Efficient camera self-calibration method for remote sensing photogrammetry

High-precision calibration to zoom lens of optical measurement machine based on FNN

We proposed a calibration method for high-precision zoom lenses of optical measurement machines based on Fully Connected Neural Network (FNN), using a 5-layer neural network instead of a camera calibration model, to achieve continuous calibration of zoom lenses at any zoom setting by calibrating typical zooms. From the experimental verification, the average calibration error of this method is 9.83×10^-4mm and the average measurement error at any zoom setting is 0.01317mm. The overall calibration precision is better than that of Zhang's calibration method and can meet the application requirements of a high-precision optical measurement machine. The method proposed in this paper provided a new solution and a new idea for the calibration of zoom lenses, which can be widely used in the fields of precision parts inspection and machine-vision measurement.

Hyperspectral face recognition based on sparse spectral attention deep neural networks

Inspired by the robust capability and outstanding performance of convolutional neural networks (CNN) in image classification tasks, CNN-based hyperspectral face recognition methods are worthy of further exploration. However, hyperspectral imaging poses new challenges including high data dimensionality and interference between bands on spectral dimension. High data dimensionality can result in high computational costs. Moreover, not all bands are equally informative and discriminative. The usage of a useless spectral band may even introduce noises and weaken the performance. For the sake of solving those problems, we proposed a novel CNN framework, which adopted a channel-wise attention mechanism and Lasso algorithm to select the optimal spectral bands. The framework is termed as the sparse spectral channel-wise attention-based network (SSCANet) where the SSCA-block focuses on the inter-band channel relationship. Different from other methods which usually select the useful bands manually or in a greedy fashion, SSCA-block can adaptively recalibrate spectral bands by selectively emphasizing informative bands and suppressing less useful ones. Especially, a Lasso constraint strategy can zero out the bands during the training of the network, which can boost the training process by making the weights of bands sparser. Finally, we evaluate the performance of the proposed method in comparison of other state-of-the-art hyperspectral face recognition algorithms on three public datasets HK-PolyU, CMU, and UWA. The experimental results demonstrate that SSCANet based method outperforms the state-of-the-art methods for face recognition on the benchmark.

Pre-flight calibration of a multi-angle polarimetric satellite sensor directional polarimetric camera

The directional polarimetric camera (DPC), developed by Anhui Institute of Optics and Fine Mechanics, Chinese Academy of Science, is a satellite sensor used to observe the polarization and directionality of the earth's reflectance. It acquires the two-dimensional image of the earth with a large field of view (118.74°) and a high spatial resolution (3.3 km) in 8 spectral bands. The first DPC was successfully launched onboard the GaoFen-5 satellite in May 2018, subject to the Chinese high-resolution earth observation program. In this paper, a set of systematic and complete pre-flight calibrations of the DPC are proposed to ensure the effective characterization for in-flight calibration, so as to ensure the accuracy of DPC measured radiation polarization data and the reliability of inversion results. Since the geometric calibration method of the DPC has been presented in an early companion paper [Appl. Opt. 59 226 (2020)], this paper will not introduce it in detail. Instead, the geometric calibration results of each spectral band together with a discussion on the origin of differences between spectral bands are analyzed, and the error analysis of the method is conducted. The results of the DPC geometric calibration is that the residuals of all spectral bands are less than 0.1 pixel. For radiometric calibration, the radiometric models of non-polarized bands and polarized bands are derived in detail, respectively, and the specific calibration methods with error analysis, equipment, and main results with their related accuracies for each parameter of the radiometric models are described. To verify the accuracy of calibration parameters, a series of polarization detection accuracy verification experiments based on a non-polarized radiation source, a polarizing system, and a natural scene were carried out. The experimental results show that the maximum deviation of degree of polarization between the set values of the polarizing system and measured values of the DPC at the corresponding positions of four field of view angles of 0, 15, 30, and 45 degrees of each polarized spectral band is 0.009, 0.004, and 0.003, respectively. The average error in measuring the degree of polarization of a non-polarized light source by all pixels in the three polarized bands is 0.0043, 0.0046, and 0.0037, respectively. And the relative deviations of each field of view are within 0.020 when the DPC and CE318N simultaneously measure the DoLP of sky. All of these prove the effectiveness of the pre-flight calibration.

Geometric calibration method based on a two-dimensional turntable for a directional polarimetric camera

The directional polarimetric camera (DPC) is a polarization sensor with ultra-wide-angle and low-distortion imaging characteristics. Geometric calibration is usually the first essential step before remote sensing satellites are launched. In this paper, a geometric calibration method based on a two-dimensional turntable and a rotation matrix with high precision, simple operation, and wide application range is proposed for the directional polarimetric camera. Instead of precisely adjusting the position of the sensor on the turntable, the method effectively eliminates the errors caused by the mechanical axis of the turntable and the optical axis of the sensor not being adjusted to the same direction through the rotation transformation of the coordinate system. The geometric calibration experiments of the directional polarimetric camera were carried out with the method of Chen et al. [Optik121, 486 (2010)10.1016/j.ijleo.2008.08.004OTIKAJ0030-4026] and the proposed method. The experimental results showed the calibration residual of the proposed method was less than 0.1 pixel while Chen's method was 0.3 pixel. The mean reprojection error and root mean square error of the proposed method were reduced to 0.06352 pixel and 0.06961 pixel, respectively. The geometric calibration parameters obtained by the proposed method were used for geometric correction of the in-orbit images of the DPC, and the results also prove the effectiveness and superiority of the proposed method.

Research on Nonlinear Control Method of Underactuated Gantry Crane Based on Machine Vision Positioning

The movement of the gantry crane is controlled by an symmetry underactuated system, and has poor robustness in precise positioning. A new active control method based on the machine vision positioning is proposed in this paper, and the trajectories are planned after the detection of starting and ending points. A new type of energy storage function is given in this paper, and a coupling control law is derived to minimize the load swing in the process of precise positioning. The equilibrium point of the closed-loop system is checked though Lyapunov and LaSalle’s theorems, and the calculation results are verified through experimental investigations. The results show that the equilibrium points are asymptotically stable, and the proposed control method is of better robustness. The findings provide a new kind of control method with higher efficiency, and can help with the precise control of gantry cranes.

Flexible calibration method of an FPP system based on a geometrical model and NLSM with fewer parameters

Fringe projection profilometry (FPP) technology is an important method for 3D reconstruction. In this paper, we proposed a flexible calibration method of an FPP system based on the imaging principle and geometrical structure of the system. The target coordinates are only related to its pixel coordinates and phase. First, the fringe images are projected onto the calibration plate, and the phase can be calculated through the four-step phase-shifting method. Then, the pixel coordinates of the feature points can be located with the binarized fringe images and the centroid method. Finally, the calibration parameters are calculated by the nonlinear least-squares method (NLSM). The reconstructed experiment of 162 testing points was carried out, and the result shows that the maximum relative errors on coordinates X, Y, and h are 0.27%, 0.42%, and 0.59%, respectively. The other two surface reconstruction experiments also verify the feasibility of the calibration method.