ABSTRACT Maturity has a significant impact on the quality and safety of baking foods. Current maturity detection baked foods faces problems insufficient capability for small objects, low accuracy, poor efficiency. This paper proposes real‐time classification method Baking Detection (BMD) by improving baseline (SSD). Firstly, dataset with 29,360 images is constructed using small‐sample enhancement strategies, including Generative Adversarial Network (GAN), Random Augmentation (RA), multiple...

Semi-supervised semantic segmentation relieves the reliance on large-scale labeled data by leveraging unlabeled data. Recent semi-supervised approaches mainly resort to pseudo-labeling methods exploit However, unreliable can undermine semi-supervision processes. In this paper, we propose an algorithm called Multi-Level Label Correction (MLLC), which aims use graph neural networks capture structural relationships in Semantic-Level Graphs (SLGs) and Class-Level (CLGs) rectify erroneous...

<title>Abstract</title> The defects in the metal coating surface of annular ceramic workpiece have significant effects on conductivity and reliability. Due to irregularity, small area, few sample number defects, it is difficult achieve efficient accurate inspection. This paper presents a defect inspection framework based deep learning for workpiece. Firstly, an image acquisition system designed, characteristics are analyzed. Then, data set constructed through five augmentation strategies...

One of the key problems in detecting metal surface defects is that lighting angles have great influence on defect features information image. A multi-angle multivariate image analysis approach was proposed to improve accuracy and reliability detection results. By adjusting height selectively, images with could be obtained used constitute data, where each channel representation different angles. It based Multivariate Image Analysis (MIA) technique extract information. The effective were...

This paper presents a design of planar snake robot which consists bottom-wheeled modular parts chained with servo actuator joints. Serpentine curve locomotion is recognized as the most efficient gait for robots control in terms motor torques and friction forces. However, traditional serpentine model based on fixed parameters. There slipping problem when change shape. Changing parameters to implement given manoeuvring motion causes slip. The improved equations are proposed here attaching new...

Highlights A spring buffering system was added to a longitudinal vibration picking mechanism for berry shrub fruits. Vibration models of the part and side branch were established obtain relationship. Blueberry experiments conducted verify performance mechanism. Abstract . Current mechanisms are generally rigid structures, which can easily cause jam produce impacts, damaging branches. In view this situation, vibratory fruits is proposed in study with provide effect. branch, analysis performed...

Abstract Active rehabilitation can use electro-encephalogram (EEG) signals to identify the patient's left and right leg movement intentions for training, which helps stroke patients recover better faster. However, lower limb robot based on EEG has low recognition accuracy so far. A classification method of motor imagery is proposed enable accurately control their legs. Firstly, aiming at unstable characteristics signals, an experimental protocol was constructed multijoint trajectory planning...

To realise end-to-end visual identification, positioning, and angle detection of cylindrical batteries, a rotated object recognition method based on YOLOv5 is proposed. Firstly, aiming at the problems battery appearance scale variation surface reflection, recursive gated convolution feature fusion module was added to neck network enhance multi-scale extraction. Secondly, considering boundary problem range, circular smooth label introduced after prediction network, logistic regression...

During the packaging process, it is essential to detect steel stamping characters inside box identify any missing or repeated characters. Currently, manual detection suffers from low efficiency and a high false rate. To address these challenges, character recognition analysis system based on machine vision has been developed. The enhanced YOLOv7 method was employed for identification, complemented by statistical approach achieve automated judgment detection. issue of size disparity between...

To address the issue that sparse point cloud maps constructed by SLAM cannot provide detailed information about measured objects, and image sequence-based measurement methods have problems with large data volume cumulative errors, this paper proposes a scene method integrates sequences an improved LiDAR SLAM. By introducing plane features, positioning accuracy of is enhanced, real-time odometry poses are generated. Simultaneously, system captures object using synchronized cameras, NeRF used...

In view of the problem about uneven image acquisition and inaccurate edge extraction in pipeline detection process, a robot defect inspection method based on adaptive enhancement is proposed. Firstly, single-scale Retinex algorithm designed, which uses guided filter to estimate illumination component Value image, gets equilibrium by Gamma correction, so as realize enhancement. Then, traditional Canny improved, using bilateral filtering smooth image. Besides, images are segmented iterative...

To improve the motion diversity of planar end-traction rehabilitationrobot, a variable height rotating mechanism for end-tractionupper limb rehabilitation robot (ULRR) is proposed based onthe serial–parallel mechanism.

Aiming at the problem of complicated process and limited scope in visual inspection tunnel defects, a high resolution image mosaic method based on camera calibration for patrol robot system is proposed. Patroling specific track with module tunnel, designed mobile greatly simplifies defect detection process, saves lot manpower material resources by using cylindrical fast calibration. Through accurate panoramic obtained without scene constraints, which beneficial subsequent comprehensive...

A cylindrical image mosaic method based on fast camera calibration in multi-scene is proposed to solve the problems of scene limitation and complex process using parameter. Firstly, accurate corner feature checkerboard board used make it overlapping field view two adjacent images. Then, sequence pre-processed by extraction, precision matching, so that registration parameters between images be stitched can solved accurately quickly. After that, projection maintain visual consistency images,...