This paper describes a novel assembly approach for chloroplast genomes. It contains two modular steps. In the first step, based on the hypothesis that chloroplasts genomes are over-represented compared to the nuclear genome in the plant's cell, we assemble contigs with a De Bruijn graph based approach using short reads with a high k-mer coverage. Connections between oriented contigs are also provided here. The second step determines the order and the orientation of the contigs (scaffolding). Taking advantage of the knowledge that chloroplast genomes posses well studied circular structure, we develop a particular formulation of the scaffolding problem, called Nested Inverted Fragments Scaffolding. It aims to assemble highly conserved inverted repeats. We formulate it as an optimisation problem and we prove that it is NP-Complete. To solve the problem we propose and implement an integer linear programming formulation. We evaluate our method on a set of real instances (a benchmark of 42 chloroplast genomes) and show that it obtains notable achievements with respect to the quality of the results. To further estimate the performance of our scaffolding module, we test it on huge artificially created instances. The results demonstrate an excellent behaviour of our integer formulation as even very large instances have been solved at the first Branch & Bounds node.