Human action recognition is an essential process in surveillance video analysis, which is used to understand the behavior of people to ensure safety. Most of the existing methods for HAR use computationally heavy networks such as 3D CNN and two-stream networks. To alleviate the challenges in the implementation and training of 3D deep learning networks, which have more parameters, a customized lightweight directed acyclic graph-based residual 2D CNN with fewer parameters was designed from scratch and named HARNet. A novel pipeline for the construction of spatial motion data from raw video input is presented for the latent representation learning of human actions. The constructed input is fed to the network for simultaneous operation over spatial and motion information in a single stream, and the latent representation learned at the fully connected layer is extracted and fed to the conventional machine learning classifiers for action recognition. The proposed work was empirically verified, and the experimental results were compared with those for existing methods. The results show that the proposed method outperforms state-of-the-art (SOTA) methods with a percentage improvement of 2.75% on UCF101, 10.94% on HMDB51, and 0.18% on the KTH dataset.
Tomatoes will be harvested at the earlier stage for longer market life. Therefore while grading tomatoes, maturity as well the quality need to be considered. In this paper automatic and non destructive grading of cherry tomatoes based on maturity and quality has been proposed. This algorithm comprises two phases of grading. First phase grades tomatoes in terms of maturity and the second phase grades matured tomatoes in terms of quality. Using the concept of change in external color during different maturity stages, a color based maturity estimation algorithm has been proposed. And to classify the tomato in the second phase, texture, color and shape features correlated to external and internal characteristics are extracted from the surface of tomato. To extract the tomato surface alone from the background and leaves a new color based segmentation based on Euclidean distance has also been proposed. Then the extracted features are given to K-Nearest Neighbor based Support Vector Machine classifier to classify the matured fruit into three classes Class I, II and III. This classifier outperforms both SVM and KNN classifier in terms of accuracy and computation time.
8 September 2022 in Wireless Personal Communications
This paper presents a statistical view of the texture retrieval problem by combining the two related steps, feature extraction and similarity measurement. Based on spectral representation of texture images under Fourier transform, rotation invariant signatures of orientation spectrum distribution are extracted. Peak Distribution Vector (PDV) obtained on the spectral signatures capture texture properties invariant to image and surface rotation. The PDV is used to measure the similarity measurement by computing sum of square distance between query and data base images. The method is applied to content based retrieval system with a database of over 1000 randomly chosen texture images from photometric texture database. Experimental results indicate that the new method significantly improves the retrieval rates compared with the Zhang's approaches while it retains comparable levels of computational complexity.
This dataset provides three classes of hyperspectral images: pure, insecticide-immersed, and fungicide-immersed apples with different concentrations of fertilizers. The hyperspectral images were calibrated under white and dark correction and enhanced using contrast enhancement. In order to know the variations in the level of fertilizers used, we soaked the apples in 2 different concentrations of chemicals i.e., 1ml or 1g of fertilizer in 1 liter of water as low concentration, and 3ml or 3g of fertilizer in 1 liter of water as high concentration. The proposed dataset will help in finding the consumption level of fertilizers (pesticides) in apples.