High dynamic stripe image enhancement for reliable center extraction in complex environment

High-precision laser vision sensors, as an important mean of 3D data acquisition, are being widely used for online high-speed measurement in complex environment. However, due to the complexity of the scene environment, the stripe images show high dynamic, low signal to noise ratio, which seriously affect the reliability and detection accuracy. In this paper, a high dynamic stripe image brightness enhancement method is proposed. The Retinex theory is adopted to enhance stripe brightness, which makes the gray value of stripe in cross-section is close to Gaussian distribution. Meanwhile, the stripe is segmented correctly based on the reflectivity, which ensures low quality stripes can be enhanced correctly. In that case, the central coordinate of stripe can be easily extracted. Finally, a multi-thread flow acceleration and serialized output program are used to meet the requirement of real-time measurement. The physical experiments indicate that this method can improve the system reliability greatly, which has important practical application value.