At present, most scene text recognition methods achieve good performance by training models on many synthetic data. However, many data lead to huge storage space and large amount of calculation. And there is a gap between synthetic and real data. To solve these problems, we use a few real data to train a novel proposed model named spatial attention contrastive network (SAC-Net). The SAC-Net consists of a background suppression network (BSNet), a feature encoder, an attention decoder (ADEer), and a feature contrastive network (FCNet). The BSNet based on U-Net is used to reduce the interference of background. For relatively low prediction accuracy brought by connectionist temporal classification, we design an ADEer to improve performance by using convolutional attention mechanism. Based on data augmentation, we design a FCNet which belongs to contrastive learning. Finally, our SAC-Net is almost equivalent to the state-of-the-art model trained on a few real data for word accuracy on six benchmark test datasets.
In vehicle monitoring, recognizing graphic vehicle identification number (VIN) on the car frame is a particularly important step. While text recognition methods have made great progress, automatic graphic vehicle VIN recognition is still challenging. In VIN images, the VIN text is engraved on the car frame, with complex background and arbitrary orientation, which make it extremely difficult for recognition. We propose an efficient framework for recognizing rotational VIN. First, combining lightweight convolutional neural network and per-pixel segmentation in the output layer, we achieve fast and excellent VIN detection. Second, we take the VIN recognition task as a sequential position-dependent classification problem. By attaching sequential classifiers, we predict VIN text without character segmentation. Finally, we introduce a VIN dataset, which contains 2000 raw rotational VIN images and 90,000 horizontal VIN images for validating our framework. Experiments results show that the framework we proposed achieves good performance in VIN detection and recognition. By automatically identifying the VIN, we can quickly confirm the vehicle’s identity and help vehicle monitoring and tracking.
A novel method to SAR image segmentation based on fuzzy support vector machine (FSVM) for object recognition is proposed. First, the feature of river domain is extracted by analyzing bridge and its background in SAR image. Then, FSVM is used to make classification model by training example data for segmenting river region. Last, the summation minimum of direction energy is used as the rule for identifying a bridge in the binary image of river class. Experimental results show that the proposed method can effectively recognize more than one bridge over water in complex scene.
This paper handles with the problem of bridge recognition in Synthetic Aperture Radar (SAR) images. Based on features
analysis of bridges, rivers and land in different spatial resolution SAR images, a method of multi-scale analysis is
proposed. Firstly, for preventing noise, filtering the original medium resolution image is performed. And the image will
be down-sampled to ten times. Secondly, rivers will be extracted automatically by dynamic programming in low
resolution level and the bridge candidates' position will be obtained by apices locating. Finally, the mapped image will
be cut out to become regions of interest (ROI). And the bridge target regions will be detected by using constant false
alarm rate (CFAR). The example results indicate that the processing speed for bridge recognition can be greatly
improved and the precision of recognition can also be ensured.
Utilizing linear feature is now widely used in building detection. These linear feature-based methods are simple but low accuracy and time-consuming. This paper proposes a novel and efficient method of automatically detecting buildings based on multi-characteristic fusion from remote sensing images. The method firstly adopts Canny algorithm to detect edges lines from images. Then utilizing the feature of building distribution and the Hough transform, it employs ISODATA clustering algorithms to detect the main orientations of buildings. This clustering analysis method could filter edge lines and help to get latent edges of building objects. After that, the edges were linked to get the buildings' shape according to some linking rules. However there exit large amounts of false detection objects. In order to reduce them, a series of geometrical characteristics (such as the corner characteristic, the shadow characteristic, etc) and gray characteristic of buildings as criteria were brought up as the building judgments to eliminate them. We put forward the corresponding algorithm to extract each characteristic, later the fusion method based on the maximum membership principle in fuzzy pattern recognition was introduced to combine all these algorithm results together, and at last successfully detect buildings. The large number of experiment results show that this new method in this paper, compared with common linear feature-based building detection methods, is of high speed, more accurate and has good robustness. This new method is especially fit for practical applications in relatively complicated environments.
The change of grain size and structure in several metal films on different substrate by Ar+ bombarding has been shown. The experimental results show that the size of film grain will become bigger and even form single crystal under special bombarding conditions. It is also found that some factors of substrate, such as lattice structure, lattice parameter, and deposited surface, etc., have a great influence upon its growth. A new mechanism of crystal growth has been investigated.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.