Abstr act. With the development of information security, the traditional image encryption algorithm has been far from to ensuring the security of images in the transmission process. This paper presents a new image encryption algorithm, which can improve the security of image during transmission more effectively. The traditional scrambling algorithm based on Arnold transformation only applies to the square area, which is a big limitation. Focus on this, a multi-region algorithm for image scrambling encryption model is proposed, which splits the non-square image to multiple square regions, and scrambles each region. Experimental results show that the new algorithm improves the image security effectively to avoid deciphering, and it also can restore the image as same as the original image almost, which reaches to the purposes of image safe and reliable transmission. Intr oductionKeywords: Arnold transform. image encryption· image scrambling· multi-regional·scrambling degree.Digital image scrambling can make an image into a completely different meaningless image during transformation, and it is a preprocessing during hiding information of the digital image, which also known as information disguise. Image scrambling technology depends on data hiding technology which provides non-password security algorithm for information hiding. Data hiding technology led to a revolution in the warfare of network information, because it brought a series of new combat algorithms, and a lot of countries pay a lot of attentions on this area. Network information warfare is an important part of information warfare, and its core idea is to use public network for confidential data transmission. The image after scrambling encryption algorithms is chaotic, so attacker cannot decipher it.Some improved digital watermarking technology can apply scrambling method to change the distribution of the error bit in the image to improve the robustness of digital watermarking technology. Arnold scrambling algorithm has the feature of simplicity and periodicity, so it is used widely in the digital watermarking technology [1] (Arnold transform is proposed by V. I. Arnold in the research of ergodic theory, it is also called catmapping, and then it is applied to digital image). According to the periodicity of Arnold scrambling, the original image can be restored after several cycles. Because the periodicity of Arnold scrambling depends on the image size, it has to wait for a long time to restore an image. Generally, the cycle of Arnold transformation is not directly proportional to the image degree [2] . Currently, Arnold scrambling algorithm is base on square digital image in most literature, and these images are mostly N×N pixels of the digital image. However, most of the digital images are non-square in the real world, so that we cannot use Arnold scrambling algorithm widely [3] . To improve the Arnold scrambling algorithm, we will improve the original Arnold scrambling algorithm, so that we can apply Arnold scrambling algorithm to M×N non-square pixel d...

With the advantage of using only a limited number of samples, few-shot learning has been developed rapidly in recent years. It is mostly applied in the object classification or detection of a small number of samples which is typically less than ten. However, there is not much research related to few-shot detection, especially one-shot detection. In this paper, the multifeature information-assisted one-shot detection method is proposed to improve the accuracy of one-shot object detection. Specifically, two auxiliary modules are applied to the detection algorithm: Semantic Feature Module (SFM) and Detail Feature Module (DFM), which, respectively, extract semantic feature information and detailed feature information of samples in the support set. Then these two kinds of information are then calculated with the feature image extracted from the query image to obtain the corresponding auxiliary information that is used to complete one-shot detection. Thanks to the two auxiliary modules, which can retain more semantic and detailed information of samples in the support set, the proposed method can enhance the utilization rate of sample feature information and improve object detection accuracy by 2.97% compared to the benchmark method.

With the advancement of artificial intelligence (AI) and the upgrading of intelligent manufacturers, the development of intelligent manufacturing is now propelled by the replacement of inefficient traditional assembly machines and operators with machine vision (MV)-based industrial robots. The classic job recognition and positioning algorithm has multiple shortcomings, such as high complexity, manual design of similarity function, and susceptibility to noise disturbance. To solve these shortcomings, this study presents a fast job recognition and sorting method based on image processing. Firstly, the extraction approach for wavelet moment features and wavelet descriptors was introduced, and the feature fusion based on echo state network (ESN) was detailed. Then, the authors explained the idea of job template matching, and described how to measure similarity and terminate the measurement during template matching. Experimental results fully manifest the effectiveness of our strategy for fast job recognition and sorting. Our method offers a new solution to rapid recognition and sorting of objects in other fields.

No abstract

Video superresolution (VSR) aims to reconstruct a high-resolution video sequence from a low-resolution sequence. We propose a novel particle swarm optimization algorithm named as parameter-optimized multiple swarms PSO (POMS-PSO). We assessed the optimization performance of POMS-PSO by four standard benchmark functions. To reconstruct high-resolution video, we build an imaging degradation model. In view of optimization, VSR is converted to an optimization computation problem. And we take POMS-PSO as an optimization method to solve the VSR problem, which overcomes the poor effect, low accuracy, and large calculation cost in other VSR algorithms. The proposed VSR method does not require exact movement estimation and does not need the computation of movement vectors. In terms of peak signal-to-noise ratio (PSNR), sharpness, and entropy, the proposed VSR method based POMS-PSO showed better objective performance. Besides objective standard, experimental results also proved the proposed method could reconstruct high-resolution video sequence with better subjective quality.