<p>The automatic measurement of sheep body dimensions plays a vital role in advancing precision management in sheep farming. However, posture variability poses a major challenge, as it significantly affects the accuracy of dimensional data and limits the effectiveness of automated measurement in enhancing breeding efficiency. This study explored the impact of posture changes on body dimension parameters and proposed an automated measurement method incorporating posture compensation. Using a non-contact approach, three-view images of sheep are captured, allowing for the automatic extraction of length, width, and height parameters through the fusion of color and depth images. A convolutional neural network is employed to classify partial posture variations from top-down view images. Based on classification, a mapping model between posture and body dimensions is constructed, and a compensation algorithm using weighted data fusion is introduced to enable adaptive correction of measurement results. Experimental results show that, after compensation, the average relative errors for body length, body height, chest width, chest depth, and forelimb height are reduced to 3.11%, 1.93%, 3.38%, 2.52%, and 3.27%, respectively. Compared to uncorrected measurements, these results represent a substantial reduction in error, confirming the method’s effectiveness in addressing posture-induced inaccuracies. This research offers a novel, computer vision-based approach for non-contact calibration of sheep body dimensions.</p>