The traditional K‐singular value decomposition (K‐SVD) algorithm has poor image‐denoising performance under strong noise. An image‐denoising algorithm is proposed based on improved K‐SVD and dictionary atom optimization. First, a correlation coefficient‐matching criterion is used to obtain a sparser representation of the image dictionary. The dictionary noise atom is detected according to structural complexity and noise intensity and removed to optimize the dictionary. Then, non‐local regularity is incorporated into the denoising model to further improve image‐denoising performance. Results of the simulated dictionary recovery problem and application on a transmission line dataset show that the proposed algorithm improves the smoothness of homogeneous regions while retaining details such as texture and edge.
9 January 2020 in Computational Intelligence and Neuroscience
Shadow detection and removal in real scene images are a significant problem for target detection. This work proposes an improved shadow detection and removal algorithm for urban video surveillance. First, the foreground is detected by background subtraction and the shadow is detected by HSV color space. Using local variance and OTSU method, we obtain the moving targets with texture features. According to the characteristics of shadow in HSV space and texture feature, the shadow is detected and removed to eliminate the shadow interference for the subsequent processing of moving targets. Finally, we embed our algorithm into C/S framework based on the HTML5 web socket protocol. Both the experimental and actual operation results show that the proposed algorithm is efficient and robust in target detection and shadow detection and removal under different scenes.
20 October 2022 in Wireless Communications and Mobile Computing
It is quite simple for foreign objects to attach themselves to transmission line corridors because of the wide variety of laying and the complex, changing environment. If these foreign objects are not found and removed in a timely manner, they can have a significant impact on the transmission lines’ ability to operate safely. Due to the problem of poor accuracy of foreign object identification in transmission line image inspection, we provide an improved YOLOX technique for detection of foreign objects in transmission lines. The method improves the YOLOX target detection network by first using Atrous Spatial Pyramid Pooling to increase sensitivity to foreign objects of different scales, then by embedding Convolutional Block Attention Module to increase model recognition accuracy, and finally by using GIoU loss to further optimize. The testing findings show that the enhanced YOLOX network has a mAP improvement of around 4.24% over the baseline YOLOX network. The target detection SSD, Faster R-CNN, YOLOv5, and YOLOV7 networks have improved less than this. The effectiveness and superiority of the algorithm are proven.
Glioblastoma is a solid, infiltrating, and the most frequent highly malignant primary brain tumor. Our aim was to find the correlation between sex, age, preoperative Karnofsky performance status (KPS), presenting with seizures, and extent of resection (EOR) with overall survival (OS), progression-free survival (PFS), and postoperative KPS, along with the prognostic value of IDH1, MGMT, ATRX, EGFR, and TP53 genes mutations and of Ki67 through the analysis of a single-operator series in order to avoid the biases of a multi-operator series, such as the lack of homogeneity in surgical and adjuvant nonsurgical treatments. A randomized retrospective analysis of 122 patients treated by a single first operator at Sapienza University of Rome was carried out. After surgery, patients followed standard Stupp protocol treatment. Exclusion criteria were: (1) patients with primary brainstem and spinal cord gliomas and (2) patients who underwent partial resections (resection < 90%) or a biopsy exclusively for diagnostic purposes. Statistical analysis with a simultaneous regression model was carried out through the use of SPSS 25® (IBM). Results showed statistically significant survival increase in four groups: (1) patients treated with gross total resection (GTR) (p < 0.030); (2) patients with mutation of IDH1 (p < 0.0161); (3) patients with methylated MGMT promoter (p < 0.005); (4) patients without EGFR amplification or EGFRvIII mutation (p < 0.035). Higher but not statistically significant survival rates were also observed in: patients <75 years, patients presenting with seizures at diagnosis, patients affected by lesions in noneloquent areas, as well as in patients with ATRX gene mutation and Ki-67 < 10%.
13 May 2019 in Computational Intelligence and Neuroscience
To overcome the shortcomings of inaccurate textural direction representation and high-computational complexity of Local Binary Patterns (LBPs), we propose a novel feature descriptor named as Local Dominant Directional Symmetrical Coding Patterns (LDDSCPs). Inspired by the directional sensitivity of human visual system, we partition eight convolution masks into two symmetrical groups according to their directions and adopt these two groups to compute the convolution values of each pixel. Then, we encode the dominant direction information of facial expression texture by comparing each pixel’s convolution values with the average strength of its belonging group and obtain LDDSCP-1 and LDDSCP-2 codes, respectively. At last, in view of the symmetry of two groups of direction masks, we stack these corresponding histograms of LDDSCP-1 and LDDSCP-2 codes into the ultimate LDDSCP feature vector which has effects on the more precise facial feature description and the lower computational complexity. Experimental results on the JAFFE and Cohn-Kanade databases demonstrate that the proposed LDDSCP feature descriptor compared with LBP, Gabor, and other traditional operators achieves superior performance in recognition rate and computational complexity. Furthermore, it is also no less inferior to some state-of-the-art local descriptors like as LDP, LDNP, es-LBP, and GDP.
28 April 2022 in Mathematical Problems in Engineering
To address automatic detection of pedestrian fall events and provide feedback in emergency situations, this paper proposes an attention-guided real-time and robust method for pedestrian detection in complex scenes. First, the YOLOv3 network is used to effectively detect pedestrians in the videos. Then, an improved DeepSort algorithm is used to track by detection. After tracking, the authors extract effective features from the tracked bounding box, use the output of the last convolutional layer, and introduce the attention weight factor into the tracking module for final fall event prediction. Finally, the authors use the sliding window for storing feature maps and SVM classifier to redetect fall events. The experimental results on the CityPersons dataset, Montreal fall dataset, and self-built dataset indicate that this approach has good performance in complex scenes. The pedestrian detection rate is 87.05%, the accuracy of fall event detection reaches 98.55%, and the delay is within 120 ms.