No abstract

Modeling forest fire spread is a very complex problem, and the existing models usually need some input parameters which are hard to get. How to predict the time series of forest fire spread rate based on passed series may be a key problem to break through the current technical bottleneck. In the process of forest fire spreading, spread rate and wind speed would affect each other. In this paper, three kinds of network models based on Long Short-Term Memory (LSTM) are designed to predict fire spread rate, exploring the interaction between fire and wind. In order to train these LSTM-based models and validate their effectiveness of prediction, several outdoor combustion experiments are designed and carried out. Process data sets of forest fire spreading are collected with an infrared camera mounted on a UAV, and wind data sets are recorded using a anemometer simultaneously. According to the close relationship between wind and fire, three progressive LSTM based models are constructed, which are called CSG-LSTM, MDG-LSTM and FNU-LSTM, respectively. A Cross-Entropy Loss equation is employed to measure the model training quality, and then prediction accuracy is computed and analyzed by comparing with the true fire spread rate and wind speed. According to the performance of training and prediction stage, FNU-LSTM is determined as the best model for the general case. The advantage of FNU-LSTM is further demonstrated by doing comparison experiments with the normal LSTM and other LSTM based models which predict both fire spread rate and wind speed separately. The experiment has also demonstrated the ability of the model to the real fire prediction on the basis of two historical wildland fires.

A single VLP-16 LiDAR estimation method based on a single-frame 3D laser point cloud is proposed to address the problem of estimating negative obstacles’ geometrical features in structured environments. Firstly, a distance measurement method is developed to determine the estimation range of the negative obstacle, which can be used to verify the accuracy of distance estimation. Secondly, the 3D point cloud of a negative obstacle is transformed into a 2D elevation raster image, making the detection and estimation of negative obstacles more intuitive and accurate. Thirdly, we compare the effects of a StatisticalOutlierRemoval filter, RadiusOutlier removal, and Conditional removal on 3D point clouds, and the effects of a Gauss filter, Median filter, and Aver filter on 2D image denoising, and design a flowchart for point cloud and image noise reduction and denoising. Finally, a geometrical feature estimation method is proposed based on the elevation raster image. The negative obstacle image in the raster is used as an auxiliary line, and the number of pixels is derived from the OpenCV-based Progressive Probabilistic Hough Transform to estimate the geometrical features of the negative obstacle based on the raster size. The experimental results show that the algorithm has high accuracy in estimating the geometric characteristics of negative obstacles on structured roads and has a practical application value for LiDAR environment perception research.

Motion sensors in smart wristbands/watches have been widely used to sense users' level of movement and animation. Some studies have further recognised activity contexts using these sensors, such as walking, sitting and running. However, in applications requiring understanding of more complex activities such as interactions with other people or objects, it is necessary to recognise the fine-grained arm action during user interactions with other people or objects. A method to recognise a set of arm actions on a fine-grained level (e.g. checking the wristband, drinking water etc.) is proposed. Motion signals from the accelerometer and gyroscope are transformed into the frequency domain using the short-time Fourier transform. Then, the action patterns are represented by the motion spectrum mixture model and action dynamics are modelled by continuous density hidden Markov models. Tested on a dataset collected from 23 subjects, the method shows satisfying performance and efficiency.