Grain depot personnel nonstandard operation detection method based on improved YOLOv3 algorithm

Abstract

The invention discloses a grain depot personnel nonstandard operation detection method based on an improved YOLOv3 algorithm. The grain depot personnel nonstandard operation detection method comprises the following steps: carrying out image acquisition on grain depot personnel nonstandard operation behaviors; inputting the acquired image into a grain depot non-standard operation behavior detection network in an upper computer; outputting a result picture with an accurate mark of the non-standard operation behavior of the personnel; and displaying the result and storing same in the upper computer. The grain depot non-standard operation behavior detection network comprises a trunk layer constructed based on a YOLOv3 network and a feature fusion output layer, and a scale context selection attention module SCA is embedded in a Y2 layer and a Y3 layer of the feature fusion output layer respectively. The invention overcomes the defects in the prior art, and provides the improved YOLOv3 algorithm-based detection method for the irregular operation of the grain depot personnel, which is more suitable for grain depot scene detection and has stronger network characterization capability.

Description

technical field [0001] The invention relates to the fields of computer vision and image recognition, in particular to a method for detecting irregular operations of grain depot personnel based on the improved YOLOv3 algorithm. Background technique [0002] As an important field in computer vision, target detection is widely used in reality. Its goal is to detect the object target that needs to be recognized in a given image, and determine the category of the object and its position in the image. Before the large-scale application of deep learning in the field of computer vision, the progress of target detection accuracy is relatively slow. For example, traditional manual construction feature algorithms such as HOG features and Haar features plus classification algorithms such as SVM algorithm and Adaboost algorithm are used to improve accuracy. more difficult things. The convolutional neural network AlexNet that appeared in the ImageNet image classification competition show...

AI Technical Summary

Problems solved by technology

However, the number of prior frames in the YOLOv3 algorithm is pre-selected, and its size is obtained by clustering the detected objects on the ImageNet dataset, which is not applicable to the grain depot operation scenario. In terms of scale and target prediction at different locations, YOLOv3 also has insufficient network representation capabilities and poor detection results, and YOLOv3 directly uses concatenation of global small-scale features and fine-grained features in the feature fusion stage of the network. method fusion, the global scene cannot be well considered when detecting small targets

Images