A decision tree algorithm was developed to classify the freeze/thaw status of the surface soil based on the cluster analysis of samples such as frozen soil,thawed soil,desert and snow,along with microwave emission and scattering characteristics of the frozen/thawed soil.The algorithm included five SSM/I channels(19V,19H,22V,37V,85V)and three crucial indices including scattering index,37GHz vertical polarization brightness temperature and 19GHz polarization difference,and took into consideration the scattering effect of desert and precipitation.The pureness of samples is essential to the analysis of the microwave brightness temperature characteristics,which is prior to deciding the thresholds of each node of the decision tree.We have selected four types of samples,including frozen soil,thawed soil,desert and snow.The frozen soil has some special microwave emission and scattering characteristics different from the thawed soil:① lower thermodynamic temperature and brightness temperature;② higher emissivity;③ stronger volume scattering,and the brightness temperature decreased with increasing frequency.The threshold of each node of the decision tree can be determined by using cluster analysis of three vital indices,and calculating the average and standard differences of each type and each index.The 4cm-depth soil temperature on the Qinghai-Tibetan Plateau observed by Soil Moisture and Temperature Measuring System of GEWEX-Coordinated Enhanced Observing Period,were used to validate the classification results.The total accuracy can reach about 87%.A majority of misclassification occurred near the freezing point of soil,about 40% and 73% of the misclassified cases appeared when the surface soil temperature is between-0.5—0.5℃ and-2.0—2.0℃,respectively.Furthermore,the misclassification mainly occurred during the transition period between warm and cold seasons,namely April-May and September-October.Based on this decision tree,a map of the number of frozen days during Oct.2002 to Sep.2003 in China was produced by composing 5 days classification results due to the swath coverage of SSM/I.The accuracy assessment for pixels with more than 15 frozen days(less than 15 meaning the short time frozen soil)was carried out with the regions of permafrost and seasonally frozen ground in map of geocryological regionalization and classification in China as reference data(Zhou et al.,2000),and the total classification accuracy was 91.66%,while the Kappa coefficient was 80.5%.The boundary between frozen and thawed soil was well consistent with the southern limit of seasonally frozen ground.A long time series surface frozen/thawed dataset can be produced using this decision tree,which may provide indicating information for regional climate change studies,regional and global scale carbon cycle models,hydrologic model and land surface model so on.
Object detection tasks for sonar image confront two major challenges, scarcity of dataset and perturbation of noise, which cause overfitting to models. The state-of-the-art object detection designed for optical images cannot address the issues because of the inherent differentiation between the optical image and sonar image. To tackle this problem, in this paper, we propose an adversarial training method to generalize the detector by introducing perturbation with specific noise property of sonar images during training stage. We design a sideway network which we name Noise Adversarial Network (NAN). The NAN is embedded into the state-of-the-art detector to generate adversarial examples which serve as assistant decision-making items to predict both class and bounding box, aiming to improve the generalization and noise robustness of the detector. To provide prior knowledge of noise perturbation to NAN, we also design a Noise Block (NB) for introducing noise in the upstream layers, which further improves noise robustness. Following the Faster R-CNN framework, the results of our experiments indicate a 8.9% mAP boost on our sonar image dataset. The detector equipped with NAN and NB also outperforms the baseline on noised test sets. Furthermore, it gains a 2.4% mAP boost on the optical image dataset PASCAL VOC 2007.
Conventional matched-field processing (MFP) for acoustic source localization is sensitive to environmental mismatches because it is based on the wave propagation model and environmental information that is uncertain in reality. In this paper, a mode-predictable sparse Bayesian learning (MPR-SBL) method is proposed to increase robustness in the presence of environmental uncertainty. The estimator maximizes the marginalized probability density function (PDF) of the received data at the sensors, utilizing the Bayesian rule and two hyperparameters (the source powers and the noise variance). The replica vectors in the estimator are reconstructed with the predictable modes from the decomposition of the pressure in the representation of the acoustic normal mode. The performance of this approach is evaluated and compared with the Bartlett processor and original sparse Bayesian learning, both in simulation and using the SWellEx-96 Event S5 dataset. The results illustrate that the proposed MPR-SBL method exhibits better performance in the two-source scenario, especially for the weaker source.
'/%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)