Unexploded ordnance (UXO) pose a significant threat to post-conflict communities, and current efforts to locate bombs rely on time-intensive and dangerous in-person enumeration. Very high resolution (VHR) sub-meter satellite images may offer a low-cost and high-efficiency approach to automatically detect craters and estimate UXO density. Machine-learning methods from the meteor crater literature are ill-suited to find bomb craters, which are smaller than meteor craters and have high appearance variation, particularly in spectral reflectance and shape, due to the complex terrain environment. A two-stage learning-based framework is created to address these challenges. First, a simple and loose statistical classifier based on histogram of oriented gradient (HOG) and spectral information is used for a first pass of crater recognition. In a second stage, a patch-dependent novel spatial feature is developed through dynamic mean-shift segmentation and SIFT descriptors. We apply the model to a multispectral WorldView-2 image of a Cambodian village, which was heavily bombed during the Vietnam War. The proposed method increased true bomb crater detection by over 160 percent. Comparative analysis demonstrates that our method significantly outperforms typical object-recognition algorithms and can be used for wide-area bomb crater detection. Our model, combined with declassified records and demining reports, suggests that 44 to 50 percent of the bombs in the vicinity of this particular Cambodian village may remain unexploded.
In this paper, a strain-gauge-type force sensor which can be directly mounted on the existing drop-weight device and used to calibrate the piezoelectric high-pressure sensor is designed. The absolute quasistatic pressure calibration principle and the measurement principle of the strain-gauge-type force sensor are described. Based on the mathematical relation model between force and pressure, the influence factors of pressure calibration accuracy are discussed, and the empirical formula for correcting the peak pressure is obtained. To analyze the mechanical strength of the force sensor, the modal parameters of the force sensor in the impact direction, and the pressure loss in the transmission process, a finite element simulation based on ANSYS software is performed. In addition, the static characteristic of the force sensor and the dynamic characteristic of the new drop-weight system composed of the force sensor, the mounting frame, and weight stacks are analyzed. To verify the accuracy of the empirical formula, the absolute quasistatic pressure calibration experiment is performed based on the calibration device. The final experimental results show that the accuracy of pressure calibration can be effectively improved by means of empirical formula correction. Taking the peak pressure measured by the reference pressure sensors as the standard value, the maximum calculation difference can be reduced from 3.42% (obtained by the direct calculation method) to 0.70% by the empirical formula correction method.
This paper applies back propagation neural network (BPNN) optimized by genetic algorithm (GA) for the prediction of pressure generated by a drop-weight device and the quasi-static calibration of piezoelectric high-pressure sensors for the measurement of propellant powder gas pressure. The method can effectively overcome the slow convergence and local minimum problems of BPNN. Based on test data of quasi-static comparison calibration method, a mathematical model between each parameter of drop-weight device and peak pressure and pulse width was established, through which the practical quasi-static calibration without continuously using expensive reference sensors could be realized. Compared with multiple linear regression method, the GA-BPNN model has higher prediction accuracy and stability. The percentages of prediction error of peak pressure and pulse width are less than 0.7% and 0.3%, respectively.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.