Mobile applications have become a ubiquitous part of our daily life, providing users with access to various services and utilities. Text input, as an important interaction channel between users and applications, plays an important role in core functionality such as search queries, authentication, messaging, etc. However, certain special text (e.g., -18 for Font Size) can cause the app to crash, and generating diversified unusual inputs for fully testing the app is highly demanded. Nevertheless, this is also challenging due to the combination of explosion dilemma, high context sensitivity, and complex constraint relations. This paper proposes InputBlaster which leverages the LLM to automatically generate unusual text inputs for mobile app crash detection. It formulates the unusual inputs generation problem as a task of producing a set of test generators, each of which can yield a batch of unusual text inputs under the same mutation rule. In detail, InputBlaster leverages LLM to produce the test generators together with the mutation rules serving as the reasoning chain, and utilizes the in-context learning schema to demonstrate the LLM with examples for boosting the performance. InputBlaster is evaluated on 36 text input widgets with cash bugs involving 31 popular Android apps, and results show that it achieves 78% bug detection rate, with 136% higher than the best baseline. Besides, we integrate it with the automated GUI testing tool and detect 37 unseen crashes in real-world apps.

LiDAR is an active remote sensing technology, which have been used in forest monitoring over twenty years. The separation of branches and leaves is a key prerequisite for estimating some forest parameters. At present, branch and leaf separation algorithms based on terrestrial laser scanning (TLS) point cloud data are matured and numerous. However, they are unable to meet the requirements of large-scale forest inventory and research using UAV-LiDAR data. This study proposed a novel branch and leaf separation method mainly utilizing the near-infrared intensity information of UAV-LiDAR data, as the reflectance of branch and leaf is quite different. This method enabled branch and leaf separation over large scene areas with a straightforward, efficient, and less interactive way. Our study would improve the ability of forest resource investigation, management, and ecological research over large scales.

Terrestrial lidar is capable of resolving trees at the branch/leaf level with accurate and dense point clouds. The separation of leaf and wood components is a prerequisite for the estimation of branch/leaf-scale biophysical properties and realistic tree model reconstruction. Most existing methods have been tested on trees with similar structures; their robustness for trees of different species and sizes remains relatively unexplored. This study proposed a new graph-based leaf–wood separation (GBS) method for individual trees purely using the xyz-information of the point cloud. The GBS method fully utilized the shortest path-based features, as the shortest path can effectively reflect the structures for trees of different species and sizes. Ten types of tree data—covering tropical, temperate, and boreal species—with heights ranging from 5.4 to 43.7 m, were used to test the method performance. The mean accuracy and kappa coefficient at the point level were 94% and 0.78, respectively, and our method outperformed two other state-of-the-art methods. Through further analysis and testing, the GBS method exhibited a strong ability for detecting small and leaf-surrounded branches, and was also sufficiently robust in terms of data subsampling. Our research further demonstrated the potential of the shortest path-based features in leaf–wood separation. The entire framework was provided for use as an open-source Python package, along with our labeled validation data.