MULTILEVEL DATA FUSION FOR THE DETECTION OF TARGETS USING MULTISPECTRAL IMAGE SEQUENCES

An approach is presented to the long range automatic detection of vehi cles, using multisensor image sequences, The method is tested on a dat abase of multispectral image sequences, acquired under diverse operati onal conditions, The approach consists of two parts, The first part us es a semisupervised approach, based on texture parameters, for detecti ng stationary targets, Far each type of sensor one learning image is c hosen, Texture parameters are calculated at each pixel of the !earning images and are combined using logistic regression into a value that r epresents the conditional probability that the pixel belongs to a targ et given the texture parameters, The actual detection algorithm applie s the same combination to the texture features calculated on the remai nder of the database (test images), When the results of this feature-l evel fusion are stored as an image, the local maxima correspond to lik ely target positions, These feature-level-fused images are calculated for each sensor. In a sensor fusion step, the results obtained per sen sor are then combined again, Region growing around the local maxima is then used to detect the targets, The second part of the algorithm sea rches for moving targets, To detect moving vehicles, any motion of the sensor must be detected first, if sensor motion is detected, it is es timated using a Markov random field model, Available prior knowledge a bout the sensor motion is used to simplify the motion estimation. The estimate is used to warp past images onto the current image in a tempo ral fusion approach and moving targets are detected by thresholding th e difference between the original and warped images. Decision level fu sion combines the results from both parts of the algorithm, (C) 1998 S ociety of Photo-Optical Instrumentation Engineers.