A method for detecting beef cattle body shape parameters based on machine vision

Abstract

A method for detecting beef cattle body size parameters based on machine vision, which relates to the fields of image processing and breeding technology, and particularly relates to a method for detecting beef cattle body size parameters. Aiming at the measurement of beef cattle body size, this paper proposes a detection algorithm combining image processing and deep learning. This method first preprocesses the beef cattle images under a specific scale, and uses image feature calculation and deep learning methods, combined with visual computing technology to measure the following parameters of beef cattle: body height, body straight length, cannon circumference, chest width, chest depth, hip angle width, cross height, body diagonal length, hip width, etc. This method provides a convenient, simple and low-cost method for detecting beef cattle body size parameters, which is convenient for small and medium-sized farmers.

Description

Technical Field

[0001] The present invention relates to the technical fields of image processing and breeding technology, and particularly relates to a method for detecting beef cattle body shape parameters. Background Art

[0002] With the improvement of people's living standards, the beef cattle industry is becoming increasingly important in agricultural production and the national economy. Automation and industrialization are the inevitable trends of the beef cattle industry. In view of the problems that the manual measurement of beef cattle body shape parameters has low efficiency, low measurement accuracy, and the measurement results are easily restricted by the technical level, fatigue degree of the measurer and the accuracy of the measuring instrument itself, computer vision technology is applied to measure the main body shape parameters of beef cattle and manage the image information. Computer vision technology can obtain a large number of parameters and information from the acquired images without contacting the measurement object, and is particularly suitable for the detection and comprehensive quality evaluation of organisms such as animals, plants, and agricultural products. Summary of the Invention

[0003] This method provides a convenient, simple, and low-cost method for detecting beef cattle body shape parameters for small and medium-sized beef cattle farmers.

[0004] This method can realize the measurement of parameters such as the weight, body height, and body slant length of cattle only through the photos of cattle, without the need to purchase equipment such as weighing scales and calipers, and the measurement is simpler and more convenient.

[0005] A method for detecting beef cattle body shape parameters based on machine vision includes the following steps:

[0006] Step S1: Take images;

[0007] Step S2: Feature extraction;

[0008] Step S3: Calculate body shape parameters.

[0009] Preferably, the step of taking images in step S1 of the present invention includes the following steps:

[0010] S11. Lead the cattle into the shooting rack, fasten the slow rope, clamp the binding clip, and a scale should be fixed on the binding rack to make the cattle body in a natural state and the spine keep straight;

[0011] S12. Use a mobile phone camera to take a side photo of the cattle; the camera plane is parallel to the beef cattle, take a side photo of the beef cattle, and the scale should be included in the photo when taking the photo;

[0012] S13. Point the camera directly at the buttocks of the beef cattle and take a photo of the buttocks of the beef cattle, and the scale should be included in the photo when taking the photo;

[0013] S14. In front of the cattle, take a photo of the beef cattle, and the scale should be included in the photo when taking the photo;

[0014] S15. Above the cattle, take a top view of the beef cattle, and include the scale in the shot.

[0015] Preferably, the feature extraction in step S2 of the present invention includes the following steps:

[0016] S21. First, preprocess each image, including de - jittering, noise filtering, and image enhancement;

[0017] S22. Revise the shape and size of the image according to the scale;

[0018] S23. Calculate the circumscribed rectangle of the images taken in each direction;

[0019] S24. Use traditional feature extraction methods and deep networks to extract the positions of feature points, and the positions of the feature points include the cattle shoulder, cattle waist angle, the landing point of the front limb of the cattle, the highest point of the withers, the scapula, the spine, the sternum, and the intersection of the end of the front limb and the abdomen.

[0020] Preferably, the calculation of body type parameters in step S3 of the present invention includes the following steps:

[0021] S31. Use shape calculation techniques, such as approximate circumscribed ellipse and line recognition, to calculate body height, body diagonal length, body straight length, chest girth, cannon circumference, hind leg girth, chest width, chest depth, nine - length, loin angle width, loin height, ischial tuberosity width, and hip width;

[0022] S32. Calculate body length index, body trunk index, chest girth index, and cannon circumference index;

[0023] S33. According to the previous parameters, use a BP network to calculate the body weight.

[0024] The present invention adopts a standard image - processing process to calculate parameters such as the body weight, body length, and body height of beef cattle through pictures, without the need to purchase a weighing scale, calipers, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is the shooting stand used in the present invention.

[0026] Figure 2 is the scale used in the present invention.

[0027] Figure 3 is the back view of the beef cattle.

[0028] Figure 4 is the side view of the beef cattle.

[0029] Among them: 1. Body height, 2. Body diagonal, 3. Body straight length, 4. Chest girth, 5. Cannon circumference, 6. Hind leg girth, 7. Chest width, 8. Chest depth, 9. Length, 10. Loin angle width, 11. Loin height, 12. Ischial tuberosity width, 13. Hip width. Specific implementation mode

[0030] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings:

[0031] A method for detecting beef cattle body size parameters based on machine vision includes the following steps:

[0032] Step S1: Take images

[0033] S11. Lead the cattle into the shooting rack, fasten the slow rope, clamp the binding clip, and a scale should be fixed on the binding rack to make the cattle body in a natural state and the spine keep straight;

[0034] S12. Use a mobile phone camera to take a side view photo of the cattle; the camera plane is parallel to the beef cattle, and take a side view photo of the beef cattle, and the scale should be included in the photo when taking the photo;

[0035] S13. Direct the camera at the buttocks of the beef cattle and take a photo of the buttocks of the beef cattle. The scale should be included in the photo when taking the photo;

[0036] S14. In the front of the cattle, take a photo of the beef cattle, and the scale should be included in the photo when taking the photo;

[0037] S15. Above the cattle, take a top view photo of the beef cattle, and the scale should be included in the photo when taking the photo.

[0038] Step S2: Feature extraction

[0039] S21. First, preprocess each image, including de-shaking, noise filtering, and image enhancement;

[0040] S22. Revise the shape and size of the image according to the scale;

[0041] S23. Calculate the circumscribed rectangle of the images taken in each direction;

[0042] S24. Use traditional feature extraction methods and deep networks to extract the positions of feature points, and the positions of the feature points include the cattle shoulder, cattle waist angle, cattle front limb landing point, highest point of withers, scapula, spine, sternum, intersection of the front limb end and the abdomen.

[0043] Step S3: Body size parameter calculation

[0044] S31. Using shape calculation techniques, approximate circumscribed ellipse and line recognition, calculate body height, body oblique length, body straight length, chest girth, cannon circumference, hind leg girth, chest width, chest depth, nine lengths, loin angle width, loin height, ischial tuberosity width, hip width;

[0045] S32. Calculate body length index, body trunk index, chest girth index and cannon circumference index;

[0046] S33. According to the previous parameters, use the BP network to calculate the body weight.

Claims

1. A method for detecting beef cattle body size parameters based on machine vision, characterized in that It includes the following steps: Step S1: Take images; it includes the following steps: Step S11: Lead the cattle into the shooting stand, fasten the slack rope, clamp the binding clip, and a scale should be fixed on the binding stand to keep the cattle body in a natural state with the spine straight; Step S12: Use a mobile phone camera to take a side photo of the cattle; the camera plane is parallel to the beef cattle, and take a side photo of the beef cattle while including the scale in the photo; Step S13: Point the camera directly at the buttocks of the beef cattle and take a photo of the buttocks while including the scale in the photo; Step S14: In front of the cattle, take a photo of the beef cattle while including the scale in the photo; Step S15: Above the cattle, take a top view photo of the beef cattle while including the scale in the photo; Step S2: Feature extraction, which includes the following steps: Step S21: First, preprocess each image, including de-shaking, noise filtering, and image enhancement; Step S22: Revise the shape and size of the image according to the scale; Step S23: Calculate the circumscribed rectangle of the images taken in each direction; Step S24: Use traditional feature extraction methods and deep networks to extract the positions of feature points, and the positions of the feature points include the cattle shoulder, cattle waist angle, front limb landing point of the cattle, highest point of the withers, scapula, spine, sternum, intersection point of the front limb end and the abdomen; Step S3: Calculate body type parameters.

2. The method for detecting beef cattle body shape parameters based on machine vision according to claim 1, wherein The calculation of the body type parameters in the above step S3 includes the following steps: Step S31: Use shape calculation techniques, approximate circumscribed ellipse, line recognition, to calculate body height, body diagonal length, body straight length, chest girth, cannon circumference, hind leg girth, chest width, chest depth, nine lengths, loin angle width, loin height, ischial tuberosity width, hip width; Step S32: Calculate body length index, body trunk index, chest girth index, and cannon circumference index; Step S33: According to the previous parameters, use the BP network to calculate the body weight.

Images