Research On Snapshot Multispectral 3D Imaging Method

With the development and application of computer vision,how to recover depth from the two-dimensional image information of the target scene quickly and accurately has become an important topic.It plays an important role in the fields of industrial inspection,medical imaging,and robotics.At present,the existing vision-based depth recovery methods cannot meet the requirements of multiple indicators such as accuracy,time efficiency,and measurement range at the same time.In this context,this paper proposes a snapshot multi-spectral 3D imaging method.In this method,images are captured through a snapshot multispectral camera combined with a longitudinal chromatic aberration lens.Multiple spectral images with different degrees of defocus can be acquired by one imaging.These images are analyzed and processed on the basis of defocus blur theory to realize monocular,fast and accurate depth recovery in single frame.The main research contents are as follows:(1)A depth recovery method based on the ratio of edge gradients of multispectral images.This method is based on geometric optical imaging and defocus degradation model.The depth can be recovered by establishing the relationship between the gradient ratio of different images and the real depth.Aiming at this method,a parameter calibration method is proposed.The main process of this method is to scale the curve of point spread function and depth and compare the sharpness of images.In the process of calculating the edge gradient of these images,the Canny operator is used to detect the edge.(2)A depth recovery method based on differential operation of edge gradient of multispectral image.This method uses the relationship curve between the differential operation value of the edge gradient of the scale image and the depth to recover the depth.After restoring the depth information at the edge of the image,the global depth information is calculated through a soft matting algorithm.Then the L0 gradient minimization algorithm is used to correct the recovery results.(3)Near-natural color reproduction of multi-spectral images.With reference to the Bayer array color interpolation algorithm,snapshot multispectral images are separated using bilinear interpolation algorithm.Then the color of the multispectral images are reproduced with the help of white balance algorithm.(4)Verification experiment of the multi-spectral 3D imaging method.This experiment aims to analyze and compare the above two deep recovery methods.It can be seen from the experimental results,that the depth recovery method based on the differential operation of the edge gradient of the multispectral images can achieve depth recovery in the depth range of 5-54 cm,and the deviation is less than ±1cm.The average processing time of a single frame is 0.083 s.This method has a larger measurement range and less processing time.Finally,the color reproduction images are fused with the three-dimensional data.Near-natural color multispectral three-dimensional imaging that is more suitable for the observation of human can be obtainedThe innovation points of the snapshot multi-spectral 3D imaging method proposed in this paper are as follows:(1)Fast and accurate monocular single-frame depth recovery can be realized,with less calculation,lower algorithm complexity,and simpler calibration steps.(2)The characteristic spectrum information of the scene can be acquired while recovering in depth,which is helpful for the feature recognition of threedimensional objects.(3)The device is simple and portable,with low power consumption,which is convenient for outdoor dynamic detection.