Unsupervised-learning-based calibration method in microscopic fringe projection profilometry.

Microscopic fringe projection profilometry (MFPP) technology is widely used in 3D measurement. The measurement precision performed by the MFPP system is closely related to the calibration accuracy. However, owing to the shallow depth of field, calibration in MFPP is frequently influenced by low-quality target images, which would generate inaccurate features and calibration parameter estimates. To alleviate the problem, this paper proposes an unsupervised-learning-based calibration robust to defocus and noise, which could effectively enhance the image quality and increase calibration accuracy. In this method, first, an unsupervised image deblurring network (UIDNet) is developed to recover a sharp target image from the deteriorated one. Free from capturing strictly paired images by a specific vision system or generating the dataset by simulation, the unsupervised deep learning framework can learn more accurate features from the multi-quality target dataset of convenient image acquisition. Second, multi-perceptual loss and Fourier frequency loss are introduced into the UIDNet to improve the training performance. Third, a robust calibration compensation strategy based on 2D discrete Fourier transform is also developed to evaluate the image quality and improve the detection accuracy of the reference feature centers for fine calibration. The relevant experiments demonstrate that the proposed calibration method can achieve superior performance in terms of calibration accuracy and measurement precision.

Small object intelligent detection method based on adaptive recursive feature pyramid.

As we all know, YOLOv4 can achieve excellent detection performance in object detection and has been effectively applied in many fields. However, the inconsistency of scale features affects the prediction accuracy of the path aggregation network (PANet) in YOLOv4 for small objects, resulting in low detection accuracy. This paper presents YOLOv4, which uses an adaptive recursive path aggregation network (AR-PANet) to improve the detection accuracy of small objects. First, the output characteristics of the PANet are fed back into the backbone network by using a recursive structure to enrich the characteristic information of the object. Second, an adaptive approach is developed to eliminate conflicting information in multi-scale feature space, thereby enhancing scale invariance and promoting feature extraction accuracy for small objects. Finally, the CBAM is used to map the multi-scale features obtained from the AR-PANet to independent channels and spatial dimensions to achieve feature refinement, thus improving the detection accuracy of small objects. Experimental results show that our proposed method can effectively improve the accuracy of small object detection in multiple datasets, addressing this challenging problem with impressive results. Thus, our proposed approach has great potential and valuable applications in the fields of remote sensing and intelligent transportation.

Attention-Driven Memory Network for Online Visual Tracking.

A memory mechanism has attracted growing popularity in tracking tasks due to the ability of learning long-term-dependent information. However, it is very challenging for existing memory modules to provide the intrinsic attribute information of the target to the tracker in complex scenes. In this article, by considering the biological visual memory mechanisms, we propose the novel online tracking method via an attention-driven memory network, which can mine discriminative memory information and enhance the robustness and reliability of the tracker. First, to reinforce effectiveness of memory content, we design a novel attention-driven memory network. In the network, the long memory module gains property-level memory information by focusing on the state of the target at both the channel and spatial levels. Meanwhile, in reciprocity, we add a short-term memory module to maintain good adaptability when confronting drastic deformation of the target. The attention-driven memory network can adaptively adjust the contribution of short-term and long-term memories to tracking results under the weighted gradient harmonized loss. On this basis, to avoid model performance degradation, an online memory updater (MU) is further proposed. It is designed to mining for target information in tracking results through the Mixer layer and the online head network together. By evaluating the confidence of the tracking results, the memory updater can accurately judge the time of updating the model, which guarantees the effectiveness of online memory updates. Finally, the proposed method performs favorably and has been extensively validated on several benchmark datasets, including object tracking benchmark-50/100 (OTB-50/100), temple color-128 (TC-128), unmanned aerial vehicles-123 (UAV-123), generic object tracking -10k (GOT-10k), visual object tracking-2016 (VOT-2016), and VOT-2018 against several advanced methods.

Adaptive parameter estimation for the expanded sandwich model.

An expanded-sandwich system is a nonlinear extended block-oriented system in which memoryless elements in conventional block-oriented systems are displaced by memory submodels. Expanded-sandwich system identification has received extensive attention in recent years due to the powerful ability of these systems to describe actual industrial systems. This study proposes a novel recursive identification algorithm for an expanded-sandwich system, in which an estimator is developed on the basis of parameter identification error data rather than the traditional prediction error output information. In this scheme, a filter is introduced to extract the available system information based on miserly structure layout, and some intermediate variables are designed using filtered vectors. According to the developed intermediate variables, the parameter identification error data can be obtained. Thereafter, an adaptive estimator is established by integrating the identification error data compared with the classic adaptive estimator based on the prediction error output information. Thus, the design framework introduced in this research provides a new perspective for the design of identification algorithms. Under a general continuous excitation condition, the parameter estimation values can converge to the true values. Finally, experimental results and illustrative examples indicate the availability and usefulness of the proposed method.

A novel recursive learning estimation algorithm of Wiener systems with quantized observations.

In this paper, a novel recursive learning identification approach is proposed to estimate the parameters of the Wiener systems with quantized output. By using a filter with adaptive performance, the data preprocessing is achieved based on the system data. To derive the error information of parameter estimation, some filtered and intermediate variables are developed. Based on the estimation error and initial parameter data, a novel loss function is established, in which the estimation precision can be raised by force of the estimation error data and the convergence rate can be improved based on the initial parameter data. By minimizing the loss function, a novel recursive learning estimator is derived where the performance of the modified gain is improved due to the utilization of the observed data. Under the continuous excitation condition, the convergence analysis shows that the estimation error can converge to zero. Finally, illustrative examples and a real-life experiment are performed to validate the obtained results and efficiency of the proposed algorithm.

Robust adaptive identification for sandwich systems with unknown time-delay.

A discrete time, robust adaptive estimator is developed to identify the time-delay and sandwich systems parameters. To obtain explicitly the expression of delay parameter, the observation and augmented data are reconstructed. By using the filter operator and some auxiliary vectors, the parameter identification error vector is derived. Based on the parameter identification error term and initial estimation error term, a novel criterion function is invented. In comparison to the common criterion function, a classy estimation property is provided based on the criterion function in that paper because the identification error term can lift estimation accuracy, and the initial estimation error term speeds up the convergence. Under the persistent excitation condition, convergence of the developed estimator is analyzed. The availability and superiority of proposed identification scheme are verified by both numerical simulation and a turntable servo system.

Hardness Recognition of Robotic Forearm Based on Semi-supervised Generative Adversarial Networks.

The hardness recognition is of great significance to tactile sensing and robotic control. The hardness recognition methods based on deep learning have demonstrated a good performance, however, a huge amount of manually labeled samples which require lots of time and labor costs are necessary for the training of deep neural networks. In order to alleviate this problem, a semi-supervised generative adversarial network (GAN) which requires less manually labeled samples is proposed in this paper. First of all, a large number of unlabeled samples are made use of through the unsupervised training of GAN, which is used to provide a good initial state to the following model. Afterwards, the manually labeled samples corresponding to each hardness level are individually used to train the GAN, of which the architecture and initial parameter values are inherited from the unsupervised GAN, and augmented by the generator of trained GAN. Finally, the hardness recognition network (HRN), of which the main architecture and initial parameter values are inherited from the discriminator of unsupervised GAN, is pretrained by a large number of augmented labeled samples and fine-tuned by manually labeled samples. The hardness recognition result can be obtained online by importing the tactile data captured by the robotic forearm into the trained HRN. The experimental results demonstrate that the proposed method can significantly save the manual labeling work while providing an excellent recognition precision for hardness recognition.

SIFT flow for large-displacement object tracking.

Traditional tracking methods place an emphasis on how to cope with the variations in target appearance effectively. However, when the motion displacement of the target between image frames becomes larger, these methods may be unstable. This paper presents a novel (to our knowledge) visual object tracking method. In this method, we first introduce scale-invariant feature transform (SIFT) flow into the tracking problem and develop a real-time motion prediction method to capture large displacement between consecutive image frames. Then we use belief propagation (BP) to convert the problem of finding maximum a posteriori probability (MAP) to globally minimizing an energy function to get the best matching pairs of points for producing good candidate regions of the target. And last, the refined point trajectories are obtained according to the bidirectional flow field consistency estimation and covariance region descriptor matching, which can update model states efficiently so as to achieve enhanced robustness for visual tracking. Compared with the state-of-art tracking methods, the experimental results demonstrate that the proposed algorithm shows favorable performance when the object undergoes large motion displacement between image frames.

Anomaly detection based on local nearest neighbor distance descriptor in crowded scenes.

We propose a novel local nearest neighbor distance (LNND) descriptor for anomaly detection in crowded scenes. Comparing with the commonly used low-level feature descriptors in previous works, LNND descriptor has two major advantages. First, LNND descriptor efficiently incorporates spatial and temporal contextual information around the video event that is important for detecting anomalous interaction among multiple events, while most existing feature descriptors only contain the information of single event. Second, LNND descriptor is a compact representation and its dimensionality is typically much lower than the low-level feature descriptor. Therefore, not only the computation time and storage requirement can be accordingly saved by using LNND descriptor for the anomaly detection method with offline training fashion, but also the negative aspects caused by using high-dimensional feature descriptor can be avoided. We validate the effectiveness of LNND descriptor by conducting extensive experiments on different benchmark datasets. Experimental results show the promising performance of LNND-based method against the state-of-the-art methods. It is worthwhile to notice that the LNND-based approach requires less intermediate processing steps without any subsequent processing such as smoothing but achieves comparable event better performance.