Object Detection via Feature Synthesis Using MDL-Based Genetic Programming

Abstract

In this paper, we use genetic programming (GP) to synthesize composite operators and composite features from combinations of primitive operations and primitive features for object detection. The motivation for using GP is to overcome the human experts' limitations of focusing only on conventional combinations of primitive image processing operations in the feature synthesis. GP attempts many unconventional combinations that in some cases yield exceptionally good results. To improve the efficiency of GP and prevent its well-known code bloat problem without imposing severe restriction on the GP search, we design a new fitness function based on minimum description length principle to incorporate both the pixel labeling error and the size of a composite operator into the fitness evaluation process. To further improve the efficiency of GP, smart crossover, smart mutation and a public library ideas are incorporated to identify and keep the effective components of composite operators. Our experiments, which are performed on selected training regions of a training image to reduce the training time, show that compared to normal GP, our GP algorithm finds effective composite operators more quickly and the learned composite operators can be applied to the whole training image and other similar testing images. Also, compared to a traditional region-of-interest extraction algorithm, the composite operators learned by GP are more effective and efficient for object detection.

Combinatorial Fusion Criteria for Real-Time Tracking

“…In the feature-based approach, the crucial problem is determining which features remain robust to occlusion. Lin and Bhanu [15] introduce a feature synthesis strategy for target recognition based on genetic programming approach. Compositions of primitive features are learned that produce improve discrimination between target classes as long as the targets are not overlapping or occluded.…”

An Improved Scheme for Target Discrimination in High-Resolution SAR Images

“…However, owing to the complexity of the problem and the abundance of data, even the most experienced experts cannot provide the optimal combination of features according to the concrete discrimination application. The key point of the selection problem is that we need a universal feature-selection method for different discrimination applications [18]. Furthermore, the method should be capable of identifying the best feature subsequences by searching the initial feature set.…”

Polynomial modeling for time-varying systems based on a particle swarm optimization algorithm

“…Genetic programming (GP) [25,26] is a commonly used evolutionary computation method which is used to generate polynomial models for various systems such as chemical plants [38], time series systems [21], nonlinear dynamic systems [56], object classification systems [1,65], machine learning systems [27], feature selection systems [43], object detection systems [37], speech recognition systems [11], control systems [5] and mechatronic systems [61]. The GP starts by creating a random initial population of individuals, each of which represents the structure of a polynomial model.…”