We present a framework for edge detection in range images acquired by a time of flight laser sensor. Our edge detection approach is inspired by [Edge Detection in Range Images Based on Scan Line Approximation], in the context of which edge detection via scan line approximation with geometric parametric models is performed. The main drawback of this edge detector, namely the scan line over-segmentation problem is addressed by the introduction of a simple merging step. In addition, we incorporate a method for detection of the noisy data points created by the effect of laser beam splitting between surfaces of different ranges. Finally, a procedure for fine localization of the edge points is introduced. Experimental results on a variety of target object configurations demonstrate that our edge detection framework exhibits increased robustness and accuracy with regard to [Edge Detection in Range Images Based on Scan Line Approximation]. These characteristics in combination with the computational efficiency of our approach, allows for its usage as a component of a real time system for automatic unloading of piled box-like objects.