Conventional geographic object-based image analysis (GEOBIA) land cover classification methods by using very high resolution images are hardly applicable due to their complex ground truth and manually selected features, while convolutional neural networks (CNNs) with many hidden layers provide the possibility of extracting deep features from very high resolution images. Compared with pixel-based CNNs, superpixel-based CNN classification, carrying on the idea of GEOBIA, is more efficient. However, superpixel-based CNNs are still problematic in terms of their processing units and accuracies. Firstly, the limitations of salt and pepper errors and low boundary adherence caused by superpixel segmentation still exist; secondly, this method uses the central point of the superpixel as the classification benchmark in identifying the category of the superpixel, which does not allow classification accuracy to be ensured. To solve such problems, this paper proposes a region-based majority voting CNN which combines the idea of GEOBIA and the deep learning technique. Firstly, training data was manually labeled and trained; secondly, images were segmented under multiresolution and the segmented regions were taken as basic processing units; then, point voters were generated within each segmented region and the perceptive fields of points voters were put into the multi-scale CNN to determine their categories. Eventually, the final category of each region was determined in the region majority voting system. The experiments and analyses indicate the following: 1. region-based majority voting CNNs can fully utilize their exclusive nature to extract abstract deep features from images; 2. compared with the pixel-based CNN and superpixel-based CNN, the region-based majority voting CNN is not only efficient but also capable of keeping better segmentation accuracy and boundary fit; 3. to a certain extent, region-based majority voting CNNs reduce the impact of the scale effect upon large objects; and 4. multi-scales containing small scales are more applicable for very high resolution image classification than the single scale.

Traditional and convolutional neural network (CNN)-based geographic object-based image analysis (GeOBIA) land-cover classification methods prosper in remote sensing and generate numerous distinguished achievements. However, a bottleneck emerges and hinders further improvements in classification results, due to the insufficiency of information provided by very high-spatial resolution images (VHSRIs). To be specific, the phenomenon of different objects with similar spectrum and the lack of topographic information (heights) are natural drawbacks of VHSRIs. Thus, multisource data steps into people’s sight and shows a promising future. Firstly, for data fusion, this paper proposed a standard normalized digital surface model (StdnDSM) method which was actually a digital elevation model derived from a digital terrain model (DTM) and digital surface model (DSM) to break through the bottleneck by fusing VHSRI and cloud points. It smoothed and improved the fusion of point cloud and VHSRIs and thus performed well in follow-up classification. The fusion data then were utilized to perform multiresolution segmentation (MRS) and worked as training data for the CNN. Moreover, the grey-level co-occurrence matrix (GLCM) was introduced for a stratified MRS. Secondly, for data processing, the stratified MRS was more efficient than unstratified MRS, and its outcome result was theoretically more rational and explainable than traditional global segmentation. Eventually, classes of segmented polygons were determined by majority voting. Compared to pixel-based and traditional object-based classification methods, majority voting strategy has stronger robustness and avoids misclassifications caused by minor misclassified centre points. Experimental analysis results suggested that the proposed method was promising for object-based classification.