Place recognition is an important technique for autonomous mobile robotic applications. While single-modal sensor-based approaches have shown satisfactory performance, cross-modal place recognition remains underexplored due to the challenge of bridging the cross-modal heterogeneity gap. In this work, we introduce an instance-aware cross-modal place recognition approach, named InsCMPR. We design a novel instance-aware modality alignment module, which aligns multi-modal data at both pixel-level and instance-level by leveraging a pre-trained vision foundation model SAM. Then a novel dual-branch hybrid Mamba-Transformer network is proposed to efficiently enhance the distinctiveness of the produced descriptors by integrating global features with local instance features. Experimental results on the KITTI, NCLT, and HAOMO datasets show that our proposed methods achieve state-of-the-art performance while operating in real time. We will open source the implementation of our method at: https://github.com/nubot-nudt/InsCMPR.

As lunar exploration missions grow increasingly complex, ensuring safe and autonomous rover-based surface exploration has become one of the key challenges in lunar exploration tasks. In this work, we have developed a lunar surface simulation system called the Lunar Exploration Simulator System (LESS) and the LunarSeg dataset, which provides RGB-D data for lunar obstacle segmentation that includes both positive and negative obstacles. Additionally, we propose a novel two-stage segmentation network called LuSeg. Through contrastive learning, it enforces semantic consistency between the RGB encoder from Stage I and the depth encoder from Stage II. Experimental results on our proposed LunarSeg dataset and additional public real-world NPO road obstacle dataset demonstrate that LuSeg achieves state-of-the-art segmentation performance for both positive and negative obstacles while maintaining a high inference speed of approximately 57 Hz. We have released the implementation of our LESS system, LunarSeg dataset, and the code of LuSeg at: https://github.com/nubot-nudt/LuSeg.

Multimodal 3D object detectors leverage the strengths of both geometry-aware LiDAR point clouds and semantically rich RGB images to enhance detection performance. However, the inherent heterogeneity between these modalities, including unbalanced convergence and modal misalignment, poses significant challenges. Meanwhile, the large size of the detection-oriented feature also constrains existing fusion strategies to capture long-range dependencies for the 3D detection tasks. In this work, we introduce a fast yet effective multimodal 3D object detector, incorporating our proposed Instance-level Contrastive Distillation (ICD) framework and Cross Linear Attention Fusion Module (CLFM). ICD aligns instance-level image features with LiDAR representations through object-aware contrastive distillation, ensuring fine-grained cross-modal consistency. Meanwhile, CLFM presents an efficient and scalable fusion strategy that enhances cross-modal global interactions within sizable multimodal BEV features. Extensive experiments on the KITTI and nuScenes 3D object detection benchmarks demonstrate the effectiveness of our methods. Notably, our 3D object detector outperforms state-of-the-art (SOTA) methods while achieving superior efficiency. The implementation of our method has been released as open-source at: https://github.com/nubot-nudt/ICD-Fusion.

According to reresearch, the corner detection-based threshold segmentation algorithm has been found to outperform the Otsu method in handling mixed pixels in canopy imaging. This paper aims to describe the principle of the corner detection-based threshold segmentation algorithm and compare its classification performance to that of the Otsu method through experiments. The results indicate that compared to the Otsu method, the corner detection-based threshold segmentation algorithm achieves higher accuracy in classifying mixed pixels, preserves more canopy information in overexposed areas of the image, effectively reduces the misclassification of mixed pixels, and greatly improves the accuracy of hemispherical photography leaf area index inversion results. Specifically, the correlation coefficient R2of the corner detection-based threshold segmentation algorithm is shown to increase from 0.8 to 0.9, demonstrating its superior performance.

In the indirect measurement of Leaf Area Index (LAI), previous studies commonly assumed that the actual LAI can be obtained by subtracting the Woody Area Index (WAI) from the Plant Area Index (PAI), which includes the entire plant. This can be calculated using the equation LAI = PAI - WAI. However, this study, conducted using 3D model simulations and the DHP (Digital Hemispherical Photography) method to measure LAI, discovered that this conclusion may not be correct. The LAI value obtained by subtracting WAI directly from PAI showed significant discrepancies from the true LAI value in the simulated environment. Through the re-derivation and verification of the LAI calculation equation, this study proposes a more reasonable explanation for this discrepancy and presents the following conclusions based on the 3D simulation experiment and hemispherical photography method: using PAI-WAI to calculate LAI leads to overestimation of the values, while the calculation method described in this study obtained more accurate LAI values in the simulation experiments (with higher R2, lower RMSE and MAE).

Studies have shown that camera auto-exposure underestimates the LAI (leaf area index) measured by DHP (digital hemispheric photography) to varying degrees. To address this problem, this paper proposes the use of multi-exposure fusion to reconstruct information from canopy images to compensate for the loss of information caused by overexposure or underexposure of canopy images acquired by the camera in auto-exposure mode. By fusing a series of canopy images with different exposure times from the same canopy layer, the LAI is then calculated using DHP on the fused image. Experimental results show that the method improves the R2from 0.698 to 0.837 and reduces the RMSE from 0.87 to 0.37 compared with the automatic exposure mode of the camera, with LAI-2200 measurements as a reference. This method contributes to resolving the problem of underestimating LAI in DHP caused by the automatic exposure mode, thereby improving the accuracy of DHP.

Hemispherical photography is an important method of leaf area index (LAI) measurement, but the intermediate processes such as image segmentation and clumping index estimation will introduce errors. In this paper, an end-to-end model was proposed to directly estimate LAI from a hemispheric image based on GhostNet, which avoids errors introduced by the intermediate processing. Hemispherical images of crops and forest vegetation obtained from Shihezi, Xinjiang and Xiong'an, Hebei, and the measured values using LAI-2200 were utilized as the dataset. Compared with the LAI-2200 measurement results, the analysis results show that the correlation between them is extremely significant, with$\mathrm{R}^{2}=0.80,\ \text{RMSE}=0.65,\ \text{MAE}=0.46$. The experimental results show that the estimation model based on GhostNet can accurately estimate the LAI of the hemispheric image and is suitable for timely and accurate estimation of the LAI value of various types of vegetation through edge devices.

Optimal exposure time is essential for accurate measurement of LAI (leaf area index) by DHP (digital hemispheric photography) method, The results showed that LAI was underestimated in different degrees under automatic exposure mode. To address this problem, in this paper we constructed a model of light intensity and optimal exposure time by studying the quantitative relationship between light intensity under the canopy and exposure time. The result shows that using the exposure based on this model rather than the automatic exposure, the comparison of LAI from the LAI-2200 and digital photographs is greatly improved, with R2 increasing from 0.398 to 0.845, and RMSE decreasing from 1.298 to 0.293. The method helps to solve the uncertainty of optimal exposure time in DHP method and improve the accuracy of DHP method in LAI measurement.