Press line detection is one of the core functions of road test instrument, but because of complex and diverse, road test environment has been effect is not ideal, because of the traditional algorithms based on color information to detect, so will be seriously affected by the following several parts:
1, the influence of the next steps, especially the steps of color similar to the color line, lead to false identification algorithm,
2, the steps outside dirt or the effects of floor tile,
3, line is incomplete, vague,
4, the influence of the sun's rays, the strong and weak light can create the algorithm the serious interference
5, online there are a lot of dust dirt covered, which leads to the line is not clean
With the traditional color space transformation, hough transform method such as do the road test, the effect is very poor is these reasons,
To this end, on the basis of using a large amount of measured data, I chose a more robust lane line features as the basis of pressure test, and then use data analysis to determine the pressure detection scheme and model of complete effect is ideal, to clean the environment effect is very good, the environment not clean effect is good, the accuracy is much higher than the market products, and is now ready to complete the development environment of Java, c + +, can directly use,
If you have any other scenarios on the model of demand, as specialists in the field of research, so to optimize the model as soon as possible, can also meet the needs of other scenarios,
The whole algorithm process:
Algorithm results show as follows, for these "is considered the" environment, at present the product market effect is poorer, but the environment and a lot of actually exist:
CodePudding user response:
Looks very cow force ah, I use the hough transform, the corrosion coefficient of expansion coefficient is badCodePudding user response:
These it is difficult to do by using the,,, because in many domestic lane line environment variableCodePudding user response:
Can you contact me 1992631313, bull looks good at @qq.com