Automatic building segmentation from aerial imagery is an important and challenging task because of the variety of backgrounds, building textures and imaging conditions. Currently, research using variant types of fully convolutional networks (FCNs) has largely improved the performance of this task. However, pursuing more accurate segmentation results is still critical for further applications such as automatic mapping. In this study, a multi-constraint fully convolutional network (MC–FCN) model is proposed to perform end-to-end building segmentation. Our MC–FCN model consists of a bottom-up/top-down fully convolutional architecture and multi-constraints that are computed between the binary cross entropy of prediction and the corresponding ground truth. Since more constraints are applied to optimize the parameters of the intermediate layers, the multi-scale feature representation of the model is further enhanced, and hence higher performance can be achieved. The experiments on a very-high-resolution aerial image dataset covering 18 km ${}^2$ and more than 17,000 buildings indicate that our method performs well in the building segmentation task. The proposed MC–FCN method significantly outperforms the classic FCN method and the adaptive boosting method using features extracted by the histogram of oriented gradients. Compared with the state-of-the-art U–Net model, MC–FCN gains 3.2% (0.833 vs. 0.807) and 2.2% (0.893 vs. 0.874) relative improvements of Jaccard index and kappa coefficient with the cost of only 1.8% increment of the model-training time. In addition, the sensitivity analysis demonstrates that constraints at different positions have inconsistent impact on the performance of the MC–FCN.

Multi-source remote sensing imagery has become widely accessible owing to the development of data acquisition systems. In this paper, we address the challenging task of the semantic segmentation of buildings via multi-source remote sensing imagery with different spatial resolutions. Unlike previous works that mainly focused on optimizing the segmentation model, which did not enable the severe problems caused by the unaligned resolution between the training and testing data to be fundamentally solved, we propose to integrate SR techniques with the existing framework to enhance the segmentation performance. The feasibility of the proposed method was evaluated by utilizing representative multi-source study materials: high-resolution (HR) aerial and low-resolution (LR) panchromatic satellite imagery as the training and testing data, respectively. Instead of directly conducting building segmentation from the LR imagery by using the model trained using the HR imagery, the deep learning-based super-resolution (SR) model was first adopted to super-resolved LR imagery into SR space, which could mitigate the influence of the difference in resolution between the training and testing data. The experimental results obtained from the test area in Tokyo, Japan, demonstrate that the proposed SR-integrated method significantly outperforms that without SR, improving the Jaccard index and kappa by approximately 19.01% and 19.10%, respectively. The results confirmed that the proposed method is a viable tool for building semantic segmentation, especially when the resolution is unaligned.

The automatic digitizing of paper maps is a significant and challenging task for both academia and industry. As an important procedure of map digitizing, the semantic segmentation section is mainly relied on manual visual interpretation with low efficiency. In this study, we select urban planning maps as a representative sample and investigate the feasibility of utilizing U-shape fully convolutional based architecture to perform end-to-end map semantic segmentation. The experimental results obtained from the test area in Shibuya district, Tokyo, demonstrate that our proposed method could achieve a very high Jaccard similarity coefficient of 93.63% and an overall accuracy of 99.36%. For implementation on GPGPU and cuDNN, the required processing time for the whole Shibuya district can be less than three minutes. The results indicate the proposed method can serve as a viable tool for urban planning map semantic segmentation task with high accuracy and efficiency.

When mosaicking orthoimages, especially in urban areas with various obvious ground objects like buildings, roads, cars or trees, the detection of optimal seamlines is one of the key technologies for creating seamless and pleasant image mosaics. In this paper, we propose a new approach to detect optimal seamlines for orthoimage mosaicking with the use of deep convolutional neural network (CNN) and graph cuts. Deep CNNs have been widely used in many fields of computer vision and photogrammetry in recent years, and graph cuts is one of the most widely used energy optimization frameworks. We first propose a deep CNN for land cover semantic segmentation in overlap regions between two adjacent images. Then, the energy cost of each pixel in the overlap regions is defined based on the classification probabilities of belonging to each of the specified classes. To find the optimal seamlines globally, we fuse the CNN-classified energy costs of all pixels into the graph cuts energy minimization framework. The main advantage of our proposed method is that the pixel similarity energy costs between two images are defined using the classification results of the CNN based semantic segmentation instead of using the image informations of color, gradient or texture as traditional methods do. Another advantage of our proposed method is that the semantic informations are fully used to guide the process of optimal seamline detection, which is more reasonable than only using the hand designed features defined to represent the image differences. Finally, the experimental results on several groups of challenging orthoimages show that the proposed method is capable of finding high-quality seamlines among urban and non-urban orthoimages, and outperforms the state-of-the-art algorithms and the commercial software based on the visual comparison, statistical evaluation and quantitative evaluation based on the structural similarity (SSIM) index.

Accurate human path forecasting in complex and crowded scenarios is critical for collision avoidance of autonomous driving and social robots navigation. It still remains as a challenging problem because of dynamic human interaction and intrinsic multimodality of human motion. Given the observation, there is a rich set of plausible ways for an agent to walk through the circumstance. To address those issues, we propose a spatio-temporal model that can aggregate the information from socially interacting agents and capture the multimodality of the motion patterns. We use mixture density functions to describe the human path and predict the distribution of future paths with explicit density. To integrate more factors to model interacting people, we further introduce a coordinate transformation to represent the relative motion between people. Extensive experiments over several trajectory prediction benchmarks demonstrate that our method is able to forecast various plausible futures in complex scenarios and achieves state-of-the-art performance.