Abstract. Based on Machine Vision, the paper research a method to detect obstacles and calculate the distance between the obstacle and moving vehicles. First it uses the shadow detection method to determine approximate location of the obstacle and adopt corner detection method to detect the obstacle contour information. Then, the minimum bounding rectangle method is used to accurately framed obstacles vehicles. Finally, based on the principle of the distance in the camera, the obstacle distance measurement formula is built. Experimental results show that the algorithm can accurately locate obstacles and measure the actual distance between the moving vehicle and the obstacles.

As the location method of indoor fingerprint position needs to be conducted in a region with signal coverage, the signal intensity's information of all Access Points has to be collected to build a fingerprint database. And not all information in the database will be of positive service to the positional accuracy. On account of increasing data dimensions in the database, rising complexity of algorithm and the climbing number of required experimental samples, a dimension disaster could be arisen. Based on KPCA algorithmic method, in this paper, positioning performance will be improved even under noisy circumstance by preprocessing the position fingerprint data and some data space occupied by positioning system will be saved through decreasing the data dimensions and reduced information of redundancy.

Based on the RSSI fingerprint matching positioning technology is widely used in WIFI indoor positioning field, due to the influence of multipath effect, scattering, refraction, staff walking, propagation distance and other aspects, indoor signal propagation is easy to introduce noise into the acquisition signal database, leading to the final positioning result is not too ideal. So, this paper proposes a fingerprint library filtering algorithm based on neighborhood mean filter, introduce the concept of acquisition noise on the basis of traditional algorithm, removes the noise of the original data and reduces the average error of location. The results show the effectiveness of this method, and the improved algorithm makes the positioning accuracy greatly improved.