No abstract

No abstract

Methods to detect directly aphids based on convolutional neural networks (CNNs) are unsatisfactory because aphids are small and usually are specially distributed. To enhance aphid detection efficiency, a framework based on oriented FAST and rotated BRIEF (ORB) and CNNs (EADF) is proposed by us to detect aphids in images. Firstly, the key point is to find regions of aphids. Points generated by the ORB algorithm are processed by us to generate suspected aphid areas. Regions are fed into convolutional networks to train the model. Finally, images are detected in blocks with the trained model. In addition, in order to solve the situation that the coordinates are not uniform after the image is segmented, we use a coordinate mapping method to unify the coordinates. We compare current mainstream target detection methods. Experiments indicate that our method has higher accuracy than state-of-the-art two-stage methods that the AP value of RetinaNet with EADF is 0.385 higher than RetinaNet without it and the Cascade-RCNN with EADF is more than without it by 43.3% on value of AP, which demonstrates its competency.

Without ground-truth data, trajectory anomaly detection is a hard work and the result lacks of interpretability. Moreover, in most current methods, trajectories are represented by geometric features or their low-dimensional linear combination, and some hidden features and high-dimensional combined features cannot be found efficiently. Meanwhile, traditional methods still cannot get rid of the limitation of space attributes. Therefore, a novel trajectory anomaly detection algorithm is present in this article. Unsupervised learning mechanism is used to overcome nonground-truth problem and deep representation method is used to represent trajectories in a comprehensive way. First, each trajectory is partitioned into segments according to its open angles, then the shallow features at each point of a segment are extracted and. In this way, each segment is represented as a feature sequence. Second, shallow features are integrated into auto-encoder-based deep feature fusion model, and the fusion feature sequences can be extracted. Third, these fused feature sequences are grouped into different clusters using a unsupervised clustering algorithm, and then segments which quite differ from others are detected as anomalies. Finally, comprehensive experiments are conducted on both synthetic and real data sets, which demonstrate the efficiency of our work.

No abstract