Piglet pressed detection method based on deep learning model

Abstract

The invention discloses a piglet pressed detection method based on a deep learning model, and relates to the technical field of breeding, and the technical scheme comprises steps: arranging a microphone array in a pigsty of a pig delivery room; collecting sound of two columns through the microphone; judging the position of a sound source according to the collected sound; and judging whether the piglet is pressed or not by combining a pressed piglet sounding model established based on the CNN and a spectrogram. The beneficial effects of the invention are that the method provided by the scheme liberates the demand of a delivery room for labor force, reduces the production management cost of an enterprise, and improves the digital intelligence level of livestock breeding. Besides, from the perspective of production, by means of the scheme and method, the death and culling rate of the piglets can be reduced, more piglets can be rescued, the piglets can be converted into growing-finishing pigs which continue to be bred, and great help is provided for further improving the operating profit of animal husbandry enterprises.

Description

technical field [0001] The invention relates to the field of breeding technology, in particular to a method for detecting piglets being crushed based on a deep learning model. Background technique [0002] According to statistics, in the process of sow farrowing, 1 out of every 10 piglets will be crushed to death by the sow. Nearly half of the deaths occurred in the first 3 days after birth. If the sow stands up within 3 minutes of being on the piglet, the piglet has a chance of survival. When this kind of incident occurs, the traditional method is that when the breeder finds that the piglet is being crushed, he rushes to the sow and makes it stand up and take the piglet by patting it. Stand up again, so this method of relying on human subjective will and behavior to rescue piglets is time-consuming and labor-intensive, and cannot effectively solve this pain point. [0003] In addition, the complexity of the livestock breeding environment also poses a huge challenge to th...

AI Technical Summary

Problems solved by technology

When this kind of incident occurs, the traditional method is that when the breeder finds that the piglet is being crushed, he rushes to the sow and makes it stand up and take the piglet by patting it. Stand up again, so this method of relying on human subjective will and behavior to rescue piglets is time-consuming and labor-intensive, and cannot effectively solve this pain point

[0003] In addition, the complexity of the livestock breeding environment also poses a huge challenge to the stability of some automated farming equipment. For example, the high temperature, high humidity, and high dust farming environment will have a greater impact on some automated measurement sensors. Some studies have used images or Sound technology solves this problem. It is hoped that by collecting the sound near the delivery bed where the sow is lying in real-time and non-contact, using pattern recognition technology to identify whether the piglet sends out a distress signal, and provide timely feedback information to the actuator (for example: mechanical arm , electrical stimulation devices, etc.) and professional breeders, to rescue the piglets in the first time, but there are no mature commercial products and solutions on the market to achieve this, and the recognition accuracy of piglets based on image / sound detection technology is in some cases. Aspects are lower than human ear / eye detection, and cannot be identified with 100% accuracy. The patent of the present invention aims at this problem, combined with actual production performance indicators, and proposes a statistical method for evaluating the effectiveness of piglet pressure detection methods. This method can analyze Whether the detected suppressed events are effectively processed, so that similar technologies have a broad research and improvement space, and evaluate their application feasibility

Images