To overcome the IMM algorithm is easy divergence and low tracking accuracy when dealing with complex maneuvering situations, this paper propose an improved interactive multiple model strong tracking square room cubature Kalman filter (IIMM-STSRCKF) algorithm under the idea of real-time dynamic adjustment of gain matrix and transition probability matrix. The algorithm has been improved in two aspects, on the one hand, the algorithm uses the idea of a strong tracking filter to deduce a new method for time-varying fading factor, and introduce it into square root of the state error covariance matrix of the SRCKF, which improves the tracking accuracy for strong maneuver. On the other hand, the probability difference between two consecutive time points in the IMM submodel is used to adjust the Markov probability transfer matrix to adaptively improve the switching speed of the submodel and the rationality of the allocation. By comparing with IMM-CKF algorithm by maneuvering target tracking case, and results show that the IIMM-STSRCKF algorithm has better tracking performance in nonmaneuvering, weak maneuvering and strong maneuvering cases.

Deep convolutional neural networks (CNNs) have been widely used to obtain high-level representation in various computer vision tasks. However, in the field of remote sensing, there are not sufficient images to train a useful deep CNN. Instead, we tend to transfer successful pre-trained deep CNNs to remote sensing tasks. In the transferring process, generalization power of features in pre-trained deep CNNs plays the key role. In this paper, we propose two promising architectures to extract general features from pre-trained deep CNNs for remote scene classification. These two architectures suggest two directions for improvement. First, before the pre-trained deep CNNs, we design a linear PCA network (LPCANet) to synthesize spatial information of remote sensing images in each spectral channel. This design shortens the spatial “distance” of target and source datasets for pre-trained deep CNNs. Second, we introduce quaternion algebra to LPCANet, which further shortens the spectral “distance” between remote sensing images and images used to pre-train deep CNNs. With five well-known pre-trained deep CNNs, experimental results on three independent remote sensing datasets demonstrate that our proposed framework obtains state-of-the-art results without fine-tuning and feature fusing. This paper also provides baseline for transferring fresh pretrained deep CNNs to other remote sensing tasks.

In order to improve filtering precision and restrain divergence caused by sensor faults or model mismatches for target tracking, a new adaptive unscented Kalman filter (N-AUKF) algorithm is proposed. First of all, the unscented Kalman filter (UKF) problem to be solved for systems involving model mismatches is described, after that, the necessary and sufficient condition with third order accuracy of the standard UKF is given and proven by using the matrix theory. In the filtering process of N-AUKF, an adaptive matrix gene is introduced to the standard UKF to adjust the covariance matrixes of the state vector and innovation vector in real time, which makes full use of normal innovations. Then, a covariance matching criterion is designed to judge the filtering divergence. On this basis, an adaptive weighted coefficient is applied to restrain the divergence. Compared with the standard UKF and existing adaptive UKF, the proposed UKF algorithm improves the filtering accuracy, rapidity and numerical stability remarkably, moreover, it has a good adaptive capability to deal with sensor faults or model mismatches. The performance and effectiveness of the proposed UKF is verified in a target tracking mission.

Event cameras are bio-inspired vision sensors measuring brightness changes (referred to as an ‘event’) for each pixel independently, instead of capturing brightness images at a fixed rate using conventional cameras. Asynchronous event data mixed with noise information is challenging for event-based vision tasks. In this paper, we propose a broad learning network for object detection using the event data. The broad learning network consists of two distinct layers, a feature-node layer and an enhancement-node layer. Different to convolutional neural networks, the broad learning network can be extended by adding nodes into layers during training. We design a gradient descent algorithm to train network parameters, which creates an event-based broad learning network in an end-to-end manner. Our model outperforms state-of-the-art models, specifically, because of the small scale and increased speed displayed by our model during training. This demonstrates the superiority of event cameras towards online training and inference.