Research On Triple-color Cable Sequence Detection Method

In the modern industrial production process,many kinds of workpieces to be processed need to be arranged in order according to the specified order.However,due to various external factors in the automated production process,the initial arrangement order of the workpieces is usually disordered,so the order is specified.The problem of workpieces ordering has become a challenging research topic.In the field of serial interface soldering applications,the soldering between the wires and the joints must meet the specified sequence requirements to ensure the safety of industrial production.Specifically,wires of different colors are soldered to different joint locations.The existing cable detection methods have achieved some results,but have the following deficiencies:(1)The sorting of wires based on manual inspection is inefficient and costly,resulting in a waste of a large amount of resources;(2)The high-intensity manual inspection method is easy to cause the missing and misjudgment of the workpieces,which cannot meet the requirements of current industrial automation production;(3)The existing image-based cable detection algorithm lacks versatility,and the robustness of the detection results is not ideal.In view of the above problems,this paper takes the three-color cable in industrial production as the research object,and studies the three-color cable sequential detection technology around the visual word bag model and deep learning model in the image processing field.The details are as follows:(1)In this paper,in the current industrial production,the sequential detection of cables is time-consuming and laborious,and the efficiency is not high.The sequential detection of cables is studied.By reading a large amount of literature,it is firstly determined to be simple and effective.The Bag of Words(BOW)model is the main entry point for research.In this paper,the preprocessing part of the triple-color cable image will be added,that is,the feature is better represented by the Region of Interest Location(ROIL)method.Secondly,the performance of the Region of Interest Location method under different thresholds and different visual lexicons is studied.The Bag of Words model using the Region of Interest Location(ROIL)method can effectively reduce the number of invalid features in the cable image,thus improving the accuracy of model classification.(2)The Bag of Words model is a statistically related model structure.Local feature quantization is used to sparsely express some features.It is mostly applied in image representation in the traditional sense.In this paper,we use a deep learning model and a Bag of Words model to combine the triple-color cable sequence detection method.The ability to better express triple-color cable images,while the Bag of Words model can solve the problem that small-area targets are not easy to detect.Experiments show that the VGG-BOW model has better recognition accuracy,the accuracy rate is 0.61% higher than the original VGG network,showing a better detection effect.(3)Some traditional target detection algorithms extract features by hand.External factors,acquisition factors,and artificial conditions all have some effects on accuracy.At the same time,artificial extraction features are not efficient and generalization performance is not strong.In recent years,the traditional target detection algorithm lags far behind the target detection algorithm using deep learning in the detection accuracy.The convolutional neural network can extract features,efficiency and the results are much stronger than the features of manual design and extraction.This paper proposes four aspects to improve Faster R-CNN: optimized anchor,ELU activation function,simplified VGG network and the region of interest location method.The proposed model not only improves the accuracy of detection but also reduces the detection time,the experiment proves the validity and feasibility of the proposed model,and the proposed method shows excellent detection performance.