A robust event-driven approach to always-on object recognition

<div> <div> <div> <p>We propose a neuromimetic architecture able to perform always-on pattern recognition. To achieve this, we extended an existing event-based algorithm [1], which introduced novel spatio-temporal features as a Hierarchy Of Time-Surfaces (HOTS). Built from asynchronous events acquired by a neuromorphic camera, these time surfaces allow to code the local dynamics of a visual scene and to create an efficient event-based pattern recognition architecture. Inspired by neuroscience, we extended this method to increase its performance. Our first contribution was to add a homeostatic gain control on the activity of neurons to improve the learning of spatio-temporal patterns [2]. A second contribution is to draw an analogy between the HOTS algorithm and Spiking Neural Networks (SNN). Following that analogy, our last contribution is to modify the classification layer and remodel the offline pattern categorization method previously used into an online and event-driven one. This classifier uses the spiking output of the network to define novel time surfaces and we then perform online classification with a neuromimetic implementation of a multinomial logistic regression. Not only do these improvements increase consistently the performances of the network, they also make this event-driven pattern recognition algorithm online and bio-realistic. Results were validated on different datasets: DVS barrel [3], Poker-DVS [4] and N-MNIST [5]. We foresee to develop the SNN version of the method and to extend this fully event-driven approach to more naturalistic tasks, notably for always-on, ultra-fast object categorization. </p> </div> </div> </div>

The Path of Film and Television Animation Creation Using Virtual Reality Technology under the Artificial Intelligence

The purpose is to provide a more reliable human-computer interaction (HCI) guarantee for animation works under virtual reality (VR) technology. Inspired by artificial intelligence (AI) technology and based on the convolutional neural network—support vector machine (CNN-SVM), the differences between animation works under VR technology and traditional animation works are analyzed through a comprehensive analysis of VR technology. The CNN-SVM gesture recognition algorithm using the error correction strategy is designed based on HCI recognition. To have better recognition performance, the advantages of depth image and color image are combined, and the collected information is preprocessed including the relations between the times of image training iterations and the accuracy of different methods in the direction of the test set. After experiments, the maximum accuracy of the preprocessed image can reach 0.86 showing the necessity of image preprocessing. The recognition accuracy of the optimized CNN-SVM is compared with other algorithm models. Experiments show that the accuracy of the optimized CNN-SVM has an upward trend compared with the previous CNN-SVM, and the accuracy reaches 0.97. It proves that the designed algorithm can provide good technical support for VR animation, so that VR animation works can interact well with the audience. It is of great significance for the development of VR animation and the improvement of people’s artistic life quality.

A robust event-driven approach to always-on object recognition

<div> <div> <div> <p>We propose a neuromimetic architecture able to perform always-on pattern recognition. To achieve this, we extended an existing event-based algorithm [1], which introduced novel spatio-temporal features as a Hierarchy Of Time-Surfaces (HOTS). Built from asynchronous events acquired by a neuromorphic camera, these time surfaces allow to code the local dynamics of a visual scene and to create an efficient event-based pattern recognition architecture. Inspired by neuroscience, we extended this method to increase its performance. Our first contribution was to add a homeostatic gain control on the activity of neurons to improve the learning of spatio-temporal patterns [2]. A second contribution is to draw an analogy between the HOTS algorithm and Spiking Neural Networks (SNN). Following that analogy, our last contribution is to modify the classification layer and remodel the offline pattern categorization method previously used into an online and event-driven one. This classifier uses the spiking output of the network to define novel time surfaces and we then perform online classification with a neuromimetic implementation of a multinomial logistic regression. Not only do these improvements increase consistently the performances of the network, they also make this event-driven pattern recognition algorithm online and bio-realistic. Results were validated on different datasets: DVS barrel [3], Poker-DVS [4] and N-MNIST [5]. We foresee to develop the SNN version of the method and to extend this fully event-driven approach to more naturalistic tasks, notably for always-on, ultra-fast object categorization. </p> </div> </div> </div>

Colour Recognition Algorithm Based on Colour Mapping Knowledge for Wooden Building Image

In the cross-media image reproduction technology, the accurate transfer and reproduction of colour between different media are an important issue in the reproduction process, and the colour mapping technology is the key technology to effectively maintain the image details and improve the level of colour reproduction. Wooden structure in the image colour and colour piece is different, the image of each colour of visual perception is not independent, and every colour in the image pixels is affected by the surrounding pixels, but in the process of image map, without thinking of the pixel space, adjacent pixels of mutual influence in particular, do not let a person particularly be satisfied with the resulting map figure. In the process of image processing by traditional colour mapping algorithm, the colour distortion caused by colour component is ignored and the block diagram of colour mapping system is constructed. With the continuous development of mapping recognition algorithms, the maximum and minimum brightness values in the image are mapped to the maximum and minimum brightness values of the display device by linear mapping algorithm according to the flow of the established recognition algorithm. By establishing the colour adjustment method of the colour mapping image, the processing effect of the mapping algorithm is analysed. The results show that the brightness deviation of the image is reduced and the colour resolution is improved by the colour brightness compensation.

THE EFFECT OF DIFFERENT OCCUPATIONAL BACKGROUND NOISES ON VOICE RECOGNITION ACCURACY

Voice recognition has become an integral part of our lives, commonly used in call centers and as part of virtual assistants. However, voice recognition is increasingly applied to more industrial uses. Each of these use cases has unique characteristics that may impact the effectiveness of voice recognition, which could impact industrial productivity, performance, or even safety. One of the most prominent among them is the unique background noises that are dominant in each industry. The existence of different machinery and different work layouts are primary contributors to this. Another important characteristic is the type of communication that is present in these settings. Daily communication often involves longer sentences uttered under relatively silent conditions, whereas communication in industrial settings is often short and conducted in loud conditions. In this study, we demonstrated the importance of taking these two elements into account by comparing the performances of two voice recognition algorithms under several background noise conditions: a regular Convolutional Neural Network (CNN) based voice recognition algorithm to an Auto Speech Recognition (ASR) based model with a denoising module. Our results indicate that there is a significant performance drop between the typical background noise use (white noise) and the rest of the background noises. Also, our custom ASR model with the denoising module outperformed the CNN based model with an overall performance increase between 14-35% across all background noises. . Both results give proof that specialized voice recognition algorithms need to be developed for these environments to reliably deploy them as control mechanisms.

A Deep Neural Network-Based Target Recognition Algorithm for Robot Scenes

Intelligent robots are a key vehicle for artificial intelligence and are widely employed in all aspects of everyday life and work, not just in the industry. One of the talents required for intelligent robots to complete their jobs is the capacity to identify their environment, which is a crucial obstacle to be overcome. Deep learning-based target identification algorithms currently do not fully leverage the link between high-level semantic and low-level detail information in the prediction step and hence are less successful in recognizing tiny target objects. Target recognition via vision sensors has also improved in accuracy and efficiency because of the development of deep learning. However, due to the insufficient usage of semantic information and precise texture information of underlying characteristics, tiny target recognition remains a difficulty. To address the aforementioned issues, we propose a target detection method based on a jump-connected pyramid model to improve the target detection performance of robots in complex scenarios. In order to verify the effectiveness of the algorithm, we designed and implemented a software system for target detection of intelligent robots and performed software integration of the proposed algorithm model with excellent experimental results. These experiments reveal that, when compared to other algorithms, our suggested algorithm’s characteristics have higher flexibility and robustness and can deliver a higher scene classification accuracy rate.

Digital Twins by Physical Education Teaching Practice in Visual Sensing Training System

The paper expects to improve the efficiency and intelligence of somatosensory recognition technology in the application of physical education teaching practice. Firstly, the combination of induction recognition technology and the Internet is used. Secondly, through the Kinect sensor, bone data are acquired. Finally, the hidden Markov model (HMM) is used to simulate the experimental data. On the simulation results, a gait recognition algorithm is proposed. The gait recognition algorithm is used to identify the motion behaviour, and the results are displayed in the Web (World Wide Web) end built by the cloud server. Meantime, in view of the existing problems in the practice of physical education, combined with the establishment and operation of the Digital Twins (DTs) system, the camera source recognition architecture is carried out since the twin network and the two network branches share weights. This paper analyses these problems since the application of somatosensory recognition technology and puts forward the improvement methods. For the single problem of equipment in physical education, this paper puts forward the monitoring and identification function of the cloud server. It is to transmit data through Hypertext Transfer Protocol (HTTP) and locate and collect data through a monitoring terminal. For the lack of comprehensiveness and balance of sports plans, this paper proposes a scientific training plan and process customization based on Body Mass Index (BMI), analyses real-time data in the cloud, and makes scientific customization plans according to different students’ physical conditions. Moreover, 25 participants are invited to carry out the exercise detection and analysis experiment, and the joint monitoring of their daily movements is tested. This process has completed the design of a feasible and accurate platform for information collection and processing, which is convenient for managers and educators to comprehensively and scientifically master and manage the physical level and training of college students. The proposed method improves the recognition rate of the camera source to some extent and has important exploration significance in the field of action recognition.

Design and Implementation of Image Edge Detection Algorithm on FPGA

Image processing is an important task in data processing systems for applications such as medical sectors, remote sensing, and microscopy tomography. Edge recognition is a sort of image division method that is used to simplify the image records so as to reduce the amount of data to be processed. Edges are considered the most important in image processing because they are used to characterize the boundaries of an image. The performance of the Canny edge recognition algorithm remarkably surpasses the present edge recognition technology in various computer visualization methods. The main drawback of using Canny edge boundary is that it consumes lot of period due to its complex computation. In order to tackle this problem a hybrid edge recognition method is proposed in block stage to locate edges with no loss. It employs the Sobel operator estimate method to calculate the value and direction of the gradient by substituting complex processes by hardware cost savings, traditional non-maximum suppression adaptive thresholding block organization, and conventional hysteresis thresholding. Pipeline was presented to lessen latency. The planned strategy is simulated using Xilinx ISE Design Suite14.2 running on a Xilinx Spartan-6 FPGA board. The synthesized architecture uses less hardware to detect edges and operates at maximum frequency of 935 MHz.

Strawberry Maturity Recognition Algorithm Combining Dark Channel Enhancement and YOLOv5

Aiming at the problems of low accuracy of strawberry fruit picking and large rate of mispicking or missed picking, YOLOv5 combined with dark channel enhancement is proposed. In “Fengxiang” strawberry, the criterion of “bad fruit” is added to the conventional three criteria of ripeness, near-ripeness, and immaturity, because some of the bad fruits are close to the color of ripe fruits, but the fruits are small and dry. The training accuracy of the four kinds of strawberries with different ripeness is above 85%, and the testing accuracy is above 90%. Then, to meet the demand of all-day picking and address the problem of low illumination of images collected at night, an enhancement algorithm is proposed to enhance the images, which are recognized. We compare the actual detection results of the five enhancement algorithms, i.e., histogram equalization, Laplace transform, gamma transform, logarithmic variation, and dark channel enhancement processing under the different numbers of fruits, periods, and video tests. The results show that combined with dark channel enhancement, YOLOv5 has the highest recognition rate. Finally, the experimental results demonstrate that YOLOv5 is better than SSD, DSSD, and EfficientDet in terms of recognition accuracy, and the correct rate can reach more than 90%. Meanwhile, the method has good robustness in complex environments such as partial occlusion and multiple fruits.