Group-housed pig detection in video surveillance of overhead views using multi-feature template matching

Li, Bo; Liu, Longshen; Shen, Mingxia; Sun, Yuanyuan; Lu, Mingzhou

doi:10.1016/j.biosystemseng.2019.02.018

Cited by 24 publications

(19 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The RPN then completes the following two tasks: (1) determining whether the anchors are targets or non-targets (2k); (2) performing coordinate correction on the target anchors (4k). In the classification layer branch, two scores (target and non-target) are generated for each anchor; in the regression layer branch, the parameterizations of the four coordinates are corrected, and shown as Equation (1). For each target anchor:…”

Section: Rois Generation Based On Rpnmentioning

confidence: 99%

“…With the development of artificial intelligence and automation technology, utilizing video cameras to monitor the health and welfare of pigs has become more important in the modern pig industry. In group-housed environments, instance segmentation of pigs includes detection, which automatically obtains the positions of all pigs, and segmentation, which distinguishes each pig in the images [1]. Many high-level and intelligent pig farming applications, such as pig weight estimation [2], pig tracking [3], and behavior recognition [4][5][6][7], require accurate detection and segmentation of pig objects in complex backgrounds.…”

Section: Introductionmentioning

confidence: 99%

“…Pig detection and seg-mentation methods include two categories based on non-deep learning and deep learning algorithms. These non-deep learning approaches have more mature technologies and have been widely applied in video surveillance of pigs [1,[10][11][12]. Guo et al [10] proposed a pig's foreground detection method based on the combination of a mixture of Gaussians and threshold segmentation.…”

Section: Introductionmentioning

confidence: 99%

“…A recognition rate with 91.86% was obtained by this algorithm. Li et al [1] combined appearance features and the template matching framework to detect each pig under a group-housed environment. From these related studies, we found that these non-deep learning methods have some disadvantages, and that feature extraction and recognition of the pigs are separated.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Automatic Detection and Segmentation for Group-Housed Pigs Based on PigMS R-CNN

Yuan

Liang

et al. 2021

Sensors

View full text Add to dashboard Cite

Instance segmentation is an accurate and reliable method to segment adhesive pigs’ images, and is critical for providing health and welfare information on individual pigs, such as body condition score, live weight, and activity behaviors in group-housed pig environments. In this paper, a PigMS R-CNN framework based on mask scoring R-CNN (MS R-CNN) is explored to segment adhesive pig areas in group-pig images, to separate the identification and location of group-housed pigs. The PigMS R-CNN consists of three processes. First, a residual network of 101-layers, combined with the feature pyramid network (FPN), is used as a feature extraction network to obtain feature maps for input images. Then, according to these feature maps, the region candidate network generates the regions of interest (RoIs). Finally, for each RoI, we can obtain the location, classification, and segmentation results of detected pigs through the regression and category, and mask three branches from the PigMS R-CNN head network. To avoid target pigs being missed and error detections in overlapping or stuck areas of group-housed pigs, the PigMS R-CNN framework uses soft non-maximum suppression (soft-NMS) by replacing the traditional NMS to conduct post-processing selected operation of pigs. The MS R-CNN framework with traditional NMS obtains results with an F1 of 0.9228. By setting the soft-NMS threshold to 0.7 on PigMS R-CNN, detection of the target pigs achieves an F1 of 0.9374. The work explores a new instance segmentation method for adhesive group-housed pig images, which provides valuable exploration for vision-based, real-time automatic pig monitoring and welfare evaluation.

show abstract

Section: Rois Generation Based On Rpnmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Automatic Detection and Segmentation for Group-Housed Pigs Based on PigMS R-CNN

Yuan

Liang

et al. 2021

Sensors

View full text Add to dashboard Cite

show abstract

“…Dong et al [5] proposed to learn the corresponding target contour model from the segmented image and then used the boost classifier to find the contour of the target in the image so as to obtain the position information of each part of the human body. e literature [6,7] uses the HOG method to extract the information of each part of the human body in the image and then uses the classical algorithm support vector machine and random forest to identify and classify. Cui et al [8] found the global optimal features from many features such as Fourier descriptors, shape context, edges, and gradients to quickly and accurately complete the backprojection process from features to three-dimensional poses.…”

Section: Introductionmentioning

confidence: 99%

Human Motion Gesture Recognition Based on Computer Vision

Zhang

Chen³

2021

Complexity

View full text Add to dashboard Cite

Human motion gesture recognition is the most challenging research direction in the field of computer vision, and it is widely used in human-computer interaction, intelligent monitoring, virtual reality, human behaviour analysis, and other fields. This paper proposes a new type of deep convolutional generation confrontation network to recognize human motion pose. This method uses a deep convolutional stacked hourglass network to accurately extract the location of key joint points on the image. The generation and identification part of the network is designed to encode the first hierarchy (parent) and the second hierarchy (child) and show the spatial relationship of human body parts. The generator and the discriminator are designed as two parts in the network, and they are connected together in order to encode the possible relationship of appearance and, at the same time, the possibility of the existence of human body parts and the relationship between each part of the body and its parental part coding. In the image, the key nodes of the human body model and the general body posture can be identified more accurately. The method has been tested on different data sets. In most cases, the results obtained by the proposed method are better than those of other comparison methods.

show abstract