Image target detection method in smart home environment

Abstract

The invention relates to an image target detection method in a smart home environment. The method comprises the following steps: pre-training a model through ImageNet data, fusing multiple image enhancement modes by adopting a random seed to carry out home data enhancement and expansion preprocessing operation, introducing hole convolution by utilizing a feature extraction network, and carrying out model re-training on a processed home data set by adopting pre-trained model parameters; storing the secondarily trained model, packaging the model, and performing k-means clustering analysis on images in an image library and a detection library to form a specific target detection feature library; and when a single home image is input, carrying out feature extraction on the input image by usingthe feature extraction network to obtain four coordinates of a prediction frame, carrying out regression and classification calculation on the prediction frame, and outputting a detection result through non-maximum suppression. The target detection requirement in the smart home environment is met.

Description

technical field [0001] The invention relates to an image target detection method in a smart home environment. Background technique [0002] Object detection is an important technology in computer vision, and it has a wide range of applications in automotive autonomous driving, intelligent robot technology, intelligent security and other fields. The classic target detection methods include the detection method based on HOG features proposed by Dalal in 2005, and the Deformable PartModel (DPM) detection method proposed by Felzenswalb et al. in 2008. This method first uses the gradient operator to calculate The HOG feature of the target object is classified by the method of sliding window + SVM, and it performs well in target detection. [0003] In recent years, with the substantial improvement of computing performance, the rapid development of artificial intelligence and neural networks, various computer vision processing methods based on deep learning have been widely used. ...

AI Technical Summary

Problems solved by technology

Currently widely used neural network-based target detection methods are mainly divided into two categories: one is "two-stage detector", which divides target detection into two steps, first determines the candidate frame and then identifies the target in the area , the detection accuracy of this type of method is relatively high, the detection speed is relatively low, generally only up to 5fps, typical networks include RCNN, FAST-RCNN, FASTER-RCNN, etc.; the other type is "one-stage detector", this type of method Use the idea of ​​regression to complete the detection and recognition of the rear box at the same time, and realize end-to-end detection and recognition. Typical networks include YOLO, SSD, etc. The detection speed of this type of method is extremely fast, but the detection accuracy is relatively low

Images