Research On Cattle Image Based On Machine Vision And Sparse Reconstruction

In the current period,the consumption of beef and milk products by the residents has driven the scale of cattle breeding.The growth status detection in the cattle breeding process is an indispensable link.The most direct way is to measure each body size.The traditional manual measurement mostly fixes the cattle,which requires the cattle to have a correct posture and a wide field.and many people cooperate to complete the measurement quickly.Usually,the measurement is rough,and the result is not very accurate.This paper mainly uses the knowledge of image processing,combines the relevant knowledge of machine vision with sparse reconstruction,and completes the measurement of bovine body ruler by non-contact measurement with the help of MATLAB and other platforms.The main research work and content are:(1)Acquisition of cattle body image.Here,a digital camera is used to capture the image of the target cow.The collection site is selected from an outdoor pasture in a large pastoral area in Inner Mongolia.The camera is used to capture the image of the cow from multiple angles by panning the camera.(2)Image preprocessing.The first is the grayscale processing of the color image,and then the background is removed,and then the median filtering,noise reduction,sharpening and other operations are performed to improve the quality of the picture.The Canny algorithm and the OR operation are used to combine fuzzy segmentation and edge information.Image segmentation,remove broken edges,and perform digital morphological closure operations on the divided image to fill small holes in objects,connect adjacent objects,and smooth their boundaries.(3)Feature point detection and matching.Here,by comparing the SIFT and SURF algorithms in the number of feature points and the accuracy of matching,the SIFT algorithm is finally selected for feature point extraction and matching,and the RANSAC algorithm is used to remove the feature points that are incorrectly matched Finally,the images were weighted and fused.(4)Sparse reconstruction.Here we use the weighted fusion image to perform SFM operation,that is,to restore the structure from motion,and generate sparse point clouds on software and platforms such as VisualSFM and MeshLab to complete the sparse reconstruction process of the target cattle body contour.(5)Measurement of bovine body height and body length.The sparse reconstruction method is used to reconstruct the outline of the cow body,and the outline is highlighted.The cattle ear tag is used to calibrate to determine the correspondence between the length of the pixels in the image and the actual data.According to the definition of body height and body straight length,the image Find the measuring point on the top,and mark the body height and body straight length,and calculate the actual distance of body height and body straight length through conversion based on the pixel distance of the marking line segment.After comparing the selected data,we can see that the maximum relative errors of body height and body straight length obtained by image measurement are 0.47% and 0.08%,which are within the allowable error range required by body measurement,and the measurement data error is small.Compared with the traditional manual measurement,the measurement safety is improved,the measurement accuracy is high,and the measurement result is accurate.It can be applied to the daily measurement of the cattle body of the herdsman in modern livestock.