A system and method for automatic tracking and monitoring video acquisition of farm targets

Abstract

The invention relates to a system and method for automatically tracking and monitoring video collection of farm targets. The system includes a patrol robot and a host computer monitoring system; Set up a square limit device, the mechanical arm is installed on the inspection track through the connecting rod, the camera is fixed at the other end, and the host computer monitoring system controls the inspection robot; method steps: based on the detection and positioning results of the breeding target, control the movement of the mechanical arm, and collect directional video , the upper computer monitoring system or the breeding personnel judge the abnormal behavior of the breeding target, determine the location where the suspected abnormal behavior occurs, control the robot to move to the designated position, and collect the fixed-point video; adopt the improved SSD detection method and the FD‑SSD detection method to detect excessive aggregation of the breeding target or fighting behavior, increase the time of video collection, and regularly video collection; the whole system and method are reasonable in design, easy to operate, and can realize real-time and automatic monitoring of breeding targets.

Description

technical field [0001] The invention belongs to the field of automatic tracking and monitoring of farm video, in particular to a system and method for automatically tracking and monitoring video of a farm target. Background technique [0002] my country is the largest livestock and poultry breeding and consuming country in the world. The total production of livestock and poultry meat accounts for half of the world's total meat. The traditional livestock and poultry breeding is inefficient and cannot meet the growing market demand. From the free-range farming mode of small and medium-sized households to the large-scale and intensive farming mode; in intensive farming, the traditional abnormal monitoring method of breeding targets relies on a large amount of manpower and material resources, which is inefficient and difficult to guarantee the effect. Automatic and intelligent monitoring is the future development trend ; Applying machine vision to livestock and poultry farming can...

AI Technical Summary

Problems solved by technology

[0002] my country is the largest livestock and poultry breeding and consuming country in the world. The total production of livestock and poultry meat accounts for half of the world's total meat. The traditional livestock and poultry breeding is inefficient and cannot meet the growing market demand. From the free-range farming mode of small and medium-sized households to the large-scale and intensive farming mode; in intensive farming, the traditional abnormal monitoring method of breeding targets relies on a large amount of manpower and material resources, which is inefficient and difficult to guarantee the effect. Automatic and intelligent monitoring is the future development trend ; Applying machine vision to livestock and poultry breeding can obtain real-time information on the behavior of breeding targets without interfering with their normal life

[0003] However, in the process of real-time acquisition of breeding target video in the farm, due to the limitation of camera shooting angle and shooting range, there will be visual blind spots, which will affect the tracking and monitoring of breeding targets. In order to automatically monitor breeding targets and achieve full coverage of breeding target videos, traditional methods The monitoring method requires the installation of a large number of cameras, which seriously affects the economy of breeding target tracking and monitoring, and also brings difficulties to the analysis and processing of multi-source video data. A similar system is used in indoor indicator light recognition by Zhejiang Guozi Company. The inspection robot uses the method of converting the collected images into the HSV color space type to calculate the average brightness of the indicator light area in each target image, and count the number of times the indicator lights are on within the preset collection time period. It can process a limited number of images in a period of time, and the real-time performance is poor; Panzhihua University proposed a video detection technology that integrates Gaussian model and Canny operator for traffic information collection. When Canny operator performs edge extraction, the edge in the image It can only be marked once, and the possible image noise should not be marked as an edge, and it cannot be automatically tracked and shot; Tianjin Ya'an Technology Co., Ltd. proposes a video surveillance system for target detection and tracking based on a multi-feature deep neural network. The YOLO framework based on the deep learning algorithm pre-trained the deep learning model, which cannot automatically judge and select the target to be tracked

Images