Occlusion is one of the most challenging issues in visual surveillance. In the real road surveillance systems, there are different kinds of objects to be tracked in real time. These objects may occlude each other, which makes their detection quite difficult. This paper proposes a simple but efficient occlusion-adaptive multi-object tracking approach to resolve this issue. Our approach considers three object states, including normal state, occluded state and split state. In the normal state, it tracks the object according to the spatial continuity. In the occluded state, it establishes a constant velocity model to estimate the position of an occluded object. In the split state, it rematches an object with one object before occlusion according to their appearance features. Experimental results show that our approach can correctly track multiple objects under both partial and total occlusion in real time.
The rationale of polarimetric optimization techniques is to enhance the phase quality of the interferograms by combining adequately the different polarization channels available to produce an improved one. Different approaches have been proposed for polarimetric persistent scatterer interferometry (PolPSI). They range from the simple and computationally efficient BEST, where, for each pixel, the polarimetric channel with the best response in terms of phase quality is selected, to those with high-computational burden like the equal scattering mechanism (ESM) and the suboptimum scattering mechanism (SOM). BEST is fast and simple, but it does not fully exploit the potentials of polarimetry. On the other side, ESM explores all the space of solutions and finds the optimal one but with a very high-computational burden. A new PolPSI algorithm, named coherency matrix decomposition-based PolPSI (CMD-PolPSI), is proposed to achieve a compromise between phase optimization and computational cost. Its core idea is utilizing the polarimetric synthetic aperture radar (PolSAR) coherency matrix decomposition to determine the optimal polarization channel for each pixel. Three different PolSAR image sets of both full- (Barcelona) and dual-polarization (Murcia and Mexico City) are used to evaluate the performance of CMD-PolPSI. The results show that CMD-PolPSI presents better optimization results than the BEST method by using either $D_{\mathrm{ A}}$ or temporal mean coherence as phase quality metrics. Compared with the ESM algorithm, CMD-PolPSI is 255 times faster but its performance is not optimal. The influence of the number of available polarization channels and pixel's resolutions on the CMD-PolPSI performance is also discussed.
Coal fire is a worldwide disaster that wastes massive energy and seriously pollutes the environment. Accurate acquisition of abnormal LST (land surface temperature) caused by underground coal fire is essential for coal fire monitoring and extinguishing. As a remote sensing technique, UAV (unmanned aerial vehicle) thermography can obtain LST images with very high spatial resolution and it has been used for coal fire monitoring. However, the accuracy of the UAV thermography obtained LST images (i.e., UAV LST images) has not yet been well studied, and the scale effect of UAV thermography for coal fire monitoring has not been discussed in previous studies. To this end, this study evaluates the accuracy of UAV LST images of coal fire areas based on the corresponding ground measurements. After that, the acquired UAV LST images are upscaled to different resolutions to simulate the LST images obtained at different observation scales. Finally, the local variance and Shannon entropy are employed to determine the optimal LST anomaly observation scale and coal fire area extraction scale. Baoan coalfield fire area, which is in Xinjiang province of China, is selected as the study area. The results show that the linear regression correlations R2 between UAV LST images and the LST values measured by thermal imaging camera and the infrared thermometer are both higher than 0.99. RMSE (Root Mean Square Error) between the thermal imaging camera LST measurements and that of UAV is 2.1 °C. When UAV LST images' resolutions are better than 7.5 m, most of the LST anomalies can be detected, and the LST anomaly information loss is relatively small (less than 17%). The resolution of 4 m is the required lowest resolution to accurately extract the areas of coal fire. When the resolution is lower than 4 m, the high-temperature abnormal boundaries caused by coal fire are being blurred, making the extraction of coal fire combustion areas unreliable.
The problem of accelerating metallic flyers to ultra high speed with strong detonating explosive slabs has been analyzed and numerically simulated in this paper, where the next stage slab is impacted by the flyer of previous stage and accelerates the next stage flyer to a higher speed. There is a high plateau in the detonation products pressure profile of the slab, to which the effective acceleration is attributed. A combination of impedance matched flyers of the final stage is impacted by the strong detonating explosive driven flyer at speed 6–7 km/s, and could be sped up over 10 km/s. This kind of high speed impactors have the advantages of simple structure, lower cost, smart design and promising in many applications of high dynamic pressure loading and high velocity impact.
With the great successful development of deep learning, stacked auto-encoder (SAE) has been widely used in POLSAR image terrain classification. In this paper, we propose a complex-valued Wishart stacked auto-encoder (CV-WSAE) classification model for POLSAR data interpretation. The proposed method stacks a complex-valued Wishart autoencoder (CV-WAE) and a complex-valued auto-encoder (CVAE) for feature extraction and connects a linear classifier for image classification. It not only expends real-valued neural network to complex-valued, but also utilizes the statistical distribution of POLSAR image. What is more, all elements of CV-WSAE including input, hidden, output, encoder and decoder layers are complex-valued, and a complex back propagation algorithm is used for training processing. The experiments of a real POLSAR data illustrate that this method can obtain good classification accuracy.
The environment management and land utilization of the abandoned mining region is critically dependent on precise monitoring and comprehensive understanding of mining subsidence. In order to overcome the shortcomings of the traditional distributed target phase optimization method in the space continuity constraints of adjacent pixels, an improved phase optimization algorithm was proposed, which combines region growing and time series interferometric synthetic aperture radar with distributed scatterers (DS-InSAR). By using 17 L-band and 51 C-band SAR images, the characteristics of temporal and spatial distribution in Peibei mining area of Xuzhou, China, were obtained during the period from 2007 to 2011 and the period from 2017 to 2020. With the long-term monitoring, the evolution of deformation in the mining area was carried out. The modified phase optimization technique has proven its ability in the density of measurement points and the influence of noise in space, which is promising for the detection of large gradient deformation and the accurate analysis of surface deformation in mining areas. The study has been concentrated toward detecting continuous subsidence in the mining region. Coal mines in operation are usually accompanied by unstable ground, and the uplift or second subsidence has sometime occurred in the closed mine region. Conclusively, the presented methodology is practically feasible for long-term deformation pattern analysis in coal-exhausted mining areas.
Differential synthetic aperture radar interferometry (D-InSAR) is limited when exploited in high-intensity mining areas, because large deformation gradients lie beyond the maximum measurable value of the D-InSAR technique which breaks the prerequisite for successfully employing of the method. The SAR amplitude-based pixel-tracking method provides an alternative way to efficiently and robustly extract the large deformation distribution particularly when the D-InSAR technique is limited by loss of coherence. In addition, the deformation in the line-of-sight direction and the deformation along the azimuth direction are also presented in this paper with 24-day interval repeat-pass high-resolution Rardarsat-2 imagery. Combining both of these techniques can help to better understand the deformation mechanisms associated with underground mining activities. The accuracies of 0.12 m in slant-range direction and 0.19 m in the azimuth direction were achieved, respectively. Besides, the profiles across the maximum deformation region have verified that the deformation occurred during two acquisition periods is far beyond the theoretical maximum deformation gradient corresponding to high-resolution C-band SAR data. The obtained surface motion infers to the mining activities and assessed damage caused by the large deformation.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)