Deep learning-based express sorting operation standard detection method

Abstract

The invention discloses a CNN and LSTM network-based logistical sorting standard detection method, and belongs to the technical field of image processing. The method comprises the following steps of:firstly obtaining a continuous skeleton data frame sequence from somatosensory equipment when an express sorting action is executed by a worker, screening 20 pieces of joint data which can represent actions from the skeleton data, converting the joint data into world coordinates, and dividing the skeleton data into two parts; selecting certain key points from the skeleton data from a part of the skeleton data to form certain geometric features formed by points, lines and surfaces so as to describe the actions, and carrying out time-domain modeling on the actions through an LSTM network; projecting the other part of skeleton data to generate joint graphs, such as a front view, a left view and a top view, at three visual angles and motion trail graphs thereof, and taking the joint graphs andthe motion trail graphs are taken as inputs of a convolutional neural network to carry out feature extraction; and finally fusing the two parts of information to carry out action judgement. The method has good timeliness, robustness and correctness, is convenient and reliable to implement and is suitable for real-time action recognition systems.

Description

technical field [0001] The invention relates to a recognition method of human behavior, in particular to a deep learning-based express sorting operation specification detection method, which belongs to the technical field of video image processing. Background technique [0002] At present, the problems in the express delivery industry include backward technical level, lack of express delivery talents, tense relationship with customers, and irregular overall environment. According to statistics, the use of traditional manual sorting not only has a large workload and high error rate, but also the maximum amount of sorting per person per day is limited; while using automatic sorting technology, the number of sorting per machine per day is five times that of manual sorting Many, the error rate is also controlled within 5 / 10,000. However, due to the complexity of the actual situation, manual sorting still accounts for a large part. The low threshold of the express delivery indu...

AI Technical Summary

Problems solved by technology

[0002] At present, the problems in the express delivery industry include backward technical level, lack of express delivery talents, tense relationship with customers, and irregular overall environment, etc.

The RNN model establishes context dependencies in the time domain and is successfully applied to variable-length sequence data processing. It can handle time well, but it cannot handle spatial features well.

Images