A non-contact image data collection platform for pigs was designed to solve the problems of the response to stress， large errors， and low efficiency caused by the manual contact measurement of body size in breeding pigs. A method of detecting back posture and estimating body size in pigs was proposed based on YOLOv11 and support vector regression （SVR） algorithm. The YOLOv11 model was used to detect the back posture in pigs and the SVR algorithm was used to process the information of body size from the results of detection to estimate the body size in pigs. The results showed that the recall and average precision of the YOLOv11 model reached 94.6% and 96.0%， respectively， indicating that the model has a good robustness of detection. The mean absolute percentage errors between the estimated and measured value of body length， chest width， and hip width obtained through the SVR algorithm was 2.78%， 2.55%， and 2.88%， respectively， indicating that the algorithm has high accuracy in measuring the body size. It is indicated that the method of detecting back posture and estimating body size in pigs based on YOLOv11 and SVR algorithm has the characteristics of lightweight and high accuracy， providing an efficient and reliable tool for breeding selection in pig farms while reducing human error and the response to stress in pigs.