Object detection in road scenes is a task that has recently become popular and it is also an important part of intelligent transportation systems. Due to the different locations of cameras in the road scenes, the size of the traffic objects captured varies greatly, which imposes a burden on the network optimization. In addition, in some dense traffic scenes, the size of the traffic objects captured is extremely small and it is easy to miss detection and to encounter false detection. In this paper, we propose an improved multi-scale YOLOv5s algorithm based on the YOLOv5s algorithm. In detail, we add a detection head for extremely small objects to the original YOLOv5s model, which significantly improves the accuracy in detecting extremely small traffic objects. A content-aware reassembly of features (CARAFE) module is introduced in the feature fusion part to enhance the feature fusion. A new SPD-Conv CNN Module is introduced instead of the original convolutional structure to enhance the overall computational efficiency of the model. Finally, the normalization-based attention module (NAM) is introduced, allowing the model to focus on more useful information during training and significantly improving detection accuracy. The experimental results demonstrate that compared with the original YOLOv5s algorithm, the detection accuracy of the multi-scale YOLOv5s model proposed in this paper is improved by 7.1% on the constructed diverse traffic scene datasets. The improved multi-scale YOLOv5s algorithm also maintains the highest detection accuracy among the current mainstream object detection algorithms and is superior in accomplishing the task of detecting traffic objects in complex road scenes.

No abstract

No abstract

In the automatic online vision-based inspector on pharmaceutical production line, vibration disturbance from electromechanical system mostly leads to failure using traditional inter-frame difference algorithm. In this paper, a real-time vibration-insensitive inter-frame difference inspection algorithm for online impurities detection is proposed to overcome the disturbance. The inspection algorithm employs an efficient subpixel image registration method (based on a mountain-climbing searching strategy) and adaptive local threshold segmentation. Inspection can be finished in no more than 0.93 second. Experimental inspection results of false alarm rate 4.51% and omission rate 0.14% show detection validity even if vibration amplitude increases to five pixels. Then impurity classification based on Relevance Vector Machine (RVM) is carried out as feedback to monitor and improve the manufacture process. Classification accuracy above 97% satisfies the requirements for automatic production line. Keywords-vision-based inspector; subpixel image registration; image segmentation; relevance vector machineI.

No abstract