A Multiview SAR Target Recognition Method Using Inner Correlation Analysis

This paper proposes a synthetic aperture radar (SAR) image target recognition method using multiple views and inner correlation analysis. Due to the azimuth sensitivity of SAR images, the inner correlation between multiview images participating in recognition is not stable enough. To this end, the proposed method first clusters multiview SAR images based on image correlation and nonlinear correlation information entropy (NCIE) in order to obtain multiple view sets with strong internal correlations. For each view set, the multitask sparse representation is used to reconstruct the SAR images in it to obtain high-precision reconstructions. Finally, the linear weighting method is used to fuse the reconstruction errors from different view sets and the target category is determined according to the fusion error. In the experiment, the tests are conducted based on the MSTAR dataset, and the results validate the effectiveness of the proposed method.

View Synthesis: LiDAR Camera versus Depth Estimation

Depth-Image-Based Rendering (DIBR) can synthesize a virtual view image from a set of multiview images and corresponding depth maps. However, this requires an accurate depth map estimation that incurs a high compu- tational cost over several minutes per frame in DERS (MPEG-I’s Depth Estimation Reference Software) even by using a high-class computer. LiDAR cameras can thus be an alternative solution to DERS in real-time DIBR ap- plications. We compare the quality of a low-cost LiDAR camera, the Intel Realsense LiDAR L515 calibrated and configured adequately, with DERS using MPEG-I’s Reference View Synthesizer (RVS). In IV-PSNR, the LiDAR camera reaches 32.2dB view synthesis quality with a 15cm camera baseline and 40.3dB with a 2cm baseline. Though DERS outperforms the LiDAR camera with 4.2dB, the latter provides a better quality-performance trade- off. However, visual inspection demonstrates that LiDAR’s virtual views have even slightly higher quality than with DERS in most tested low-texture scene areas, except for object borders. Overall, we highly recommend using LiDAR cameras over advanced depth estimation methods (like DERS) in real-time DIBR applications. Neverthe- less, this requires delicate calibration with multiple tools further exposed in the paper.

Repeated Multiview Imaging for Estimating Seedling Tiller Counts of Wheat Genotypes Using Drones

Early generation breeding nurseries with thousands of genotypes in single-row plots are well suited to capitalize on high throughput phenotyping. Nevertheless, methods to monitor the intrinsically hard-to-phenotype early development of wheat are yet rare. We aimed to develop proxy measures for the rate of plant emergence, the number of tillers, and the beginning of stem elongation using drone-based imagery. We used RGB images (ground sampling distance of 3 mm pixel-1) acquired by repeated flights (≥ 2 flights per week) to quantify temporal changes of visible leaf area. To exploit the information contained in the multitude of viewing angles within the RGB images, we processed them to multiview ground cover images showing plant pixel fractions. Based on these images, we trained a support vector machine for the beginning of stem elongation (GS30). Using the GS30 as key point, we subsequently extracted plant and tiller counts using a watershed algorithm and growth modeling, respectively. Our results show that determination coefficients of predictions are moderate for plant count (R2=0.52), but strong for tiller count (R2=0.86) and GS30 (R2=0.77). Heritabilities are superior to manual measurements for plant count and tiller count, but inferior for GS30 measurements. Increasing the selection intensity due to throughput may overcome this limitation. Multiview image traits can replace hand measurements with high efficiency (85–223%). We therefore conclude that multiview images have a high potential to become a standard tool in plant phenomics.

Low-delay single holoscopic 3D computer-generated image to multiview images

Due to the nature of holoscopic 3D (H3D) imaging technology, H3D cameras can capture more angular information than their conventional 2D counterparts. This is mainly attributed to the macrolens array which captures the 3D scene with slightly different viewing angles and generates holoscopic elemental images based on fly’s eyes imaging concept. However, this advantage comes at the cost of decreasing the spatial resolution in the reconstructed images. On the other hand, the consumer market is looking to find an efficient multiview capturing solution for the commercially available autostereoscopic displays. The autostereoscopic display provides multiple viewers with the ability to simultaneously enjoy a 3D viewing experience without the need for wearing 3D display glasses. This paper proposes a low-delay content adaptation framework for converting a single holoscopic 3D computer-generated image into multiple viewpoint images. Furthermore, it investigates the effects of varying interpolation step sizes on the converted multiview images using the nearest neighbour and bicubic sampling interpolation techniques. In addition, it evaluates the effects of changing the macrolens array size, using the proposed framework, on the perceived visual quality both objectively and subjectively. The experimental work is conducted on computer-generated H3D images with different macrolens sizes. The experimental results show that the proposed content adaptation framework can be used to capture multiple viewpoint images to be visualised on autostereoscopic displays.

Incremental Instance-Oriented 3D Semantic Mapping via RGB-D Cameras for Unknown Indoor Scene

Scene parsing plays a crucial role when accomplishing human-robot interaction tasks. As the “eye” of the robot, RGB-D camera is one of the most important components for collecting multiview images to construct instance-oriented 3D environment semantic maps, especially in unknown indoor scenes. Although there are plenty of studies developing accurate object-level mapping systems with different types of cameras, these methods either process the instance segmentation problem in completed mapping or suffer from a critical real-time issue due to heavy computation processing required. In this paper, we propose a novel method to incrementally build instance-oriented 3D semantic maps directly from images acquired by the RGB-D camera. To ensure an efficient reconstruction of 3D objects with semantic and instance IDs, the input RGB images are operated by a real-time deep-learned object detector. To obtain accurate point cloud cluster, we adopt the Gaussian mixture model as an optimizer after processing 2D to 3D projection. Next, we present a data association strategy to update class probabilities across the frames. Finally, a map integration strategy fuses information about their 3D shapes, locations, and instance IDs in a faster way. We evaluate our system on different indoor scenes including offices, bedrooms, and living rooms from the SceneNN dataset, and the results show that our method not only builds the instance-oriented semantic map efficiently but also enhances the accuracy of the individual instance in the scene.