CAAI Transactions on Intelligence Technology最新文献

Underwater images are often with biased colours and reduced contrast because of the absorption and scattering effects when light propagates in water. Such images with degradation cannot meet the needs of underwater operations. The main problem in classic underwater image restoration or enhancement methods is that they consume long calculation time, and often, the colour or contrast of the result images is still unsatisfied. Instead of using the complicated physical model of underwater imaging degradation, we propose a new method to deal with underwater images by imitating the colour constancy mechanism of human vision using double-opponency. Firstly, the original image is converted to the LMS space. Then the signals are linearly combined, and Gaussian convolutions are performed to imitate the function of receptive fields (RFs). Next, two RFs with different sizes work together to constitute the double-opponency response. Finally, the underwater light is estimated to correct the colours in the image. Further contrast stretching on the luminance is optional. Experiments show that the proposed method can obtain clarified underwater images with higher quality than before, and it spends significantly less time cost compared to other previously published typical methods.

Initiatives to minimise battery use, address sustainability, and reduce regular maintenance have driven the challenge to use alternative power sources to supply energy to devices deployed in Internet of Things (IoT) networks. As a key pillar of fifth generation (5G) and beyond 5G networks,IoT is estimated to reach 42 billion devices by the year 2025. Thermoelectric generators (TEGs) are solid state energy harvesters which reliably and renewably convert thermal energy into electrical energy. These devices are able to recover lost thermal energy, produce energy in extreme environments, generate electric power in remote areas, and power micro-sensors. Applying the state of the art, the authorspresent a comprehensive review of machine learning (ML) approaches applied in combination with TEG-powered IoT devices to manage and predict available energy. The application areas of TEG-driven IoT devices that exploit as a heat source the temperature differences found in the environment, biological structures, machines, and other technologies are summarised. Based on detailed research of the state of the art in TEG-powered devices, the authors investigated the research challenges, applied algorithms and application areas of this technology. The aims of the research were to devise new energy prediction and energy management systems based on ML methods, create supervised algorithms which better estimate incoming energy, and develop unsupervised and semi-supervised approaches which provide adaptive and dynamic operation. The review results indicate that TEGs are a suitable energy harvesting technology for low-power applications through their scalability, usability in ubiquitous temperature difference scenarios, and long operating lifetime. However, TEGs also have low energy efficiency (around 10%) and require a relatively constant heat source.

Urgent care clinics and emergency departments around the world periodically suffer from extended wait times beyond patient expectations due to surges in patient flows. The delays arising from inadequate staffing levels during these periods have been linked with adverse clinical outcomes. Previous research into forecasting patient flows has mostly used statistical techniques. These studies have also predominately focussed on short-term forecasts, which have limited practicality for the resourcing of medical personnel. This study joins an emerging body of work which seeks to explore the potential of machine learning algorithms to generate accurate forecasts of patient presentations. Our research uses datasets covering 10 years from two large urgent care clinics to develop long-term patient flow forecasts up to one quarter ahead using a range of state-of-the-art algorithms. A distinctive feature of this study is the use of eXplainable Artificial Intelligence (XAI) tools like Shapely and LIME that enable an in-depth analysis of the behaviour of the models, which would otherwise be uninterpretable. These analysis tools enabled us to explore the ability of the models to adapt to the volatility in patient demand during the COVID-19 pandemic lockdowns and to identify the most impactful variables, resulting in valuable insights into their performance. The results showed that a novel combination of advanced univariate models like Prophet as well as gradient boosting, into an ensemble, delivered the most accurate and consistent solutions on average. This approach generated improvements in the range of 16%–30% over the existing in-house methods for estimating the daily patient flows 90 days ahead.

Some types of software systems, like event-based and non-deterministic ones, are usually specified as rules so that we can analyse the system behaviour by drawing inferences from firing the rules. However, when the fuzzy rules are used for the specification of non-deterministic behaviour and they contain a large number of variables, they constitute a complex form that is difficult to understand and infer. A solution is to visualise the system specification with the capability of automatic rule inference. In this study, by representing a high-level system specification, the authors visualise rule representation and firing using fuzzy coloured Petri-nets. Already, several fuzzy Petri-nets-based methods have been presented, but they either do not support a large number of rules and variables or do not consider significant cases like (a) the weight of the premise's propositions in the occurrence of the rule conclusion, (b) the weight of conclusion's proposition, (c) threshold values for premise and conclusion's propositions of the rule, and (d) the certainty factor (CF) for the rule or the conclusion's proposition. By considering cases (a)–(d), a wider variety of fuzzy rules are supported. The authors applied their model to the analysis of attacks against a part of a real secure water treatment system. In another real experiment, the authors applied the model to the two scenarios from their previous work and analysed the results.

Disturbance observer-based control method has achieved good results in the car-following scenario of intelligent and connected vehicle (ICV). However, the gain of conventional extended disturbance observer (EDO)-based control method is usually set manually rather than adjusted adaptively according to real time traffic conditions, thus declining the car-following performance. To solve this problem, a car-following strategy of ICV using EDO adjusted by reinforcement learning is proposed. Different from the conventional method, the gain of proposed strategy can be adjusted by reinforcement learning to improve its estimation accuracy. Since the “equivalent disturbance” can be compensated by EDO to a great extent, the disturbance rejection ability of the car-following method will be improved significantly. Both Lyapunov approach and numerical simulations are carried out to verify the effectiveness of the proposed method.

Sparse representation is an effective data classification algorithm that depends on the known training samples to categorise the test sample. It has been widely used in various image classification tasks. Sparseness in sparse representation means that only a few of instances selected from all training samples can effectively convey the essential class-specific information of the test sample, which is very important for classification. For deformable images such as human faces, pixels at the same location of different images of the same subject usually have different intensities. Therefore, extracting features and correctly classifying such deformable objects is very hard. Moreover, the lighting, attitude and occlusion cause more difficulty. Considering the problems and challenges listed above, a novel image representation and classification algorithm is proposed. First, the authors’ algorithm generates virtual samples by a non-linear variation method. This method can effectively extract the low-frequency information of space-domain features of the original image, which is very useful for representing deformable objects. The combination of the original and virtual samples is more beneficial to improve the classification performance and robustness of the algorithm. Thereby, the authors’ algorithm calculates the expression coefficients of the original and virtual samples separately using the sparse representation principle and obtains the final score by a designed efficient score fusion scheme. The weighting coefficients in the score fusion scheme are set entirely automatically. Finally, the algorithm classifies the samples based on the final scores. The experimental results show that our method performs better classification than conventional sparse representation algorithms.

Appropriately characterising the mixed space–time relations of the contagion process caused by hybrid space and time factors remains the primary challenge in COVID-19 forecasting. However, in previous deep learning models for epidemic forecasting, spatial and temporal variations are captured separately. A unified model is developed to cover all spatio–temporal relations. However, this measure is insufficient for modelling the complex spatio–temporal relations of infectious disease transmission. A dynamic adaptive spatio–temporal graph network (DASTGN) is proposed based on attention mechanisms to improve prediction accuracy. In DASTGN, complex spatio–temporal relations are depicted by adaptively fusing the mixed space–time effects and dynamic space–time dependency structure. This dual-scale model considers the time-specific, space-specific, and direct effects of the propagation process at the fine-grained level. Furthermore, the model characterises impacts from various space–time neighbour blocks under time-varying interventions at the coarse-grained level. The performance comparisons on the three COVID-19 datasets reveal that DASTGN achieves state-of-the-art results with a maximum improvement of 17.092% in the root mean-square error and 11.563% in the mean absolute error. Experimental results indicate that the mechanisms of designing DASTGN can effectively detect some spreading characteristics of COVID-19. The spatio–temporal weight matrices learned in each proposed module reveal diffusion patterns in various scenarios. In conclusion, DASTGN has successfully captured the dynamic spatio–temporal variations of COVID-19, and considering multiple dynamic space–time relationships is essential in epidemic forecasting.