A novel self-adaptive volume reconstruction technique (SVRT) based on multiobjective optimization is proposed. Its reconstruction results for asymmetrical 3-D single-peak cosine emission coefficient fields are studied with the numerical simulation of a computer. The results show that this algorithm has faster convergences and higher reconstruction precision, and can reconstruct asymmetrical single-peak cosine emission coefficient fields perfectly with only two views. (The average error is 0.18% with no noise data.) In the experiment of argon-arc plasma diagnosis, the 3-D reconstructions of temperature and ionization coefficient fields are fulfilled with this algorithm combined with the spectrum relative-intensity method.