An intelligent vision recognition method based on deep learning for pointer meters

Due to the danger of explosive, oversize and poison-induced abnormal waste and the complex conditions in Waste-to-Energy Power Plants (WtEPPs), the manual inspection and existing waste detection algorithms are incapable to meet the requirement of both high accuracy and efficiency. To address the issues, we propose the Waste-YOLO framework by introducing the coordinate attention, convolutional block attention module, content-aware reassembly of features, improved bidirectional feature pyramid network and SIoU loss function based on YOLOv5s for high accuracy real-time abnormal waste detection. Through video acquisition, frame-splitting, manual annotation and data augmentation, we develop an abnormal waste image dataset with the four most common types (i.e. gas cans, mattresses, wood and iron sheets) to evaluate the proposed Waste-YOLO. Extensive experimental results demonstrate the superiority of Waste-YOLO to several state-of-the-art algorithms in waste detection effectiveness and efficiency to ensure production safety in WtEPPs.

Section: Yolov5s Networkmentioning

confidence: 99%

“…However, these methods are characterized by a complex process, low efficiency and accuracy. The superior performance of deep learning has led to its widespread adoption in computer vision [8,9]. Convolutional neural networks (CNNs) have been widely applied for waste detection.…”

Section: Introductionmentioning

confidence: 99%

Waste-YOLO: towards high accuracy real-time abnormal waste detection in waste-to-energy power plant for production safety

Wang,

Chen

et al. 2023

“…Sun et al [39] utilized YOLOv4 [40] to locate the meter, adopted Anam-Net [41] for semantic segmentation to extract the pointer, employed CRAFT [42] and E2E-MLT [43] for recognizing scale values and units, applied polar coordinate transformation (PCT) for scale regions, and deploy a lightweight CNN to locate the main scale line, followed by the computation of the reading data. An intelligent vision recognition method [44] based on YOLOv5 and improved U 2 -Net [45] network is proposed to improve the accuracy and efficiency of meter recognition in a complex environment. Liu et al [46] proposed a multitask cascading convolutional neural network (MC-CNN) for pointer meter automatic recognition in outdoor environments, and an adaptive attention residual module is proposed for reading meters from cropped images.…”

Section: Deep Learning Approachesmentioning

confidence: 99%

“…BBPDF enables estimation of pointer direction based on the relationship between the location of the pointer region and pointer center, in cases where fitting the pointer direction using Since other meter reading algorithms require prior reading information to perform readings, we provide it to them in advance for comparison. Method 1 [44] transforms the arc-shaped scale into a linear scale through polar coordinate transformation (PCT) according to the obtained pointers and scale masks, and then calculates the reading on the basis of the position of the pointer on the scale. Method 2 [6] corrects the dial to a standard position through perspective transformation and calculates the reading based on the linear fit of the pointer using prior reading information.…”

Section: Ablation Study Of Reading Algorithmmentioning

confidence: 99%

Scale value guided Lite-FCOS for pointer meter reading recognition

Wang,

Tian,

et al. 2023

As intelligent power grid construction advances, substation inspection becomes crucial, particularly in identifying meter readings.Existing meter reading methods are mainly based on the relationship between pointer and scale. However, these methods commonly suffer from the issue of over-relying on prior reading information, limiting readings to known meters.Hereby in this paper, we propose a method reaching more accurate and robust performance for meter reading by utilizing the unnoticed scale value.We determine the meter pointer direction with the aid of Hough transform and the pointer distribution.To detect meters from the scene as well as obtain scale values and pointers from the meter dial, we build an object detection network, named Lite-FCOS, whose backbone adopts a fast global context network (FGCNet) that is lightweight and is of powerful feature extraction capabilities.For training these, meter dial detection dataset (MDDD) and dial reading information dataset (DRID) are constructed.Lite-FCOS achieves 94.4 mAP50 and 96.7 mAP50 on the above two datasets with only 4.2M parameters and 56.2 FPS.The entire pointer meter reading recognition process only spends 52 ms on an RTX 3080Ti with a successful meter reading rate (SMR rate) of 89.6%, which indicates that the proposed method achieves promising accuracy and speed.

“…Finally, the meter reading was calculated based on the angle relationship between the pointer and the two nearest primary scales. Chen et al [38] introduced an improved YLU 2 -Net image segmentation method based on deep separable convolutions and focal loss functions. This method involves segmenting the pointer and the scale region, and then using a dimensionality reduction technique based on polar coordinate transformation to achieve meter reading.…”

Section: Competition Testsmentioning

confidence: 99%

Reading recognition of pointer meters based on an improved UNet++ network

Huo,

Bai,

Sun

et al. 2023

Pointer meters are widely used in modern industries, such as petrochemical applications, substations, and nuclear power plants. To overcome the reading errors and inaccurate measurements due to uneven or fluctuating illumination in practical applications, this paper proposes an improved UNet++ network for recognizing pointer meter readings. First, the scale invariant feature transform feature-matching algorithm is used to adjust the captured tilted meter images to a symmetrical and upright shape. Then, the UNet++ network is used to segment the scale and pointer regions in the dashboard to eliminate background interference. Furthermore, part of the convolution in the UNet++ network is replaced with dilated convolution with different expansion rates to expand the perceptual field during network training. In the UNet++ network jump connection, the attention mechanism module is also introduced in the path to enhance the region’s features to be segmented and suppress the parts of the non-segmented area. A hybrid loss function is used for the network model training to prevent the imbalance of the segmented region share. Finally, the distance method is used to read the gauge representation. Experiments were conducted to compare the performance of the proposed method with that of the original UNet++ network in terms of feasibility and precision. The experimental results showed that the recognition reading accuracy was significantly improved by the enhanced network, with the accuracy, sensitivity, and specificity reaching 98.65%, 84.33%, and 99.38%, respectively. Furthermore, when using the improved UNet++ network for numerical reading, the average relative error was only 0.122%, indicating its robustness in a natural environment.