Object detection is a fundamental task in computer vision, relying heavily on bounding box (BB) annotations with ground truth labels to train deep learning models. This approach produces impactful results when the boundaries of objects are identifiable, but struggles in adverse weather conditions such as rain, fog, and snow, where object outlines are fuzzy. One such case is the detection of Traffic lights (TLs) in fog at night, as these weather conditions cause the light to scatter in different directions, creating a Halo effect. Therefore, creating BB for manual TL detection annotation is inaccurate. Dense fog makes it difficult for annotators to determine the state (such as red, green, or yellow) and exact location of TLs. Annotation tools are also not designed for blurred images and require annotators to manually adjust parameters like brightness and zoom. To address the challenges of manual annotation, a Boolean Map Saliency (BMS) method is employed to automatically generate annotations that highlight TLs, thereby improving object detection in scenarios where manual BBs are insufficient. Results based on automated BBs and manually annotated bounded boxes are compared using Faster-Rcnn. For both results, the SHIFT dataset is used. Our proposed approach makes it possible to generate superior-quality BBs compared to previous approaches, along with an improved TL detection algorithm, especially on foggy nights when it is difficult for the sensors to detect the TLs.

In recent engineering practice, the complexity of the systems is increasing. This complexity will keep on increasing if new components are added into the system or system configuration is changed. Numerous configurations of these components are repeated or their relationship is recognizable. On the other hand, simulation tools are commonly used to track the behavior of the modeled systems over time because of their advantages against experimental setups which are the abstraction of reality. However, the ability to integrate different components at different levels of the designed model along with the different techniques used for their inter-relationships based on users requirement is a challenge. The main contribution of this paper is to give an example of how to apply composability to handle this issue. One tangible example is to study the behavior of heating, ventilation, and air-conditioning systems in large buildings. Therefore, finding a solution to model this kind of complex system in a highly composable and scalable view is promising. Composability is a system design challenge that describes how different components can be selected and combined in different configurations and different levels to satisfy users requirement with a highly reduced development cost and time in the simulation, as its advantages. Finally, the practical steps for design and implementation of a composable demand-controlled ventilation and heating system as a useful and energy-efficient smart building's technology in MATLAB/Simulink (besides its constraints) is provided.

Active diagnosis can significantly increase the reliability of a real-time system in case of fault occurrences. Root causes are identified for observed failures and the root causes are associated with suitable recovery actions such as the migration of services to spare resources or application-specific reconfiguration. Real-time databases and diagnostic multi-query graphs (DMG) are a promising technique for root-cause analysis. However, in order to ensure safety the completion of the diagnostic queries must be performed within strict timing bounds dictated by the environment. This paper presents optimization techniques for diagnostic multi-query graphs in order to minimize the make span of a root cause analysis. The optimization is split into two steps. The first step comprises the pruning of the graph nodes without affecting the semantics of diagnostic queries. Each graph node that satisfies a certain set of constraints is deleted and its query is merged with its neighborhood nodes. The constraints for pruning and merging are based on the matching of SQL operations (select or join) and the data tables between the queries. The new graph generated after pruning is a subset of the original graph based on the merged queries from the deleted nodes. The second step is based on the optimization of the diagnostic queries in each node of the DMG, by selecting the best query execution plan. After the DMG is pruned and queries are optimized the new DMG is given as an input to a scheduler to determine the ensuing make span.

Active diagnosis in real time embedded computer systems increases the overall reliability of the system by performing error detection and fault recovery. Real time databases and diagnostic queries are a common solution to realize active diagnosis. This paper presents a technique to optimize the diagnostic queries in a fault tolerant real time embedded system. A directed graph called the DMG (Diagnostic Multi-query Graph) based on the diagnostic symptoms and features is the input to the query optimization module for the processing of each query within a short worst case execution time. The diagnostic inference process is temporally and spatially decomposed by introducing intermediate inference steps called symptoms. These symptoms and diagnostic features extracted from the DMG are stored in an embedded database created in a Pervasive SQL server. The query execution is based on periods and each query node of the DMG has to finish within its time bound which is worst case execution time of the query. At first the estimated worst case execution time for each diagnostic query is calculated. After that the algorithm optimizes the diagnostic query using a class based query categorization technique. The access method for each query is selected on the basis of its type. For join queries the most optimized join order is calculated by estimating the selectivity factor based on the number of tuples present in each join order. Results presented in this context show that the diagnostic queries are optimized effectively and their estimated worst case execution time is minimized.