Research On The Key Technologies Of Stereo Vision Based On Infrared And Visible Dual Band Images

The three-dimensional profile is an important external attribute of objects,which is also an important research content of machine vision.The profile reconstruction technology has a wide range of application requirements in national defense,aerospace exploration,industrial inspection,virtual reality and other fields.Binocular stereo vision technology is an important means to restore the three-dimensional information of an object through two-dimensional images.The binocular stereo vision technology of visible band is well developed.Visible light imaging with rich contrast and color,shape information,can identify and measure the target accurately and quickly;while infrared imaging technology may perceive the temperature information of the target,and it has obvious advantages under smoke,heat and other special circumstances.In recent years,the rapid development of image fusion theory provides a new way of expanding the applications of binocular stereo vision system.This paper used the complementary information of infrared and visible images to form the source images of binocular stereo vision system.The key technologies of dual-band stereo vision were studied in order to support the target identification and monitoring in fire rescue,space exploration,UAV navigation and so forth.A binocular stereo vision system based on dual-band images was established.The image sensors of the two bands formed the stereo vision system separately,and they can capture the images simultaneously.The system can switch its base working band according to the features of the environment.The image fusion technology was used to support the different requirements under different circumstances.The main contents of the paper are as follows:1.A dual – band binocular stereo vision system which is adept to take the images captured by the four cameras from two fields and two bands was designed.By adding a dichroic mirror we achieved a roughly consistency of the perspective of the two bands cameras in the same field of view.The calibration scheme of the dual – band binocular stereo vision system was proposed and implemented.Since different material has different thermal radiation rate,through the method of pasting metal sheets on wood board,an infrared calibration target with the same layout as classical checkerboard of the visible target was designed.It can work well without the aid of external heat or electrical components.2.An adaptive image fusion method in different illumination environments was proposed.In order to enhance the system's ability to adapt to the working environment,we added a procedure to analyse the features of the imaging conditions before image fusion.We used fuzzy system to predict the illumination conditions of the environments,and established a Gauss curve in the fuzzy system's output parameter range to determine the dual-band image fusion weight coefficients adaptively.The experiment results show that our method is not sensitive to illumination changes.So it can enhance the system's applicability to different light environments.3.An adaptive image fusion algorithm under smoke environment was proposed.We combined the information of infrared and visible images to deterimine whether the images are affected by smoke.The fusion rule was determined according to the edge integrity of the visible images.The experiment results show that by using our method the fused image's sensitivity to the somke environment can be reduced.4.For the information contained in the infrared and visible images has redundancy,which can be used to improve the stereo matching effect,we proposed to match the two views of information based on the two bands of images.We designed our stereo matching algorithm based on SAD and improved it by changing the size,shape and weights of the matching window and a new similarity measure was formed.It is found that the effects of stereo matching based on the dual-band information show better results than matching with visible images only.5.The profile reconstruction experiments based on dual-band images were taken on different objects and the results verified that our system shows reliability in unsatisfactory working environments.