quantization part. In the experiments, the proposed method is superior to some of the existing methods with similar computational complexity. At the same bit rate, it significantly improves the image quality. This means that chest CT images can still be used for disease diagnosis while taking up less space. In addition, because of the low computational complexity of the proposed method, it can be more easily embedded into the CT equipment with low computational power.

The lunar surface has complex geomorphic characteristics. Since the lunar terrain entropy can reflect the amount of geomorphic information contained in the lunar terrain, this article uses the ratio of the elevation value of a local point on the lunar surface to the total elevation value of the neighborhood to calculate the local terrain entropy value of the Moon. Then, the hierarchical polishing splines algorithm is proposed to construct the digital terrain entropy model (DTEM) of the Moon, wherein the new algorithm produces a sequence of functions based on a hierarchy of coarse-to-fine control lattices to generate the modeling function, which has good modeling performance. Using the proposed algorithm, multiscale DTEMs of the Moon are constructed based on square moving windows with different sizes. From the lunar DTEMs, it can be found that the lunar terrain entropy is sensitive to the size of the square moving window and the resolution of lunar DEM, and the high-resolution lunar DTEM with suitable moving window can well show topographical variations. In addition, the lunar terrain entropy distribution models are created based on the lunar DTEMs, which is significantly important to the study of the lunar terrain entropy distribution law. Besides, two terrain parameters, i.e., surface roughness and surface slope, are selected to show that the geomorphic characteristics of the Moon can be well reflected by the lunar terrain entropy.

Digital image feature recognition is significant to industrial information applications, such as bioengineering, medical diagnosis, and machinery industry. In order to supply an effective and reasonable technology of the severity assessment mission of coronavirus disease (COVID-19), in this article, we propose a new method that identifies rich features of lung infections from a chest computed tomography (CT) image, and then assesses the severity of COVID-19 based on the extracted features. First, in a chest CT image, the lung contours are corrected for the segmentation of bilateral lungs. Then, the lung contours and areas are obtained from the lung regions. Next, the coarseness, contrast, roughness, and entropy texture features are extracted to confirm the COVID-19 infected regions, and then the lesion contours are extracted from the infected regions. Finally, the texture features and V-descriptors are fused as an assessment descriptor for the COVID-19 severity estimation. In the experiments, we show the feature extraction and lung lesion segmentation results based on some typical COVID-19 infected CT images. In the lesion contour reconstruction experiments, the performance of V-descriptors is compared with some different methods, and various feature scores indicate that the proposed assessment descriptor reflects the infected ratio and the density feature of the lesions well, which can estimate the severity of COVID-19 infection more accurately.

Surface roughness is a land-surface parameter that is widely used in terrain analysis. Some typical roughness details, which have important effects on surface analysis, fail to be characterized on previous roughness maps. The objective of this paper is to provide a more accurate small-to-large scale roughness overview. The new roughness method is designed based on a complete orthogonal system called the V-system. The V-system roughness utilizes the special functions to detect and extract the roughness characteristics from high-resolution digital elevation models (DEMs). In this paper, Lunar Orbiter Laser Altimeter-derived DEMs are used as the source data for the roughness calculation. Compared with the global root-mean-square slope and Fourier-based roughness maps, the V-system roughness maps show that more typical roughness details have been added to clearly indicate the small roughness variations on the large map. Furthermore, the reliability and practicability of V-system roughness are demonstrated based on the multiresolution DEMs. As an example, the statistical parameters of the roughness characteristics in the lunar Maria and highlands identify the fact that the highlands are rougher at all scales than the Maria. And this difference corresponds to the basic roughness property.