Method, apparatus and system for detecting carbon emission-involved gas from ruminant

A method for detecting carbon emission-involved gas from a ruminant, comprising:

• continuously receiving ear tag information of the ruminant that is collected by an identity information identification module in a monitoring range;
• determining an operation state of a feed supply system based on the ear tag information;
• and
• determining, based on collected emitted gas from the ruminant during eating, an emission rate of carbon emission-involved gas in the emitted gas when it is determined that the feed supply system is in a continuous operation state, wherein the determining an operation state of a feed supply system based on the ear tag information comprises:
• determining identity information of the ruminant based on the ear tag information;
• querying for a historical eating record of the ruminant from a target database based on the identity information;
• controlling the feed supply system to switch from a stop state to the continuous operation state so as to supply feed into a trough, if it is determined based on the historical eating record that the ruminant has not eaten within a first preset duration before a current moment;
• counting an operation duration of the feed supply system;
• controlling the feed supply system to switch from the continuous operation state to the stop state when it is determined that the operation duration has reached a second preset duration and/or the identity information identification module has not collected the ear tag information;
• and
• controlling the feed supply system to keep in the stop state if it is determined based on the historical eating record that the ruminant has eaten within the first preset duration before the current moment,
• wherein the stop state refers to a state in which a quantitative feeder is stopped, and does not put feed in a feed box into the trough;
• and the continuous operation state refers to a state in which the quantitative feeder puts the feed in the feed box into the trough in a preset mode,
• wherein before the determining, based on collected emitted gas from the ruminant during eating, an emission rate of carbon emission-involved gas in the emitted gas, the method further comprises:
• acquiring a plurality of depth image frames of the ruminant's head during eating collected by an RGB-D camera;
• inputting each depth image frame into a pre-trained target detection model to obtain a distribution feature map of feature points outputted by the target detection model;
• determining image coordinates of each feature point in each distribution feature map of the feature points, and converting the image coordinates of the feature point into spatial coordinates with the RGB-D camera as a coordinate origin, to obtain a feature point spatial coordinate data set, wherein the feature points comprise binaural feature points, binocular feature points and a mouth feature point of the ruminant;
• classifying, based on a support vector machine, spatial position relationships between all spatial coordinates in the feature point spatial coordinate data set and vent holes on the trough, and determining a head pose category of the ruminant relative to the trough;
• and
• marking effectiveness of the collected emitted gas based on the head pose category, wherein the vent holes are configured in the trough to collect the emitted gas,
• wherein the target detection model is obtained by training a basic network model by using a plurality of depth image samples of the ruminant's head and a distribution feature map label corresponding to each of the depth image samples;
• and the basic network model is generated based on ResNet as a basic network structure through replacing a classification layer of the ResNet with a deconvolution layer.