AI that detects piglet crushing and coughing sounds… boosts pig farm productivity

The productivity and profitability of pig farms are determined by the survival rate of piglets. Crushing accidents caused when sows lie on piglets, or inadequate early response to respiratory diseases, lead to pig deaths and deal a critical blow to farm profitability.

According to the ‘Handon Farms 2025 Nationwide Korean Pork Farm Performance’ released by the Korean Pork Producers Self-Help Fund Management Committee, the MSY (marketed pigs per sow per year) of the top 10% farms among professional users was 24.4, while the bottom 10% recorded 15.7, a gap of 8.7 pigs between the top and bottom groups.

This productivity gap directly translates into an income gap for farms. Converted based on a farm with 100 sows, the annual number of pigs shipped differs by as many as 870 head. If the price per shipped pig is calculated at 535,000 won (based on a shipping weight of 115 kg, carcass yield of 77.5%, and 6,000 won per kg of carcass), the annual income gap reaches as much as 465.45 million won. In other words, even with the same farm size, differences in management and survival rates alone can determine income by hundreds of millions of won.

Triple sensors track ‘behavior, body temperature, and sound’ simultaneously… cough detection accuracy exceeds 95%
As improving piglet survival rates and narrowing productivity gaps emerge as key tasks for raising the income of pig farmers, a precision monitoring system using artificial intelligence (AI) and sensor fusion technology is drawing attention.

A research team led by Prof. Jeong Seon-ok of the Department of Smart Agriculture Systems and Mechanical Engineering at Chungnam National University is promoting a ‘field demonstration project for an intelligent pig health monitoring system’ as part of the ‘2026 Agricultural Technology Industry-Academia Cooperation Support Project’ operated by the Rural Development Administration and implemented by the Korea Institute for Advancement of Technology in Agriculture. This study, led by Chungnam National University with joint participation from the National Korea Agricultural and Fisheries University (Prof. Lee Su-hyeop’s team), is at the stage of verifying the technology by applying the team’s patented technologies—an ‘AI-based pig disease management system’ and a ‘monitoring system for disease management in pig houses’—to actual pig farms.

The demonstration prototype deployed in the field integrates three types of precision sensors and a Raspberry Pi 4B-based data acquisition device (DAQ) to monitor the inside environment of pig houses in three dimensions. The three precision sensors are an RGB camera (Raspberry Pi Camera V2), a thermal imaging camera (Thermal Expert Q1), and a sound sensor (Shure V7 microphone).
 

First, the RGB camera automatically counts the number of pigs in each pen and tracks major behavior patterns such as lying down, standing, walking, and feeding. The accuracy of individual counting and behavior recognition reaches 95–97% based on the first demonstration product. The thermal imaging camera measures the maximum, minimum, and average values of pigs’ body surface temperature in real time to detect individuals suspected of fever and the risk of heat stress due to high temperatures. The sound sensor detects coughing sounds in pig houses (detection accuracy over 95%), piglet crushing risk sounds (accuracy over 90%), and abnormal vocalizations such as tail biting in the 50Hz–16kHz band, responding within 1.5 seconds.

The collected video and audio data are transmitted to a server via wireless communication (Wi-Fi), and farm owners can remotely inspect the inside status of pig pens 24 hours a day through a PC web dashboard or smartphone app. If an emergency such as abnormal sounds or fever occurs, an alert is immediately sent to the farmer’s mobile device.

The research team is building modules in pens for pregnant sows and piglets at farms in Buyeo, South Chungcheong Province, and Gimje, North Jeolla Province, while collecting data and advancing AI models. On July 21, they held a briefing session and training at the Buyeo-gun Agricultural Technology Center, attended by about 30 officials and farmers from the Buyeo branch of the Korean Pork Producers Association, and completed a technology demonstration.

This intelligent monitoring system is being evaluated as a technology that can lower piglet mortality and raise pig farm productivity, including MSY, by allowing AI to serve as an early warning system in pig house environments where it is difficult for people to remain on site 24 hours a day.

Prof. Jeong Seon-ok said, “Through a monitoring system using AI technology, it is possible to identify the condition of pigs that cannot be known unless the farm owner or staff are constantly present. After confirming a problem through monitoring, they can visit the pig house and take action,” adding, “Because abnormal conditions in pigs can be detected early, it is effective for improving productivity and reducing labor.” She added, “Through this field demonstration, we will further advance it into a technology that farms can actually use effectively.”