With the rapid development of autonomous driving technology, automated valet parking has imposed more stringent requirements on environmental perception, especially in the demanding underground parking scenarios featuring complex structures and diverse lighting conditions. Semantic segmentation, as a pivotal part of environmental perception, enables the vehicle to better understand its surroundings by assigning semantic labels to every pixel in the image. This study delves into the application of multi-scale augmentation techniques in semantic segmentation, pinpoints their limitations in underground parking scenarios, and puts forward two novel strategies: post-maximum augmentation and post-minimum augmentation. Experiments carried out on the AVM-SemSeg dataset validate that these strategies remarkably boost the model’s performance in such scenarios. Based on this, the post-min-max (PMM) augmentation method has been further developed, integrating the advantages of the two aforementioned novel strategies. It elevates the model’s predictive confidence in true positive samples and curbs the incidence of false positive misclassifications as background. The computational burden is also alleviated by reducing the scale inputs, which enhances the efficiency of the model’s multi-scale inference while still maintaining competitive results. The results manifest that the PMM method outperforms existing multi-scale inference information fusion methods in semantic segmentation tasks and can be combined with all the tested methods to notably enhance their performance on this dataset. This research not only augments the semantic segmentation capabilities of autonomous vehicles in underground parking environments but also offers vital technical support for automated valet parking tasks.

Accurate localization in challenging garage environments—marked by poor lighting, sparse textures, repetitive structures, dynamic scenes, and the absence of GPS—is crucial for automated valet parking (AVP) tasks. Addressing these challenges, our research introduces AVM-SLAM, a cutting-edge semantic visual SLAM architecture with multi-sensor fusion in a bird’s eye view (BEV). This novel framework synergizes the capabilities of four fisheye cameras, wheel encoders, and an inertial measurement unit (IMU) to construct a robust SLAM system. Unique to our approach is the implementation of a flare removal technique within the BEV imagery, significantly enhancing road marking detection and semantic feature extraction by convolutional neural networks for superior mapping and localization. Our work also pioneers a semantic prequalification (SPQ) module, designed to adeptly handle the challenges posed by environments with repetitive textures, thereby enhancing loop detection and system robustness. To demonstrate the effectiveness and resilience of AVM-SLAM, we have released a specialized multi-sensor and high-resolution dataset of an underground garage, accessible at https://yale-cv.github.io/avm-slamdataset, encouraging further exploration and validation of our approach within similar settings.

High-resolution seamless remote sensing (HRSRS) images are essential foundational data for natural resource monitoring and land-use assessment, etc. However, high spatial resolution (HR) Earth observation from satellites typically has a long revisit period, and optical images can be widely obscured by clouds. Furthermore, the geometry and radiation issues make the generation of HRSRS images a challenging task. In this letter, to systematically address these issues, we propose a robust and efficient framework designed to generate HRSRS images for regional-scale areas, integrating various mature image processing technologies and jointing super-resolution (SR) reconstruction and thick cloud removal for the first time. In particular, the frame-work can realize adaptive reconstruction of lost spatial information on demand based on the satellite observation coverage and thick cloud coverage information. The effectiveness and reliability of the framework was demonstrated by its successful application in generating a 1-m resolution quarterly seamless image of the city of Wuhan, Hubei province, China, using 34 images acquired by the Chinese Gaofen (GF)-1/2/6/7 satellites. The experimental results show that both the intermediate results and final results achieve a satisfactory visual and quantitative effect. For example, SR reconstruction improves the spatial resolution of low-resolution images from 2-m to 1-m, thus increasing the spatial frequency and entropy by about 0.62 and 0.12, respectively.

Long-tailed learning has attracted increasing attention in very recent years. Long-tailed multi-label image classification is one subtask and remains challenging and poorly researched. In this paper, we provide a fresh perspective from probability to tackle this problem. More specifically, we find that existing cost-sensitive learning methods for long-tailed multi-label classification will affect the predicted probability of positive and negative labels in varying degrees during training, and different processes of probability will affect the final performance in turn. We thus propose a probability guided loss which contains two components to control this process. One is the probability re-balancing which can flexibly adjust the process of training probability. And the other is the adaptive probability-aware focal which can further reduce the probability gap between positive and negative labels. We conduct extensive experiments on two long-tailed multi-label image classification datasets: VOC-LT and COCO-LT. The results demonstrate the rationality and superiority of our strategy.

Solar-induced chlorophyll fluorescence (SIF) is an intuitive and accurate way to measure vegetation photosynthesis. Orbiting Carbon Observatory-2 (OCO-2)-retrieved SIF has shown great potential in estimating terrestrial gross primary production (GPP), but the discontinuous spatial coverage limits its application. Although some researchers have reconstructed OCO-2 SIF data, few have considered the uneven spatial and temporal distribution of the swath-distributed data, which can induce large uncertainties. In this article, we propose a spatiotemporal constrained light gradient boosting machine model (ST-LGBM) to reconstruct a contiguous OCO-2 SIF product (eight days, 0.05°), considering the data distribution characteristics. Two spatial and temporal constraining factors are introduced to utilize the relationships between the swath-distributed OCO-2 samples, combining the geographical regularity and vegetation phenological characteristics. The results indicate that the ST-LGBM method can improve the reconstruction accuracy in the missing data areas ($R^{2}= 0.79$), with an increment of 0.05 in$R^{2}$. The declined accuracy of the traditional light gradient boosting machine (LightGBM) method in the missing data areas is well alleviated in our results. The real-data comparison with TROPOspheric Monitoring Instrument (TROPOMI) SIF observations also shows that the results of the ST-LGBM method can achieve a much better consistency, in both spatial distribution and temporal variation. The sensitivity analysis also shows that the ST-LGBM can support stable results when using various input combinations or different machine learning models. This approach represents an innovative way to reconstruct a more accurate globally continuous OCO-2 SIF product and also provides references to reconstruct other data with a similar distribution.