Cyclone detection is a classical topic and researchers have developed various methods of cyclone detection based on sea-level pressure, cloud image, wind field, etc. In this article, a deep-learning algorithm is incorporated with modern remote-sensing technology and forms a global-scale cyclone/anticyclone detection model. Instead of using optical images, wind field data obtained from Mean Wind Field-Advanced Scatterometer (MWF-ASCAT) is utilized as the dataset for model training and testing. The wind field vectors are reconstructed and fed to the deep-learning model, which is built based on a faster-region with convolutional neural network (faster-RCNN). The model consists of three modules: a series of convolutional and pooling layers as the feature extractor, a region proposal network that searches for the potential areas of cyclone/anticyclone within the dataset, and the classifier that classifies the proposed region as cyclone or anticyclone through a fully-connected neural network. Compared with existing methods of cyclone detection, the test results indicate that this model based on deep learning is able to reduce the number of false alarms, and at the same time, maintain high accuracy in cyclone detection. An application of this method is presented in the article. By processing temporally continuous data of wind field, the model is able to track the path of Hurricane Irma in September, 2017. The advantages and limitations of the model are also discussed in the article.

Mixed Poisson–Gaussian noise exists in the star images and is difficult to be effectively suppressed via maximum likelihood estimation (MLE) method due to its complicated likelihood function. In this article, the MLE method is incorporated with a state-of-the-art machine learning algorithm in order to achieve accurate restoration results. By applying the mixed Poisson–Gaussian likelihood function as the reward function of a reinforcement learning algorithm, an agent is able to form the restored image that achieves the maximum value of the complex likelihood function through the Markov Decision Process (MDP). In order to provide the appropriate parameter settings of the denoising model, the key hyperparameters of the model and their influences on denoising results are tested through simulated experiments. The model is then compared with two existing star image denoising methods so as to verify its performance. The experiment results indicate that this algorithm based on reinforcement learning is able to suppress the mixed Poisson–Gaussian noise in the star image more accurately than the traditional MLE method, as well as the method based on the deep convolutional neural network (DCNN).

The SO₂ discharged by ships causes serious pollution to the atmosphere. Therefore, the International Maritime Organization has set strict requirements on the sulfur content of marine fuel. For the first time, this study investigates the optimal detection wavelength based on the imaging technology to realize an accurate monitoring of the SO₂ concentration in ship exhaust. First, a simulation analysis model of the optimal imaging detection (SAMID) wavelength of the SO₂ concentration in ship exhaust is proposed and analyzed in this study. Next, a bench experiment is designed. The values and the range of the gas concentration values required for the simulation are obtained. Finally, based on the principle of minimum error, the optimal detection wavelengths of the single- and the dual-wavelength imaging detection technologies are determined as 287 nm and 297 nm and 298 nm, respectively. During the SO₂ concentration retrieval, the minimum values of the root mean squared error, the mean absolute error, and the mean absolute percentage error of the single- and the dual-wavelengths are 563.14 molecules/cm³, 445.11 molecules/cm³, and 347.22% and 0.62 molecules/cm³, 0.49 molecules/cm³, and 0.85%, respectively. The simulation analysis results provide a theoretical basis for the future hardware development of an optical remote sensing system based on the imaging detection technology.

The emission of SO₂ from ships is an important source of atmospheric pollution. Therefore, the International Maritime Organization (IMO) has established strict requirements for the sulfur content of marine fuel oil. In this paper, a new optical noncontact detection technique for ship exhaust emissions analysis is studied. Firstly, the single-band simulation analysis model of the imaging detection technology for SO₂ concentration in ship exhaust gas and the deep neural network model for the prediction of sulfur content were established. A bench test was designed to monitor the tail gas concentration simultaneously using online and imaging detection methods, so as to obtain the concentration data in the flue and the ultraviolet image data. The results showed that 300 nm had a higher inversion accuracy than the other two bands. Finally, a deep neural network model was trained with the SO₂ concentration data from the inversion and the engine power, and the predictive model of sulfur content in marine fuel oil was thereby obtained. When the deep learning model was used to predict sulfur content, the prediction accuracy at 300, 310, and 330 nm was 73%, 94%, and 71%, respectively.