Single-image rectification technique in forensic science

Application of Kinect gaming sensor in forensic science

Kinect sensor appears as a low-cost option for 3D modeling. This manuscript describes a methodology to test the applicability of Kinect to crime scenes. The methodology includes the comparison versus well-established scanners (Faro and Trimble). The parameters used for the comparison are the quality in the fitting of primitives, a qualitative evaluation of facial data, the data quality for different ranges, and the accuracy in the measurement of different lengths. The results show that the Kinect noise level increases with range, from 5 mm at 1.5 m range to 15 mm at 3 m range. It is considered that for detail measurements the sensor must be placed close to the target. A general measurement of a sample crime scene was analyzed. Errors in length measurements are between 2% and 10% for 3 m range. The measurement range must be limited to c. 3 m.