Robust Measurement-Driven Cardinality Balance Multi-Target Multi-Bernoulli Filter.

The multi-target tracking filter under the Bayesian framework has strict requirements on the prior information of the target, such as detection probability density, clutter density, and target initial position information. This paper proposes a novel robust measurement-driven cardinality balance multi-target multi-Bernoulli filter (RMD-CBMeMBer) for solving the multiple targets tracking problem when the detection probability density is unknown, the background clutter density is unknown, and the target's prior position information is lacking. In RMD-CBMeMBer filtering, the target state is first extended, so that the extended target state includes detection probability, kernel state, and indicators of target and clutter. Secondly, the detection probability is modeled as a Beta distribution, and the clutter is modeled as a clutter generator that is independent of each other and obeys the Poisson distribution. Then, the detection probability, kernel state, and clutter density are jointly estimated through filtering. In addition, the correlation function (CF) is proposed for creating new Bernoulli component (BC) by using the measurement information at the previous moment. Numerical experiments have verified that the RMD-CBMeMBer filter can solve the multi-target tracking problem under the condition of unknown target detection probability, unknown background clutter density and inadequate prior position information of the target. It can effectively estimate the target detection probability and the clutter density.

Unambiguous Forward-Looking SAR Imaging on HSV-R Using Frequency Diverse Array.

It is a challenge to realize wide swath imaging due to the conflict between Doppler ambiguity and range ambiguity for hypersonic vehicle (HSV) radar. In addition, there are many conditions requiring the forward-looking imaging. In a forward-looking synthetic aperture radar (SAR) system, left-right ambiguity arises, since two symmetrical targets have the same Doppler frequency magnitude. After selecting an appropriate pulse repetition frequency (PRF) to avoid Doppler ambiguity, we only need to solve the range ambiguity and left-right ambiguity. To handle these issues, this paper proposes an approach to resolve the range ambiguity and left-right ambiguity using the frequency diverse array (FDA). With the range-angle-dependent property of the transmit steering vector, FDA can distinguish the range ambiguous echoes in the spatial frequency domain. By performing transmit beamforming after range compensation, the echo from the desired range region can be extracted from ambiguous echoes. Then, the back projection (BP) algorithm is used to achieve imaging. Next, the echoes of all channels are processed by two receive beamformers, which are designed for the right and left sides, respectively. With the aforementioned procedures, an unambiguous image can be obtained. Simulation results have verified the effectiveness of the proposed approach.

Direction-of-Arrival Estimation for Circulating Space-Time Coding Arrays: From Beamspace MUSIC to Spatial Smoothing in the Transform Domain.

As a special type of coherent collocated Multiple-Input Multiple-Output (MIMO) radar, a circulating space-time coding array (CSTCA) transmits an identical waveform with a tiny time shift. It provides a simple way to achieve a full angular coverage with a stable gain and a low sidelobe level (SLL) in the range domain. In this paper, we address the problem of direction-of-arrival (DOA) estimation in CSTCA. Firstly, we design a novel two-dimensional space-time matched filter on receiver. It jointly performs equivalent transmit beamforming in the angle domain and waveform matching in the fast time domain. Multi-beams can be formed to acquire controllable transmit freedoms. Then, we propose a beamspace multiple signal classification (MUSIC) algorithm applicable in case of small training samples. Next, since targets at the same range cell show characteristics of coherence, we devise a transformation matrix to restore the rotational invariance property (RIP) of the receive array. Afterwards, we perform spatial smoothing in element domain based on the transformation. In addition, the closed-form expression of Cramer-Rao lower bound (CRLB) for angle estimation is derived. Theoretical performance analysis and numerical simulations are presented to demonstrate the effectiveness of proposed approaches.

Semisupervised Change Detection Based on Bihierarchical Feature Aggregation and Extraction Network.

With the rapid development of remote sensing (RS) technology, high-resolution RS image change detection (CD) has been widely used in many applications. Pixel-based CD techniques are maneuverable and widely used, but vulnerable to noise interference. Object-based CD techniques can effectively utilize the abundant spectrum, texture, shape, and spatial information but easy-to-ignore details of RS images. How to combine the advantages of pixel-based methods and object-based methods remains a challenging problem. Besides, although supervised methods have the capability to learn from data, the true labels representing changed information of RS images are often hard to obtain. To address these issues, this article proposes a novel semisupervised CD framework for high-resolution RS images, which employs small amounts of true labeled data and a lot of unlabeled data to train the CD network. A bihierarchical feature aggregation and extraction network (BFAEN) is designed to achieve the pixelwise together with objectwise feature concatenation feature representation for the comprehensive utilization of the two-level features. In order to alleviate the coarseness and insufficiency of labeled samples, a confident learning algorithm is used to eliminate noisy labels and a novel loss function is designed for training the model using true-and pseudo-labels in a semisupervised fashion. Experimental results on real datasets demonstrate the effectiveness and superiority of the proposed method.

Morphological characteristics, anatomical structure, and dynamic change of ascorbic acid under different storage conditions of celery.

Ascorbic acid (AsA) is a crucial antioxidant in vegetables. Celery (Apium graveolens L.) is a vegetable of Apiaceae and is rich in AsA. Till now, the effects of different storage conditions on celery morphological characteristics, anatomical features, and antioxidant accumulation are unclear. Here, the celery cvs. 'Sijixiaoxiangqin' and 'Liuhehuangxinqin' were selected as experimental materials, and the two celery plants grown for 65 days were harvested from soils and stored in light at room temperature (25 °C), darkness at low temperature (4 °C), and darkness at room temperature (25 °C) for 0, 6, 24, 30, 48, and 54 h, respectively. The results showed that celery in darkness had better water retention capacity than celery in light. Morphological changes in celery mesophyll, leaf veins, and petioles were the least in darkness at low temperature (4 °C). The weight loss rate and wilting degree in darkness at low temperature (4 °C) were the lowest, and the AsA content remained at a high level. The expression patterns of GDP-D-mannose pyrophosphorylase (AgGMP) and L-galactose dehydrogenase (AgGalDH) were similar to the change of AsA content. The results indicated that low temperature and dark was the optimized storage condition for 'Sijixiaoxiangqin' and 'Liuhehuangxinqin' celery. AgGMP and AgGalDH genes may play an important role in the accumulation of AsA in celery. This paper will provide potential references for prolonging the shelf life of celery and other horticultural crops.

Symmetric All Convolutional Neural-Network-Based Unsupervised Feature Extraction for Hyperspectral Images Classification.

Recently, deep-learning-based feature extraction (FE) methods have shown great potential in hyperspectral image (HSI) processing. Unfortunately, it also brings a challenge that the training of the deep learning networks always requires large amounts of labeled samples, which is hardly available for HSI data. To address this issue, in this article, a novel unsupervised deep-learning-based FE method is proposed, which is trained in an end-to-end style. The proposed framework consists of an encoder subnetwork and a decoder subnetwork. The structure of the two subnetworks is symmetric for obtaining better downsampling and upsampling representation. Considering both spectral and spatial information, 3-D all convolution nets and deconvolution nets are used to structure the encoder subnetwork and decoder subnetwork, respectively. However, 3-D convolution and deconvolution kernels bring more parameters, which can deteriorate the quality of the obtained features. To alleviate this problem, a novel cost function with a sparse regular term is designed to obtain more robust feature representation. Experimental results on publicly available datasets indicate that the proposed method can obtain robust and effective features for subsequent classification tasks.