Object Recognition Algorithm Research Articles

Transmittance Surface Detection and Material Identification Using Multitask ViT-SIFT Fusion

The automatic detection of transmittance surfaces (transparency and translucency) and material class identification (glass, plastic, and so on) from an image is crucial for applications, including domestic service robotics and handling fragile instruments in experimental or industrial settings. However, this is a difficult task due to the lack of expressive textures in such materials. Although previous approaches for detecting transmittance surfaces propose handcrafted features, they disregard global material information and are limited to a single item (usually glass). Material or object recognition algorithms based on convolutional neural networks (CNNs) focus on optimizing performance that requires large labeled datasets. These algorithms are better for detecting opaque objects because large architectures tend to generalize the texture from an image’s background layer. Furthermore, creating large-scale off-the-shelf dataset labeled with transmittance surface is an extremely tedious task. Thus, we propose a simple yet effective and efficient learning model that combines a shallow multitask vision transformer (ViT) with a scale-invariant feature transform (SIFT) and a shallow CNN as the backbone network. SIFT and ViT are used to capture local and global discriminative features, respectively, and their fusion can increase the performance on a small dataset. The backbone network is utilized to introduce the desirable properties of CNN. Using limited training data, this model detects transmittance surfaces and determines their material type. We conducted tests on a new benchmark dataset containing a variety of transmissive materials and three publicly available datasets for robust comparison. Results show that the proposed model achieved a new state-of-the-art accuracy of 81.78% and 77.21% on our new dataset for transmittance surface recognition and material class identification, respectively.

Testing the detection of large, secretive snakes using camera traps

AbstractNovel technologies, such as camera traps, have expanded the opportunities for species detection, especially for rare species. Corresponding changes in data processing must occur to handle the large volume of data gathered from technology like camera traps. Automated image data processing, usually by running images through different types of computer algorithms, is an overarching goal to reduce the number of images that researchers must manually review. However, differences in camera trap setups and species characteristics can make automatic processing a challenge. Here, we evaluated the detection accuracy and efficiency of a time‐lapse triggered camera trapping technique combined with a pixel change detection algorithm as part of a monitoring program for a translocated population of the rare and federally threatened Louisiana Pinesnake (Pituophis ruthveni). We paired 5 cameras with automated pit tag readers to collect observations of P. ruthveni. We evaluated an image dataset of 1,500,187 images, collected over 7 months, both manually (i.e., researchers looking at each individual picture to determine snake presence) and automatically using a change detection algorithm. There were 18 P. ruthveni observations recorded by the tag readers, 7 of which occurred while a paired camera was not operational. Ten of the tag reader P. ruthveni observations were captured by the paired camera trap, with an additional P. ruthveni observation from a paired camera trap not recorded by the tag reader. There were 132 snake observations of 13 additional species and 18 observations of unknown snakes from the camera traps. The algorithm reduced the number of images reviewers evaluated by an average of 78.5% per camera (range = 37.3–98.7%) but had a 54.5% success rate at detecting observations of P. ruthveni (47.1% for individual images), and a slightly lower 48.9% success rate detecting other large snakes. Large snakes were 4 times more likely to be flagged by the algorithm than small snakes. Our time‐lapse triggered camera trapping technique performed well with respect to P. ruthveni detection accuracy, compared to the tag readers. However, further research is needed to improve quality assurances of camera trap image filtering and object recognition algorithms across different sites or environments.

Sonar Image Target Detection Based on Style Transfer Learning and Random Shape of Noise under Zero Shot Target

With the development of sonar technology, sonar images have been widely used to detect targets. However, there are many challenges for sonar images in terms of object detection. For example, the detectable targets in the sonar data are more sparse than those in optical images, the real underwater scanning experiment is complicated, and the sonar image styles produced by different types of sonar equipment due to their different characteristics are inconsistent, which makes it difficult to use them for sonar object detection and recognition algorithms. In order to solve these problems, we propose a novel sonar image object-detection method based on style learning and random noise with various shapes. Sonar style target sample images are generated through style transfer, which enhances insufficient sonar objects image. By introducing various noise shapes, which included points, lines, and rectangles, the problems of mud and sand obstruction and a mutilated target in the real environment are solved, and the single poses of the sonar image target is improved by fusing multiple poses of optical image target. In the meantime, a method of feature enhancement is proposed to solve the issue of missing key features when using style transfer on optical images directly. The experimental results show that our method achieves better precision.

Brain inspired face recognition: A computational framework

This paper presents a new proposal of an efficient computational model of face recognition which uses cues from the distributed face recognition mechanism of the brain, and by gathering engineering equivalent of these cues from existing literature. Three distinct and widely used features – Histogram of Oriented Gradients (HOG), Local Binary Patterns (LBP), and Principal components (PCs) extracted from target images are used in a manner which is simple, and yet effective. The HOG and LBP features further undergo Principal Component Analysis (PCA) for dimensionality reduction. Our model uses multi-layer perceptrons (MLP) to classify these three features and fuse them to form a sparsely connected model. A computational theory is first developed by using concepts from the information processing mechanism of the brain. Extensive experiments are carried out using eight publicly available face datasets to validate our proposed model’s performance in recognizing faces with extreme variation of illumination, pose angle, expression, and background. We also investigate the same mechanism because of reasons discussed later, on object recognition tasks as well. Results obtained are extremely promising when compared with other face and object recognition algorithms including CNN and deep learning-based methods. This highlights that simple computational processes, if clubbed properly, can produce competing performance with best algorithms.

Efficient Object Detection Based on Masking Semantic Segmentation Region for Lightweight Embedded Processors.

Because of the development of image processing using cameras and the subsequent development of artificial intelligence technology, various fields have begun to develop. However, it is difficult to implement an image processing algorithm that requires a lot of calculations on a light board. This paper proposes a method using real-time deep learning object recognition algorithms in lightweight embedded boards. We have developed an algorithm suitable for lightweight embedded boards by appropriately using two deep neural network architectures. The first architecture requires small computational volumes, although it provides low accuracy. The second architecture uses large computational volumes and provides high accuracy. The area is determined using the first architecture, which processes semantic segmentation with relatively little computation. After masking the area using the more accurate deep learning architecture, object detection is implemented with improved accuracy, as the image is filtered by segmentation and the cases that have not been recognized by various variables, such as differentiation from the background, are excluded. OpenCV (Open source Computer Vision) is used to process input images in Python, and images are processed using an efficient neural network (ENet) and You Only Look Once (YOLO). By running this algorithm, the average error can be reduced by approximately 2.4 times, allowing for more accurate object detection. In addition, object recognition can be performed in real time for lightweight embedded boards, as a rate of about 4 FPS (frames per second) is achieved.

B-YOLOX-S: A Lightweight Method for Underwater Object Detection Based on Data Augmentation and Multiscale Feature Fusion

With the increasing maturity of underwater agents-related technologies, underwater object recognition algorithms based on underwater robots have become a current hotspot for academic and applied research. However, the existing underwater imaging conditions are poor, the images are blurry, and the underwater robot visual jitter and other factors lead to lower recognition precision and inaccurate positioning in underwater target detection. A YOLOX-based underwater object detection model, B-YOLOX-S, is proposed to detect marine organisms such as echinus, holothurians, starfish, and scallops. First, Poisson fusion is used for data amplification at the input to balance the number of detected targets. Then, wavelet transform is used to perform Style Transfer on the enhanced images to achieve image restoration. The clarity of the images and detection targets is further increased and the generalization of the model is enhanced. Second, a combination of BIFPN-S and FPN is proposed to fuse the effective feature layer obtained by the Backbone layer to enhance the detection precision and accelerate model detection. Finally, the localization loss function of the prediction layer in the network is replaced by EIoU_Loss to heighten the localization precision in detection. Experimental results comparing the B-YOLOX-S algorithm model with mainstream algorithms such as FasterRCNN, YOLOV3, YOLOV4, YOLOV5, and YOLOX on the URPC2020 dataset show that the detection precision and detection speed of the algorithm model have obvious advantages over other algorithm networks. The average detection accuracy mAP value is 82.69%, which is 5.05% higher than the benchmark model (YOLOX-s), and the recall rate is 8.03% higher. Thus, the validity of the algorithmic model proposed in this paper is demonstrated.

A Robust Framework for Object Detection in a Traffic Surveillance System

Object recognition is the technique of specifying the location of various objects in images or videos. There exist numerous algorithms for the recognition of objects such as R-CNN, Fast R-CNN, Faster R-CNN, HOG, R-FCN, SSD, SSP-net, SVM, CNN, YOLO, etc., based on the techniques of machine learning and deep learning. Although these models have been employed for various types of object detection applications, however, tiny object detection faces the challenge of low precision. It is essential to develop a lightweight and robust model for object detection that can detect tiny objects with high precision. In this study, we suggest an enhanced YOLOv2 (You Only Look Once version 2) algorithm for object detection, i.e., vehicle detection and recognition in surveillance videos. We modified the base network of the YOLOv2 by reducing the number of parameters and replacing it with DenseNet. We employed the DenseNet-201 technique for feature extraction in our improved model that extracts the most representative features from the images. Moreover, our proposed model is more compact due to the dense architecture of the base network. We utilized DenseNet-201 as a base network due to the direct connection among all layers, which helps to extract a valuable information from the very first layer and pass it to the final layer. The dataset gathered from the Kaggle and KITTI was used for the training of the proposed model, and we cross-validated the performance using MS COCO and Pascal VOC datasets. To assess the efficacy of the proposed model, we utilized extensive experimentation, which demonstrates that our algorithm beats existing vehicle detection approaches, with an average precision of 97.51%.

A Technique to Navigate Autonomous Underwater Vehicles Using a Virtual Coordinate Reference Network during Inspection of Industrial Subsea Structures

Industrial subsea infrastructure inspections using autonomous underwater vehicles (AUV) require high accuracy of AUV navigation relative to the objects being examined. In addition to traditional navigation tools with inertial navigation systems and acoustic navigation equipment, technologies with video information processing are also actively developed today. The visual odometry-based techniques can provide higher navigation accuracy for local maneuvering at short distances to objects. However, in the case of long-distance AUV movements, such techniques typically accumulate errors when calculating the AUV movement trajectory. In this regard, the present article considers a navigation technique that allows for increasing the accuracy of AUV movements in the coordinate space of the object inspected by using a virtual coordinate reference network. Another aspect of the method proposed is to minimize computational costs for AUV moving along the inspection trajectory by referencing the AUV coordinates to the object pre-calculated using the object recognition algorithm. Thus, the use of a network of virtual points for referencing the AUV to subsea objects is aimed to maintain the required accuracy of AUV coordination during a long-distance movement along the inspection trajectory, while minimizing computational costs.

Resource Efficient Edge Computing Infrastructure for Video Surveillance

The emerging edge computing applications often use high definition cameras as edge devices to capture video streams that need to be analyzed in real-time for situational understanding and answering queries. However, such devices suffer from limited energy (and hence limited computing power) and limited bandwidth available to stream the data to the edge controllers that provide much higher computing capacities. In this paper, we address these issues in the context of vehicular traffic monitoring and develop a scheme that has two components: YLLO and BATS. YLLO is a lightweight object recognition algorithm that runs on the edge device itself and substantially reduces the frame rate sent to the edge controller without removing the important information. BATS adapts the transmissions to the available bandwidth by taking advantage of further redundancy in the video stream in both single and multi-camera scenarios. We show that these mechanisms together can maintain object identification accuracy of above 95 percent, while transmitting just <inline-formula><tex-math notation="LaTeX">$\sim$</tex-math></inline-formula> 5–10 percent of all the frames recorded by the cameras.

Stylized Pairing for Robust Adversarial Defense

Recent studies show that deep neural networks (DNNs)-based object recognition algorithms overly rely on object textures rather than global object shapes, and DNNs are also vulnerable to human-less perceptible adversarial perturbations. Based on these two phenomenons, we conjecture that the preference of DNNs on exploiting object textures for decisions is one of the most important reasons for the existence of adversarial examples. At present, most adversarial defense methods are directly related to adversarial perturbations. In this paper, we propose an adversarial defense method independent of adversarial perturbations, which utilizes a stylized pairing technique to encourage logits for a pair of images and the corresponding stylized image to be similar. With stylized pairing training, DNNs can better learn shape-biased representation. We have empirically evaluated the performance of our method through extensive experiments on CIFAR10, CIFAR100, and ImageNet datasets. Results show that the models with stylized pairing training can significantly improve their performance against adversarial examples.

Warping of Radar Data Into Camera Image for Cross-Modal Supervision in Automotive Applications

We present an approach to automatically generate semantic labels for real recordings of automotive range-Doppler (RD) radar spectra. Such labels are required when training a neural network for object recognition from radar data. The automatic labeling approach rests on the simultaneous recording of camera and lidar data in addition to the radar spectrum. By warping radar spectra into the camera image, state-of-the-art object recognition algorithms can be applied to label relevant objects, such as cars, in the camera image. The warping operation is designed to be fully differentiable, which allows backpropagating the gradient computed on the camera image through the warping operation to the neural network operating on the radar data. As the warping operation relies on accurate scene flow estimation, we further propose a novel scene flow estimation algorithm which exploits information from camera, lidar and radar sensors. The proposed scene flow estimation approach is compared against a state-of-the-art scene flow algorithm, and it outperforms it by approximately 30% w.r.t. mean average error. The feasibility of the overall framework for automatic label generation for RD spectra is verified by evaluating the performance of neural networks trained with the proposed framework for Direction-of-Arrival estimation.

A Bio-Inspired Hierarchical Spiking Neural Network With Reward-Modulated STDP Learning Rule for AER Object Recognition

The inherent spike-based and event-driven computation makes spiking neural networks (SNNs) naturally suitable to provide efficient and low-latency solution in neuromorphic vision processing. In this work we propose an address-event representation (AER) object recognition system which consists of a noise filter, an event stream segmentation module and a hierarchical convolutional SNN. In order to improve the feature learning capacity of bio-inspired shallow hierarchical SNNs for AER object recognition, we apply a convolutional SNN with R-STDP learning rule and first-spike coding to further extract features after primary feature learning using Gabor filter-based event-driven convolution. The R-STDP learning rule enables the network to adjust weights by combining local STDP learning rule with a global reward signal and to make classification without using an external classifier. Experimental results show that our method significantly outperforms the existing hierarchical SNN-based recognition methods on four popular AER datasets. Moreover, our method has much better recognition ability when using very short input event stream, and can effectively learn from small size training sets. In addition, the robustness of the network to input noise is improved. Our method is very beneficial for developing neuromorphic-vision object recognition algorithm in resource-constrained applications when event streams are short and training samples are limited.

Adaptive RANSAC and extended region-growing algorithm for object recognition over remote-sensing images

Adaptive RANSAC and extended region-growing algorithm for object recognition over remote-sensing images

A novel image segmentation utilizing FUZZY-based LBP and active contour model

ABSTRACT This work presents an object recognition algorithm that combines local binary pattern (LBP) based on fuzzy logic approach and active contour model for segmenting different images to detect textured objects. Initially, images containing objects are segmented using the fuzzy logic-optimized LBP method. Then, we eliminate the image noise. Finally, utilizing a Chan-Vese active contour method, the target object of the image is highlighted. The segmentation has been compared with the classical LBP technique and the results indicated higher accuracy and quality for highlighting the object from the background. The classification of highlighted objects is performed with a convolution neural network (CNN). To authenticate the proposed approach, 140 images with the classification of 10 different objects were used. The simulation depicted that the proposed method has better results than other methods both in terms of segmentation error and performance. Significantly, CNN classification also showed a classification accuracy of 92.8%.

A Novel Object Recognition Algorithm Based on Improved YOLOv5 Model for Patient Care Robots

The rapid development of computer vision raises a new research area involving patient care robots. Such robotic systems require fast target recognition at long ranges, where detecting smaller objects is notoriously challenging due to the cameras’ low resolution and noise. Spurred by these concerns, this paper develops a novel object recognition algorithm that solves these problems. Specifically, we amend YOLOv5 with our proposed sparse detection algorithm aiming to improve detection efficiency by separating the most significant context features and constructing smaller and less computational expensive models. Furthermore, we extend FReLU and suggest a novel activation function to improve recognition accuracy, which presents an extended nonlinearity increasing the expressiveness of the activation function. Finally, we propose sine annealing, which affords a trajectory that tends to cross over barriers and escape from local minima during training phase, addressing the challenging small object detection problem. The experimental results highlight that our algorithm has a lower memory consumption (Mem) value than the traditional YOLOv5 with a 5% boost down. Additionally, our method runs twice as fast as the traditional YOLOv5 while preserving accuracy, achieving more than 14.5 FPS on a medium-capability CPU. Overall, the detection results prove that our method can faster and accurately classifies and localizes most small-scaled objects.

Underwater occlusion object recognition with fusion of significant environmental features

Complicated underwater environments, such as occlusion by foreign objects and dim light, cause serious loss of underwater object features. This certainly increases the challenge of accurate detection of underwater objects. To tackle the above problems, we proposed an underwater occlusion object recognition algorithm combined with significant environmental features. First, the salient features of the underwater object were extracted by the salient feature extraction network. Second, salient environmental features associated with the object were extracted by the environmental feature attention mechanism. Finally, the contrast graph structure of the object’s salient features and related environmental features was constructed. The model proposed is compared with the existing object recognition model in the simulation of the three datasets: the underwater image enhancement benchmark dataset, video diver dataset, and underwater image dataset. The average recognition accuracy of the proposed model is 80.07%, which is improved by 2.38%. Experiments show that the object recognition algorithm proposed is effective and superior to existing algorithms.

A sensor combination based automatic sorting system for waste washing machine parts

In order to solve the difficulties encountered in E-wastes disposal automation, this work investigates the sensor-based sorting of waste washing machine parts. To satisfy the post-processing, we divided the parts into 6 categories (Piece, Cover, Base, Shell, Spin tub and Drain hose) based on the shape differences and used object recognition algorithm to classify them. In the recycling terminal, parts were divided based on types of materials. Piece and Cover contained acrylonitrile butadiene styrene (ABS) parts and polystyrene (PS) parts, while Base, Shell, Spin tub and Drain hose were polypropylene (PP) parts only. Therefore near-infrared (NIR) spectroscopy was applied for sorting ABS, PS and PP. Algorithm for object recognition and NIR spectroscopy both reached high precision. In object recognition, algorithm of YOLOv5 and its improved models, including focal loss and channel attention mechanism, were tested. YOLOv5 with channel attention achieved the best 98.7% mean Average Precision (mAP) and the overall over 98% Average Precision (AP). In the task of NIR spectroscopy, principal component analysis (PCA) coupled with support vector machine (SVM) reached the overall 97.8% classification precision. Finally, a sorting platform based on our device and algorithm was designed.

Attention-Enhanced One-Stage Algorithm for Traffic Sign Detection and Recognition

Detection and recognition of traffic signs are the keys to creating advanced driving assistance systems. Making highly precise maps also requires the identification and extraction of such road elements as traffic signs. Traditional detection and recognition methods no longer meet today’s needs, and object recognition algorithms based on deep learning have become the mainstream solution. However, current algorithms have limitations. The recognition speed of one-stage strategy algorithms is fast, but recognition accuracy is not satisfactory especially for small objects. The accuracy of two-stage algorithms is higher, but the recognition speed is extremely slow. This paper solves these problems with a proposed parallel attention convolution module, a channel attention pyramid network, and a loss function diagonal and center point IoU based on the YOLOv3 algorithm. The improved models in this paper are compared with SSD, YOLOv3, and Faster RCNN. Experimental results show that the proposed models have some improvement over the above models: the mAP of the models with PACM, CAFPN, and DCPIoU was 76.02%, compared with SSD300, SSD500, Faster RCNN, and YOLOv3, which had improvements of 9.27%, 6.93%, 2.94, and 5.3%, respectively. And the FPS of the improved model is basically the same as the original YOLOv3, without reducing the real-time performance.

Automatic Object Detection Algorithm-Based Braille Image Generation System for the Recognition of Real-Life Obstacles for Visually Impaired People.

The global prevalence of visual impairment due to diseases and accidents continues to increase. Visually impaired individuals rely on their auditory and tactile senses to recognize surrounding objects. However, accessible public facilities such as tactile pavements and tactile signs are installed only in limited areas globally, and visually impaired individuals use assistive devices such as canes or guide dogs, which have limitations. In particular, the visually impaired are not equipped to face unexpected situations by themselves while walking. Therefore, these situations are becoming a great threat to the safety of the visually impaired. To solve this problem, this study proposes a living assistance system, which integrates object recognition, object extraction, outline generation, and braille conversion algorithms, that is applicable both indoors and outdoors. The smart glasses guide objects in real photos, and the user can detect the shape of the object through a braille pad. Moreover, we built a database containing 100 objects on the basis of a survey to select objects frequently used by visually impaired people in real life to construct the system. A performance evaluation, consisting of accuracy and usefulness evaluations, was conducted to assess the system. The former involved comparing the tactile image generated on the basis of braille data with the expected tactile image, while the latter confirmed the object extraction accuracy and conversion rate on the basis of the images of real-life situations. As a result, the living assistance system proposed in this study was found to be efficient and useful with an average accuracy of 85% a detection accuracy of 90% and higher, and an average braille conversion time of 6.6 s. Ten visually impaired individuals used the assistance system and were satisfied with its performance. Participants preferred tactile graphics that contained only the outline of the objects, over tactile graphics containing the full texture details.

Recognition of dynamic objects from UGVs using Interconnected Neural network-based Computer Vision system

In this study, moving object recognition is performed by using images from a camera mounted on an unmanned ground vehicle. A GPS coordinate-based algorithm has been developed to obtain moving object silhouettes. In order to classify these silhouettes, an interconnected artificial neural network (ICANN) architecture consisting of two stages has been developed. The method consists of two phases. In the first phase, real-time images are converted to binary images at the end of the GPS-assisted image registration process. Then, the silhouettes are extracted from the background of the images using connected component labelling. In the second phase, two interconnected neural networks are used. The first neural network classifies silhouettes as objects or noise. The second neural network divides objects into seven subclasses as pedestrians, potholes, cars, etc. Compared to CNN-based techniques, a simpler NN architecture was employed in the research, and better accuracy rates were achieved with fewer samples. Another contribution of the research is simultaneous localization and mapping (SLAM) applications can be performed in non-GPS environments using pre-recorded images containing GPS information. In experimental studies, maximum success rates of 96,1% in object classification were obtained. The results obtained were compared to YOLO, the recently popular algorithm for object recognition.