Limited statistical frame number and strong backscatter interference from smoke result in a photon-starved regime, severely limiting the depth imaging capability of array Gm-APD lidar in smoky environment. Here, we propose a depth image estimation algorithm that can significantly improve the integrity of targets in dense smoke environments when signal photons are starved. At the signal level, the algorithm improves the accuracy of extracting signal photons by constructing multi-scale superpixels. At the image level, using edge information of depth images at different scales to guide and fill the original scale depth image achieves efficient noise reduction and improves target integrity. It has been successfully demonstrated under different attenuation lengths and statistical frame numbers. Compared with other state-of-the-art methods, the proposed algorithm has maximum target recovery and structural similarity, especially for the attenuation length (AL) is 3.6 and statistical frame number is 1500 (imaging time is 75ms). This study is of great significance for the fast depth imaging of dynamic targets by array Gm-APD lidar in dense smoke environments.
Geiger mode Avalanche Photo Diode (Gm-APD) array lidar is a lidar that can perform single-photon detection. It offers a wide range of applications due to its low power consumption, small size, and extended detecting distance. There haven't been many research on this detector's target classification because of its late development and small detector array. The classification technique based on the Gm-APD array lidar point cloud is the focus of this paper's research: Firstly, the Gm- APD array lidar is utilized to perform imaging tests on four targets from various angles in order to create a target classification dataset.Following that, several data preprocessing methods were chosen and implemented based on the characteristics of the obtained data, such as filling in missing values, performing range image and intensity image interpolation, using the principle of keyhole imaging to convert the range image to point cloud data, realizing the information fusion of distance image and intensity image, and using multiple point cloud data enhancement methods. Finally, the point cloud classification networks PointNet and PointNet++ are trained on point cloud data with varying levels of preprocessing, the results are compared and analyzed, and the impact of different preprocessing methods on the classification accuracy of the two networks is determined. Inferences were made and experiments were carried out to verify the inferences. The data set preprocessing method with the highest classification accuracy of the two networks is discovered, laying the groundwork for future Gm-APD lidar target classification and detection research.
The non-uniform distribution of smoke and laser spot seriously limits the imaging ability of single-photon lidar through smoke. To this end, based on the collision theory between photons and smoke particles, this paper establishes the time-domain distribution model of scattered photons (Gamma), which considers the lidar system parameters and the characteristics of smoke particles. The shape parameter and inverse scale parameter are defined as the maximum scattering number K and the average scattering number β, respectively. The indoor test effectively illustrates the estimation ability of the model for smoke. The parameter estimation results show that the average scattering number β and the maximum scattering number K increase linearly with the increase of attenuation coefficient. This study is of great significance for the suppression of smoke backscattering and is expected to improve the weather adaptability of single-photon lidar.
LiDAR echo intensity information can reflect the reflection characteristics of the target surface, and can be used as an important data source in the aspects of LiDAR point cloud image vision, classification and feature extraction. Geiger mode avalanche photodiode (Gm-APD) has the ability of single photon detection and high range sensitivity, and is widely used in the field of lidar. The number of statistics is often taken as the target intensity information obtained. In order to make the intensity image accurately reflect the reflection characteristics of the target surface, a kind of intensity information correction method of Gm-APD lidar is proposed. By eliminating the distortion caused by the detection model and target distance of the detector, the average reflectivity estimation error can be increased from 51.97% to 8.86%. Aiming at Gm-APD lidar, the determination method of parameters in parameter estimation method is systematically described in this paper. On this basis, the calibration of the laser emitter can improve the uniformity of the target, and the standard deviation is increased from 1.1818 to 0.0050. The proposed scheme can provide a reliable data source for target recognition, classification and feature extraction based on Gm-APD intensity image.
Aiming at the problem that the background noise mixed in the target echo will affect the calculation of the target polarization degree when the traditional polarization detection system obtains the target polarization degree, based on the polarization Gm-APD detection model, a set of target echo polarization correction method is proposed. The target is imaged in a xenon lamp environment, the influence of target attitude and polarization angle on detection is explored, and the polarization imaging results are analyzed. The results show that the polarization system has a significant effect on metal materials with low surface roughness. When circularly polarized light is incident, the echo trigger probability of the metal material reaches a peak at the polarization angle of 135°. The greater the incident angle, the greater the echo depolarization and the lower the trigger probability. By inverting the distribution of echo photons, the number of background noise photons in the echo and the number of target echo photons can be obtained respectively, and a more accurate correction of the polarization degree of the target echo can be obtained. For metal materials, when the target attitude angle is 30°, the target polarization before and after correction are 0.47 and 0.57 respectively, and the target echo polarization after correction is 7% higher than that without correction. This research work provides experimental support for the effective detection and target detection of GM-APD lidar in the daytime.
Bi-directional reflection distribution function (BRDF) is a common method to study the laser scattering characteristics of targets, and it is an important parameter for the theoretical demonstration of laser active detection, target recognition and classification. Scholars at home and abroad have proposed many mature BRDF models to describe the scattering characteristics of different targets. However, almost all of these models do not take into account the effect of incident wavelength on scattering characteristics. In addition, limited by the frequency modulation range of the laser, the existing BRDF measurement devices cannot obtain the BRDF data of the target at any wavelength, which restricts the application of the existing BRDF model. In view of this limitation, a method is proposed to calculate the unknown wavelength BRDF data using the BRDF measurement data of known wavelengths. Firstly, based on the Kirchhoff approximation theory, the spatial distribution of the scattered light field of the metal aluminum target at any wavelength was simulated and analyzed. Secondly, the error of the theoretical simulation model was analyzed through the experimental data. Finally, the BRDF data at any wavelength were calculated using the simulation data and the experimental data with known wavelengths. The final results showed that at the 1064nm wavelength, the RMSE value of the calculated data obtained by this method is 0.3553, which is 0.2233 smaller than the RMSE value of the simulation data.This method is effective in calculating the BRDF of metal aluminum targets at different wavelengths.
In this paper, three algorithms are proposed to restore the fog-containing relative intensity image of lidar based on the atmospheric scattering model and dark channel prior theory. The algorithm was evaluated by analyzing the peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) of the two data sets, including the fog-free relative intensity images and the fog-containing relative intensity images and standard fog-free relative intensity images. The experimental results show that the PSNR of the two groups of data can be improved by the three algorithms to varying degrees, and the highest PSNR can reach 35.6%. The structure similarity SSIM was significantly improved, and the effect was up to three orders of magnitude.
The surface reflected wave is formed by the wave of the sea surface when the submarine is sailing underwater. The intensity of the reflected wave decreases with the increase of the depth of the submarine, which makes it difficult to detect the submarine based on the height characteristics of the surface wave. For this problem, the theory of underwater submarine detection is studied based on the relationship between the intensity image of lidar and the normal vector distribution of target surface. First, the two-scale wave theory was proposed to establish a simulation model of the sea surface wave, and the microsurface element normal vector was solved for the gridded sea surface. Under the premise of considering the shielding effect, the intensity image of the sea surface wave including the submarine reflected wave was obtained by coupling the lidar equation with the sea surface wave model. Finally, principal component analysis (PCA) and BP neural network are used to extract and recognize the characteristics of submarine and surface intensity images. The results show that at low wind speed and small wind field, the recognition rate of the submarine with a depth of 19m is more than 90%. With the increase of the depth, wind speed and wind field, the recognition rate decreases gradually. This study provides a new idea for lidar submarine reflection wave detection.
Edges are critical important for the visual appearance of images. The traditional denoising algorithms are difficult to preserve the edges of the image while removing the noise of ICCD sensing image. At the same time, it is difficult to eliminate the problems of image darkness and low resolution caused by uneven illumination. This paper proposes a multilevel filtering image denoising algorithm based on edge information fusion. The target edges detection of the image after non-local means (NL-means) filtering is carried out based on the eight-direction Sobel operator. In order to filter the false edge points and residual noise, an adaptive threshold is determined according to the mean and variance of the eight neighborhood pixels of the detected pixel. Meanwhile, homomorphic filtering is used to enhance the image contrast and uniformity. By comparing the pixel values of the edge image and the homomorphic filtered image, the final denoised image is obtained by fusing the two images. The results indicate that, compared with the traditional algorithms, the edge preserving ability of the proposed algorithm is improved by more than 20%, and the denoising ability is improved by 63.5% for building target. For specific targets (vehicle), the results demonstrate that the proposed algorithm have the maximum edge preserving index and contrast, and the minimum non-uniformity. This algorithm lays a foundation for target segmentation and recognition.
When using Gm-APD Lidar for depth imaging through realistic fog, the echo signal of the target is submerged in the background noise due to the strong absorption and scattering characteristics of the fog particles, resulting in serious defect of the recovered depth image of the target. To solve this problem, this paper proposes a dual-parameter estimation algorithm based on continuous wavelet transform (CWT) and maximum likelihood estimation (MLE) to improve the accuracy of fog signal estimation. Then the target and the fog signal are separated by estimating the fog signal of each pixel. Finally, the depth image of the separated target is processed by cross pixel complement and median filtering algorithms to improve the integrity of the target image. The experimental results show that, compared with the traditional algorithm, the target recovery of the reconstructed image is improved by 0.337, and the relative average ranging error is reduced by 0.3897. This research improves the weather adaptability of Gm-APD Lidar.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.