MEASUREMENTS AND MODELS FOR RELATING THE PHYSICAL CHARACTERISTICS OF IMAGES TO TARGET DETECTION. AMRL-TR-64-117.
暂无分享,去创建一个
Abstract : Three metrics for predicting the time required to identify targets in high-resolution radar pictures were developed. One metric, based on four automatically measured variables related to transmissivity of positive transparencies, was tested. Through multiple-regression analysis, a correlation of 0.69 between observed and predicted identification times was obtained. When these relationships were applied to a new set of radar pictures and new test subjects, the correlation coefficient was too low to be significantly different from zero with the number of pictures used. A principal reason for the low correlation was the unexpected correlation between the four transmissivity variables. The metric also had an insignificant correlation coefficient when applied to optical and infrared photographs (0.07 and 0.04, respectively). It is hypothesized that, if additional variables are utilized, the metric examined in this study may be useful as a base to develop a more effective prediction equation.
[1] G. C. Sziklai. Some studies in the speed of visual perception , 1957 .
[2] E. Jaynes. Information Theory and Statistical Mechanics , 1957 .
[3] H. P. Birmingham,et al. A Design Philosophy for Man-Machine Control Systems , 1954, Proceedings of the IRE.
[4] W. R. Bush,et al. Recognition of forms against a complex background. , 1956, Journal of the Optical Society of America.