Search

Recent deep learning (DL) detectors adopted by radar or sonar (RS) are normally trained with transfer learning, where the typical workflow is to pretrain a convolutional neural network (CNN) on external large-scale classification datasets (e.g., ImageNet) as the backbone and then finetune the entire detector on detection datasets. Though transfer learning could effectively avoid overfitting, transferred models are usually redundant and might not generalize well on RS datasets. To achieve high generalization and to eliminate the dependence on transfer learning, a self-trained target detection method is established by including Automatic Deep Learning (AutoDL) to design optimal detectors. This self-trained target detection consists of three stages. First, a derived classification dataset (DCD) consisting of image blocks of targets and backgrounds is derived from detection datasets. Then, a memory-efficient Differentiable Architecture Search algorithm with flexible search space and large inputs (FL-DARTS), which is characterized by its predefined multistride convolutions, poolings, and unique super-structure, is proposed to automatically design and self-train optimal CNNs on DCDs. Finally, self-trained AutoDL detectors are implemented with the automatic backbone designed by FL-DARTS. We evaluated three self-trained AutoDL detectors on the public SAR ship detection dataset (SSDD) and the self-made sonar common target detection dataset (SCTD). The experiments show that while the number of parameters of automatic backbones designed for SSDD and SCTD are only 11.8% and 15.2% of that of ResNet50, self-trained AutoDL detectors implemented with automatic backbones significantly outperform their transfer learning detectors and achieve state-of-the-art detection precisions and high detection speeds. Data, codes are publicly available.

The low efficiency of complex image registration has become a key problem for the signal processing of interferometric synthetic aperture sonar (InSAS). In this paper, a collaborative complex image registration method based on CPU/GPU is proposed, which provides a new method of realtime complex image registration for InSAS. The processing steps of coarse registration, salve image interpolation and wrapped phase extraction are mapped to graphics processing unit (GPU) and executed respectively. The refine registration and polynomial coefficient fitting are still executed on CPU. The complex image registration test performed on real InSAS data confirms the validity and efficiency of the proposed method.

Phase unwrapping is one of the key problems in reconstructing the digital elevation model of a scene from its interferometric synthetic aperture radar (InSAR) or interferometric synthetic aperture sonar (InSAS) data. In this letter, we propose a quality-guided and local minimum discontinuity based phase unwrapping algorithm, which enhances the precision of the unwrapped result by local minimum discontinuity optimization in low quality areas, and still keeps a high efficiency. The new algorithm can be divided into two steps. Firstly, the quality-guided phase unwrapping algorithm is performed on the original wrapped phase image, which helps to remove all the discontinuities caused by the wrapping operation quickly, but tends to spread the unwrapped errors in low quality areas. Secondly, the initial unwrapped result is divided into high and low quality areas according to its quality map, and the minimum discontinuity optimization process is performed in the low quality areas of the initial unwrapped result, which helps to remove the remaining improving loops defined as a loop with more positive jumps than negative ones. This prevents the unwrapped errors spreading from low quality areas to high quality areas and accelerates the optimization process by restricting the optimization place in low quality areas. Tests performed on InSAR and real InSAS data confirm the accuracy and efficiency of the proposed algorithm.

Complex image registration is a crucial step in interferometric synthetic aperture sonar (InSAS) signal processing. The registration quality directly influences the quality of interferogram and subsequent reconstruction of digital elevation model (DEM). After the control points are determined, a geometric transformation procedure is needed in image registration. The traditional geometric transformation method is polynomial surface fitting. The polynomial surface fitting method is most suitable for far range, but it does not perform very well when the offset changes rapidly in near range distance. In this paper, a new InSAS complex image registration method based on inverse proportional surface fitting is presented to improve the accuracy and efficiency of InSAS complex image registration. Firstly, the relationship between offset and range is analyzed using the geometric model of imaging. The offset is an inverse proportional function of range, thus the inverse proportional surface fitting model is established. Then, the parameters of the inverse proportional surface fitting model are solved with control point pairs determined. Finally, the performance of the proposed model is analyzed by comparing it with the traditional second-order polynomial surface fitting model. The evaluation has been carried out using the root mean squared error (RMSE), calculation time and correlation coefficient as the criteria. A pair of images of the Qiandao lake, Zhejiang province in eastern China, are used in the experiments to validate the proposed method. The testing results indicate that the new method has the advantage of high accuracy and less calculation cost than traditional polynomial surface fitting method.

Although organizational research posits various relationships between a firm's slack resources and performance, findings to date have been ambiguous. The purpose of this study is to attempt to reconcile previous views of the relationship between slack resources and performance through employing empirical analysis based on survey data from 47 pharmaceutical and chemical firms in Henan Province of China. The results indicate that the relationship between slack resources and the performance is inverse N-shaped. Results provide evidence of a curvilinear relationship between slack resources and performance and broadly demonstrate that relationships differ based on industry circumstances and slack resources. Additionally, this study is the first to empirically identify an inverse N- shaped relationship between slack resources and performance, indicating that, in certain industry circumstances, little and much slack resources are bad for performance. Overall, results highlight the importance of additional research into intervening factors impacting the slack-performance relationship.

Purpose: This paper aims at defining the value creation mechanism and income distribution strategies of product-service integration in order to promote product-service integration of a firm. Design/methodology/approach: This paper conducts researches quantitatively on the coordination mechanism of product-service integration by using game theory, and uses the methods of Shapley value and Equal growth rate to further discuss income distribution strategies of product-service integration. Findings: Product-service integration increases the total income of a firm and the added value of the income δπ decreases as the unit price demand variation coefficient of products and services E increases, while decreases as the marginal cost of products βP increases, decreases as the marginal cost of services βS increases. Moreover, th e findings suggest that both income distribution strategies of product-service integration based on Shapley value method and Equal growth rate method can make the product department and service department of a firm winwin and realize the Pareto improvement. The choice of what kind of distribution strategy to coordinate the actions between departments depends on the department playing dominant role in the firm. Generally speaking, for a firm at the center of market, when the product department is the main contributor to firm income, the service department will choose the income distribution strategy of product-service integration based on Shapley value method; when the service department is the main contributor to firm income, the service departmentwill choose the income distribution strategy of product-service integration based on Equal growth rate method. Research limitations/implications: This paper makes some strict assumptions such as complete information, risk neutral, linear cost function and so on and the discussion is limited to the simple relationship between product department and service department. Practical implications: Product-service integration enables a firm to improve its total income. A firm should take appropriate income distribution strategies in order to promote the its product-service integration. Originality/value: This paper introduces game theory into the research of product-service integration for the first time and it has reached some valuable conclusions, which has opened up a new field of study in the product-service area.

In this paper, borrowing the idea of distributed source model in direction-of-arrival (DOA) estimation, a novel distributed pixel model is proposed for multibaseline InSAR absolute phase estimation. And the COvariance Matching Estimation Techniques-Extend Invariance Principle (COMET-EXIP) method is used to estimate the absolute interferometric phase for multibaseline InSAR. The proposed method can provide a robust unwrapped interferometric phase even in the presence of the large image coregistration. Some simulation results verify the performance of the proposed method.

In order to exploit the value of human resources slack, this paper studies the impact of human resources slack on the product innovation of information technology firms. I develop a nonlinear U-shaped relationship model between human resource slack and product innovation, and subsequently test this model using samples of 34 Chinese firms. The empirical results show that human resource slack on the low side is negatively related to the prodct innovation of information technology firms; human resource slack beyond a certain level is positively related to the prodct innovation of information technology firms.

Although organizational slack research posits various relationships between a firm's slack resources and innovation, findings to date have been ambiguous. To provide insight into this relationship, this paper studies the impact of organizational slack on technological innovation by using empirical analysis based on the data get from 283 firms in Henan Province by questionnaire. The result shows that organizational slack is positively related to technological innovation. The analysis also finds that there is the inverted U-shaped relationship between a firm's slack resources and technological innovation, but this relationship is not significant. Overall, results highlight the importance of additional research into intervening factors impacting the slack-innovation relationship.

This paper studies the relationship between financial slack and technological innovation, and brings forward the nonlinear N-shaped relationship model between financial slack and technological innovation. The data get from 283 firms in Henan Province by questionnaire support this prediction. In the firms, too little financial slack will be positively related to technological innovation; moderate financial slack will be negatively related to technological innovation; too much financial slack will be positively related to technological innovation. Therefore, managers should formulate corresponding strategies of technological innovation basing on different level of financial slack.

Two-dimensional phase unwrapping is one of the key problems in many signal processing applications. Flynn's minimum discontinuity phase unwrapping is one of the algorithm which can solve many different kinds of phase unwrapping problems successfully, but its computation efficiency is too low. To overcome this drawback, a blocked minimum discontinuity phase unwrapping algorithm is proposed. In the proposed algorithm, the wrapped phase image is tessellated into small blocks. Each block is unwrapped separately by minimum discontinuity phase unwrapping algorithm, and then the unwrapped blocks are merged together using minimum discontinuity approach again. The merging process has been divided into two steps. The discontinuities between the blocks are minimized firstly, and then the unwrapped image is taken as a whole to minimize the total discontinuities. This proposed algorithm is described in detail and tested on an IFSAR experimental wrapped phase image and the Etna wrapped image. Results show that the proposed algorithm works well and is more efficient than the Flynn's minimum discontinuity algorithm.

Phase unwrapping is the reconstruction of a function of a grid given its values mod 2π , which is a key problem in all quantitative applications of synthetic aperture radar (SAR) interferometry and other related fields. In this paper, a path following algorithm for phase unwrapping based on priority queue and quantized quality map is proposed and tested. A discrete quality map and a priority queue are designed to accelerate the unwrapping process, which can greatly reduce the unwrapping time compared with the conventional quality- guided algorithm. The test performed on real interferometric SAR data confirms the efficiency and accuracy of the proposed algorithm.

Based on the analysis frames of the multi-task principal-agent model, this paper formulates a model of principal-agent on the condition of asymmetrical information while employees are engaged in the "official duty" and the informal technological innovation based on organizational slack. Through the analysis of the model, this paper obtains the optimal incentive contracts while R&D staff, marketing personnel and production personnel are engaged in the "official duty" and the informal technological innovation based on organizational slack at the same time, which provides the theoretical basis of formulating the incentive policies of all-involvement innovation in a firm.