The　aim　of　this　paper　is　to　propose　an　efficient　solution　to　the　problem　of　the　camera　calibration.　This　method　is　based　on　the　epipolar　geometry　constraintand　active　vision.　At　the　same　time,　the　evaluation　object　function　containing　the　intrinsic　parameters　of　the　camera　is　obtained.　The　intrinsic　parameters　of　the　camera　are　optimized　directly　by　the　genetic　algorithm.　The　proposed　method　only　needs　one　time　translation　of　the　camera　and　doesn＇t　require　any　specialized　hardware　support.　Simulations　and　experiments　with　real　images　show　that　the　technique　can　substantially　improve　the　robustness,　convenience　and　accuracy　of　the　calibration.　It　works　well　in　the　3D　reconstruction　system.