In order to solve the problem that target tracking frames are lost during the visual tracking of pigs, this research proposed an algorithm for multi target pigs tracking loss correction based on Faster R-CNN. The video of live pigs was processed by Faster R-CNN to get the object bounding box. Then, the SURF and background difference method were combined to predict whether the target pig will be occluded in the next frame. According to the occlusion condition, the maximum value of the horizontal and vertical coordinate offset of the bounding box in the adjacent two frames of the frame image in continuous N (N is the value of the video frame rate) were calculated. When bounding boxes in a video frame are merged into one bounding box, this maximum value was used to correct the current tracking frame offset in order to achieve the purpose of solving the tracking target loss problem. The experiment results showed that the success rate range of RP Faster-RCNN in the data set was 80%-97% while in term of Faster-RCNN was 40%-85%. And the average center point error of RP Faster-RCNN was 1.46 lower than Faster-RCNN which was about 2.60. The new algorithm was characterized by good robustness and adaptability, which could solve the problem of missing tracking target and accurately track multiple targets when the targets occlude each other. Keywords: object tracking, Faster-RCNN, individual pig, target occlusion DOI: 10.25165/j.ijabe.20181105.4232 Citation: Sun L Q, Zou Y B, Li Y, Cai Z D, Li Y, Luo B, et al. Multi target pigs tracking loss correction algorithm based on Faster R-CNN. Int J Agric & Biol Eng, 2018; 11(5): 192–197.

Fish swarm image segmentation provides an easy to understand and analyze representation for behavior monitoring feature extraction and image information analysis, accurate and effective image segmentation is the basis of fish shoal monitoring. In this paper, algorithm of adaptive fast clustering for fish swarm color image segmentation was proposed by combining the fish swarm hypoxia image features and K-Means++ algorithm. In the RGB color space, the color information of the channel with the maximum average brightness is retained, the others compensation was zero, the generated new image replaced the original. The fish color library was constructed using the gray distribution statistics of the fish swarm targets and background. The pixel probability distribution value in each gray scale range of the normalized gray histogram of the newly generated image is calculated, and combine with the fish group target gray scale reference statistic, the clustering value is determined by two traversing. According to the reserved channel information, the corresponding cluster fish swarm color library is selected for color clustering. The clustering result is processed by threshold transformation to finally realize fish swarm image segmentation. Experiment showed that the average range of structural similarity of our algorithm was [0.93, 1], and the average range of peak signal-to-noise ratio was [44, 50], the running time of the algorithm in this paper is 56% shorter than K-Means ++ algorithm and 71% shorter than the fuzzy clustering algorithm when processing the same images, which could met the requirements of image segmentation quality and accuracy for fish behavior detection.

An accurate and rapid pig detection algorithm based on video image processing technology can be helpful to identify abnormal pigs and take timely measures to reduce the incidence of diseases. In order to solve the problems of low computational efficiency and low precision in pig detection algorithm based on sliding windows, this paper proposed a simple and efficient pig detection algorithm. A two-level support vector machine model was trained to calculate the probabilities of sliding windows by using gradient and gray distribution features of pigs. The principal component analysis convolution kernels were trained to extract foreground and background features of pig images. The support vector machine was used to classify sliding windows to obtain windows where pigs were located, and the non-maximum suppression algorithm was used to eliminate redundant windows to complete the target detection. The experiments showed that the proposed algorithm blending gradient and gray distribution features had a higher recall rate than the BING algorithm. The recall rate was up to 99.21% using 500 windows. The classification accuracy of sliding windows in this paper was 95.21%, which was higher than that of the PCANet. By calculating the omission detection rate, the misdetection rate, and the average detection time, it can be seen that in the detection methods of the proposed algorithm, BING + PCANet, faster rcnn and yolo, the performance of the proposed algorithm was optimal.

Individual pig object detection algorithm based on Gaussian mixture model

Pig target tracking algorithm based on multi-channel color feature fusion