Artificial intelligence based target analysis with laser signals

Although classification of target types is popular in the field of computer vision, it is still a challenging task. Target types can be in the form of material types, colors, gender and living/inanimate objects depending on the area of study. In target detection systems, cameras are commonly used. The use of cameras in low and high light environments adversely affects the performance of the systems and even completely blocks the operation of the system. In addition of these, the use of cameras causes privacy concerns these days. In this thesis study, by taking advantage of the unidirectional and coherent propagation of lasers and the interaction of light with objects, the detection of types of materials, the detection of a person's gender, the detection of the color of a material and the detection of living /inanimate objects were carried out contactless at a long distance. For this purpose, laser light was reflected to the target and the laser light reflected from the target was recorded as a digital signal with the help of an electro-optical receiver sensor. These recorded laser signals were trained in the deep learning network after the data preparation steps and the target type was determined. In the experimental applications made within the scope of the target type identification study, the detection of 10 different types of materials that are often used in daily life was carried out with an average success rate of 93.62%. Using a single laser signal taken from the subjects' arm, gender determination was performed with an average success rate of 76.2%. The detection of 9 different colors was done with an average success rate of 91.3%, and the detection of living/inanimate objects with an average success rate of 99.7%.In this thesis, the effect of different atmospheric conditions such as different humidity rates, fog, daylight and darkness on the proposed target detection approaches has also been studied. In this thesis study, it was revealed that non-contact detection of a target type at a long distance using artificial intelligence-based approaches proposed with a single low-power laser signal can be performed with high performance.