Pig body size and weight estimation method based on deep learning
A technology of deep learning and body weight, applied in the field of deep learning, can solve problems such as easy to miss features, not as good as deep learning methods, etc., and achieve the effect of comprehensive feature extraction
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Example Embodiment
[0036] Example 1
[0037] A deep-study-based porcum body weight estimation method, including steps:
[0038] S1, get the pig image;
[0039] S2, using a key point detection algorithm to perform critical point detection, obtain a critical point detection result and detect the results of the pig in the picture in the picture, and retain the full image of the pig in the picture;
[0040] S3, detecting that the pig is tilted in the picture, correcting the screen of the pig, and obtains an image that is complete and not inclined in the picture in the picture;
[0041] S4, input the image input body weight estimate and calculate the size of the size according to the critical point detection results to obtain the body weight and size of the pig;
[0042] figure 1 The image screening of the image corresponding to step S1 to S3, step S1, the pig only is an ordinary planar image taken by a normal 2D color camera, and the 2D color camera of this embodiment is only schematically shown. figur...
Example Embodiment
[0047] Example 2
[0048] A deep-study-based porcum body weight estimation method, including steps:
[0049] S1, get the pig image;
[0050] S2, using an instance division algorithm to instance segmentation, a pixel mark that belongs to a pig only, the instance division algorithm is based on the Mask RCNN instance division network, and then uses a key point detection algorithm to only critical the pigs in the image. Point detection
[0051] The instance division algorithm network structure image 3 As shown, the instance segmentation process includes:
[0052] First, the image is sent into the RESNext101 feature extraction network in the Mask RCNN instance split network;
[0053] Then set a fixed number of interested points for each pixel location of the feature, and the region of interest is sent into the Mask RCNN instance split network. The network is proposed to obtain the foreground and the background, and the coordinates return, thus Get high quality of interest;
[0054] The...
PUM
Abstract
Description
Claims
Application Information
- R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap