This work proposes a hybrid experimental-numerical technique with the aim to improve strain measurements at stress concentration regions. The novel technique is performed employing the computer vision scale invariant feature transform (SIFT) algorithm and meshless methods, here termed SIFT-meshless. The SIFT is applied to perform feature points matching in two images of the specimen surface at different stages of mechanical deformation. The output data are provided as a set of displacement measurements by tracking matched feature points. This information is then used to model displacement and strain field on the surface by means of a meshless formulation based on the moving least squares approximation. By applying the proposed SIFT-meshless method, the strain distribution around a semicircular notch in a plate under bending load was investigated. The experimental results were compared with those obtained by a digital image correlation technique based on a subset approach and to simulations from finite element analysis software. The experimental results demonstrated that the present method is capable of performing reliable strain measurements at distances close to the notch where the peak strain value is expected, even in the presence of high strain gradients.