In order to solve the problem of lack an index system influencing factors and unclear evolution path prefabricated housing development in Hainan Province, a method identifying key analyzing based on back propagation neural network (BPNN) decision experiment evaluation laboratory (DEMATEL) was proposed. Firstly, constructed grounded theory; then, were revealed through expert survey BPNN-optimized DEMATEL model; finally, explored. The research results show that F1 (imperfect standards...

To enhance land use/cover change (LUCC) simulation accuracy, we introduced ViViT-ANN-CA, blending video vision transformer's spatio-temporal features extraction ability, artificial neural network's (ANN) non-linearity computing and CA's spatial computing. Compared to 3DCNN-ANN-CA, ViViT-ANN-CA showed higher accuracy in simulating water bodies vegetation, with overall improvements Hailing District Wuxi City. ViViT demonstrates comparable feature ability three-dimensional convolutional network...

While scanning pages from a thick, bound book, there are two sources of distortion in the document images: 1) shade along book 'spine', and 2) warping surface area. We propose fast method to estimate 3D shape surface. Based on this shape, we remove correct restore images. The experiments show that photometric geometric distortions mostly removed. OCR tests original restored images also presented.

The camera calibration in monocular vision represents the relationship between pixels’ units which is obtained from a and object real world. As an essential procedure, calculates three-dimensional geometric information captured two-dimensional images. Therefore, modified method based on polynomial regression proposed to simplify. In this method, parameter vector by pixel coordinates of obstacles corresponding distance values using regression. set parameter’s vectors can measure ground field...

A novel calibration and lens distortion correction of 3D reconstruction systems is introduced in this paper. Assuming zero at the center image, parameter CCD camera DLP projector separated with points all rest points. Furthermore, identification correspondence interest done automatically because designed target standard circles multicolor code pseudo-random array during calibration. The experiments show that it a fast convenient method to carry out for distortion.

Blind source separation (BSS) denotes observing mixtures of independent sources, and by making use these signals only nothing else ,recovering the original signals. Joint Approximative Diagonalization Eigenmatrices (JADE) is a statistic-based method for transforming an observed random vector into components that are as mutually possible. In this paper, we employ JADE separating complex-valued Results simulation show can separate sources form observations effectively.

In this paper, a novel triplet network is proposed for face sketch recognition.A spatial pyramid pooling layer introduced into the to deal with different sizes of images, and an attention model on image space extract features from same location in photo sketch.Our mechanism builds improves recognition accuracy by searching similar regions which include abundant information order distinguish persons photos sketches.So that cross-modality differences between images are reduced when they mapped...

Face sketch recognition identifies the face photo from a large dataset. Some traditional methods are typically used to reduce modality gap between photos and sketches gain excellent rate based on pseudo image which is synthesized using corresponded photo. However, these cannot obtain better high for all datasets, because use of extracted features lead elimination effect different modalities' images. The feature representation deep convolutional neural networks as feasible approach...

In electric power system reform environment, grid companies have to estimate the economic performance of construction project completed with a more market-oriented perspective.Therefore, efficiency post-appraisal model is proposed in this paper.In view problem evaluating benefits only from financial angle existing research, lack comprehensiveness and depth, including costs, financial, social environmental impact costs four aspects were build as index network hierarchy.On basis, ANP method...

In this paper, we present a novel Siamese graph convolution network (GCN) for face sketch recognition. To build from an image, utilize deep learning method to detect the image edges, and then use superpixel segment edge image. Each segmented region is taken as node, each pair of adjacent regions forms graph. Graphs both photo are input into GCN A matching used share messages between cross-modal graphs in model. Experiments show that can obtain high performance on several photo-sketch...

Defects such as insulator, pins, and counterweight in highvoltage transmission lines affect the stability of power system. The small targets pins unmanned aerial vehicle (UAV) inspection images occupy a proportion characteristic representations are poor which results low defect detection rate high false positive rate. This paper proposed line pin algorithm based on improved Faster R-CNN. First, pre-training weights with higher matching degree obtained transfer learning. And it is applied to...

UAV inspection of overhead transmission lines is a common means for power grid companies to conduct defect detection, but in bad weather, the performance target detection algorithm deployed on onboard equipment often decreases significantly. Based classical atmospheric light scattering model, estimation first done more accurately by separating sky area. In terms estimation, synthetic estimated from comprehensive information loss and image texture weighting, finally color bias corrected...

To improve the performance of MC-EZBC at low bit-rate, an efficient motion vector tree pruning method based on property motion-compensated temporal filtering is proposed and applied to MC-EZBC. The new improves bit-rate with almost same high bit-rate.