Article Preview
TopIntroduction
Visual media are used to provide a variety of information to individuals in the Internet era. Images have a significant role in the transmission of information in visual communication. High-quality photos can enhance the transmission effect and provide a more comprehensive and useful message. However, a number of externally undesirable elements cause the image quality to decrease during the formation and transmission of the image. Ensuring that the photographs are aesthetically pleasing requires the adoption of image processing techniques for image repair and enhancement.
Convolutional neural networks can significantly increase image identification rates allowing for more effective data mining of visual information. As artificial intelligence technology develops, people are becoming increasingly concerned about deep learning. The recognition rate is low, and the typical picture recognition technique is comparatively out-of-date. In light of the vast amount of image data available, it is clear that the classic recognition method cannot satisfy the current needs. This essay examines how image processing technology is used in the context of visual communication. The picture recognition algorithm is optimized using convolutional neural network theory.
Zheng et al. (2015) introduced a new form of convolutional neural network that combines the advantages of convolutional neural networks (CNNs) and conditional random field (CRF)-based probabilistic graphical modeling. In 2018, Adem presented a method combining a cyclic Hough transform and a convolutional neural network (CNN) algorithm for detecting diabetic retinopathy.
Xu et al. (2018) studied deep convolutional neural network based autonomous marine vehicle maneuvering. To improve it, an autonomous collision avoidance method based on vision technology as a human visual system was proposed (Xu et al., 2018). Zhao et al. (2018) proposed a method for input image data processing and autonomous motion estimation using convolutional neural networks. Gavali et al. (2019) study deep convolutional neural network for image classification on cuda platform. The image classification process is performed using LeNet network model (Gavali et al., 2019). Bosse et al. (2019) derive the concept of distortion sensitivity as a property of the reference image which compensates for the lack of perceptual relevance of a given computational quality model potential: Convolutional neural network (CNN) for image feature extraction combined with recurrent neural network (RNN) based on attentional mechanism for automatic lip-reading recognition (Bosse et al., 2019).
The use of unmanned motion platforms in the military and in daily life has gradually grown into a new research hotspot thanks to the ongoing advancements in artificial intelligence and computer science technologies. Additionally, it applies to more complex scenarios, such as those in the military, the aerospace industry, the medical field, and daily life (Lu & Li, 2019). Unmanned motion platforms must be capable of performing self-positioning, map construction, and path planning in the working environment to be able to operate autonomously. Examples of such platforms include unmanned vehicles, unmanned aerial vehicles, and mobile robots (Peng et al., 2020).
The inertial navigation system is the most popular autonomous navigation and positioning technique for unmanned motion platforms, but it has a dead reckoning offset error, and its position estimation error will keep growing over time, making it impossible to perform high-precision, long-duration autonomy. Relevant researchers concentrate on animals with autonomous navigation abilities to investigate higher-precision and more dependable navigation methods and means and improve the autonomy and intelligence of the navigation system. John O’Keefe, an American researcher, initially looked into the relationship between the hippocampus and spatial localization cognition in the animal brain in 1971 (Trnovszky et al., 2017), revealing place cells (location cells) for localization.