Inclination angle effect on landmine characteristics estimation in sandy desert using neural networks

Many places in the world are contaminated with Landmines, normally buried under shallow or deep layers of sand and mud, which causes landmine detection and/or removal to be challenging tasks. To design a reliable landmine sensing system some deep analysis and many test cases are required. In this paper, existence of landmine under the ground surface is examined and its inclination angle effect on detection is analyzed applying finite element method and artificial neural networks. Inverse analyses are used to produce ‘forward results’. Applying a contact pressure (lower than the expected landmine activation pressure) on the ground containing a landmine under its surface would produce a pressure distribution that is dependent on the landmine type, depth and inclination. COMSOL Multi-physics is applied to model sandy soil contaminated by two landmines of different types at different depths and surface pressure distribution is obtained applying external pressure load of 1kPa. Three NNs are trained applying the obtained surface pressure distribution data. The first NN is of perceptron type which classifies the introduced objects in sand. The other two NNs are of feed-forward NN type and are developed for estimating depths of two landmines of different types, one for each. The Landmine inclination angles (0°–30°) effect on detection rate is studied. The results are tabulated and justified. The results show that the anti-tank landmine is fully detected, while the anti-personnel landmine is only detected with a rate of 75%. It is also shown that landmine characteristics estimation is reliable when its inclination angle is small.