Aiming at the problems of inefficient detection caused by traditional manual inspection and unclear features in metal surface defect detection, an improved metal surface defect detection technology based on the You Only Look Once (YOLO) model is presented. The shallow features of the 11th layer in the Darknet-53 are combined with the deep features of the neural network to generate a new scale feature layer using the basis of the network structure of YOLOv3. Its goal is to extract more features of small defects. Furthermore, then, K-Means++ is used to reduce the sensitivity to the initial cluster center when analyzing the size information of the anchor box. The optimal anchor box is selected to make the positioning more accurate. The performance of the modified metal surface defect detection technology is compared with other detection methods on the Tianchi dataset. The results show that the average detection accuracy of the modified YOLO model is 75.1%, which ia higher than that of YOLOv3. Furthermore, it also has a great detection speed advantage, compared with faster region-based convolutional neural network (Faster R-CNN) and other detection algorithms. The improved YOLO model can make the highly accurate location information of the small defect target and has strong real-time performance.

Aiming at the problem of flexible sliding tactile sensing for the actual grasp of intelligent robot fingers, a double-layer sliding tactile sensor based on fiber Bragg grating (FBG) for robot fingers is proposed in this paper. Firstly, the optimal embedding depth range of FBG in the elastic matrix of polydimethylsiloxane (PDMS) was determined through finite element analysis and static detection experiments of finger tactile sensing. Secondly, the sensor structure is optimized and designed through the simulation and dynamic experiments of sliding sensing to determine the final array structure. Thirdly, the sensing array is actually pasted on the surface of the robot finger and the sensing characteristics testing platform is built to test and analyze the basic performance of the sliding tactile sensor. Then, the sensor array is actually attached to the finger surface of the robot and the sensing characteristics testing platform is built to experiment and analyze the basic performance of the sliding tactile sensor. Finally, a sliding tactile sensing experiment of robot finger grasping is conducted. The experimental results show that the sliding tactile sensor designed in this paper has good repeatability and creep resistance, with sensitivities of 12.4 pm/N, 11.6 pm/N, and 14.5 pm/N, respectively, and the overall deviation is controlled within 5 pm. Meanwhile, it can effectively sense the signals of the robot fingers during static contact and sliding. The sensor has a high degree of fit with the robot finger structure, and has certain application value for the perception of sliding tactile signals in the object grasping of intelligent robot objects.

Aiming at the problems of lateral force interference and non-uniform strain of robot fingers in the process of pressure tactile sensing, a flexible tactile sensor with a square hole structure based on fiber Bragg grating (FBG) is proposed in this paper. Firstly, the optimal embedding depth of the FBG in the sensor matrix model was determined by finite element simulation. Secondly, according to the size of the finger knuckle and the simulation analysis based on the pressure tactile sensor element for the robot finger, the square hole structure was designed, and the overall dimensions of the sensing element and size of the square hole were determined. Thirdly, the FBG was embedded in the polydimethylsiloxane (PDMS) elastic matrix to make a sensor model, and the tactile sensor was fabricated. Finally, the FBG pressure tactile sensing system platform was built by using optical fiber sensing technology, and the experiment of the FBG tactile sensor was completed through the sensing system platform. Experimental results show that the tactile sensor designed in this paper has good repeatability and creep resistance. The sensitivity is 8.85 pm/N, and the resolution is 0.2 N. The loading sensitivity based on the robot finger is 27.3 pm/N, the goodness of fit is 0.996, and the average value of interference in the sensing process is 7.63%, which is lower than the solid structure sensor. These results verify that the sensor can effectively reduce the lateral force interference and solve the problem of non-uniform strain and has high fit with fingers, which has a certain application value for the research of robot pressure tactile intelligent perception.

Object detection of overhead transmission lines is a solution for promoting inspection efficiency for power companies. However, aerial images contain many complex backgrounds and small objects, and traditional algorithms are incompetent in the identification of details of power transmission lines accurately. To address this problem, this paper develops an object detection method based on optimized You Only Look Once v5-small (YOLOv5s). This method is designed to be engineering-friendly, with the objective of maximal detection accuracy and computation simplicity. Firstly, to improve the detecting accuracy of small objects, a larger scale detection layer and jump connections are added to the network. Secondly, a self-attention mechanism is adopted to merge the feature relationships between spatial and channel dimensions, which could suppress the interference of complex backgrounds and boost the salience of objects. In addition, a small object enhanced Complete Intersection over Union (CIoU) is put forward as the loss function of the bounding box regression. This loss function could increase the derived loss for small objects automatically, thereby improving the detection of small objects. Furthermore, based on the scaling factors of batch-normalization layers, a pruning method is adopted to reduce the parameters and achieve a lightweight method. Finally, case studies are fulfilled by comparing the proposed method with classic YOLOv5s, which demonstrate that the detection accuracy is increased by 4%, the model size is reduced by 58%, and the detection speed is raised by 3.3%.

In the process of steel production, the defects on the surface of steel will adversely affect the subsequent processing of a product. Accurate detection of such defects is the key to improve production efficiency and economic benefits. In this paper, an end-to-end steel surface defect detection and size measurement system based on the YOLOv5 model is designed. Firstly, in consideration of the defect location and direction correlation in the production process, a coordinate attention mechanism is added at the head of YOLOv5 to strengthen the spatial correlation of the steel surface and an adaptive anchor box generation method based on defect shape difference feature is proposed, which realizes the detection of three main types of defects on the Pytorch deep learning framework. Secondly, BiFPN is used to strengthen the feature fusion and a transformer encoder is added to improve the performance of detecting small defects. Thirdly, calculate the conversion ratio between the pixel and the actual size according to the standard reference specimen and obtain the actual size through the pixel statistics of the defect area to achieve pixel level size measurement. Finally, the steel surface defect detection and size measurement system are designed in this paper, which consist of various hardware, related measurement, and detection algorithms. According to the experimental results, the comprehensive defect detection accuracy of this method reaches 93.6%, of which the scratch detection accuracy reaches 95.7%. The detection speed reaches 133 fps and the defect size measurement accuracy reaches 0.5 mm. Experimental result shows that the defect detection and size measurement system designed in this paper can accurately detect and measure various industrial production defects and can be applied to the actual production process.

In asphalt pavement detection, the defect scale changes greatly, mainly including mesh cracks, patches, and potholes. In the case of large scale, the texture feature is not clear, and the information is easily lost in the feature extraction process. Correspondingly, the number of small-scale holes is often very large, which also puts forward higher requirements for the detection model. In view of the above problems, this paper proposed a model for common asphalt pavement defects detection called YOLOv5-PD. In order to reduce the loss of information and expand the receptive field of the model, Big Kernel convolution was used to replace a part of the convolution in the original CSPDarknet. The texture feature information of the cracks is retained to the greatest extent. In order to enhance the detection performance of small defects, convolution channel attention mechanism was added after each feature fusion layer, and performs attention processing on the feature map after concat to find the defect location. This study used a public pavement defect dataset from Brazil. In this work, ablation experiments were carried out according to the task scenario, and the improved effects were compared and analyzed. The proposed model is compared with other versions of models and advanced models, which proves the superiority of the proposed model. The mAP of proposed model reached 73.3% and the model inference speed reached 41FPS, which can meet real time engineering application requirements.